US20120233294A1 - Adaptive web server - Google Patents
Adaptive web server Download PDFInfo
- Publication number
- US20120233294A1 US20120233294A1 US13/473,417 US201213473417A US2012233294A1 US 20120233294 A1 US20120233294 A1 US 20120233294A1 US 201213473417 A US201213473417 A US 201213473417A US 2012233294 A1 US2012233294 A1 US 2012233294A1
- Authority
- US
- United States
- Prior art keywords
- content
- request
- content file
- server
- client
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/61—Network streaming of media packets for supporting one-way streaming services, e.g. Internet radio
- H04L65/612—Network streaming of media packets for supporting one-way streaming services, e.g. Internet radio for unicast
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/80—Responding to QoS
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/50—Indexing scheme relating to G06F9/50
- G06F2209/5013—Request control
Definitions
- the present invention pertains to Internet/Intranet systems. More particularly, this invention relates to an adaptive web server that serves user access requests for the web server to less resource intensive contents in the web server when the web server is overloaded.
- An example of a data access network system is the Internet or Intranet network system.
- An Internet/Intranet network system typically includes a number of data service systems connected together via interconnect networks.
- the data service systems typically include web and/or content servers that host contents for various customers or applications.
- the customers are the owners of the contents hosted in the data service systems such that subscribers or users can access the contents via their computer terminals.
- the content servers typically utilize Internet applications, such as electronic mail, bulletin boards, news groups, and World Wide Web access.
- the hosted contents are arranged in the form of content sites within the content servers. Each site may include a number of pages (e.g., world wide web pages).
- a content site is typically for one customer of the server while a particular customer may own a number of content sites.
- the content servers may also be referred to as web servers.
- a web page is formed from a number of data and/or executable program files, such as text files, graphics files, audio files, video files, and application program files. Each of the files is referred to as an “object”.
- the web pages can be accessed by a user at a user terminal (e.g., a personal computer system or a web access device) connected to any one of the data service systems.
- HTTP Hyper Text Transfer Protocol
- the HTTP protocol is a request-and-response protocol.
- a user at a client device designates a particular web page, at least one request is generated. The number of requests is dependent upon the sophistication of the designated web page.
- a web page may include one or more “objects”.
- a multi-object page is aesthetically pleasing, but each object requires a separate request and a separate response. Therefore, the time for each request-and-response round trip plays a role in determining the total time a user must wait to view the complete web page.
- a web server can be accessed by multiple users at the same time.
- the web server typically handles the user access requests in the first-come-first-served (FIFO) fashion.
- FIFO first-come-first-served
- One disadvantage of this type of prior art web server structure is that the web server is not equipped with protection mechanism against excessive load conditions. As is known, a web server typically has limits on the number of requests it can handle per second. In addition, the web server also has limits on the response time and bandwidth that can be served. Under high loads, the web server can be overwhelmed with requests, resulting in longer response time and poor performance for users. Thus, when the number of user access requests to the web server greatly exceeds the processing capacity of the web server (i.e., overloaded), the web server either handles a request in an unbearably long period of time or simply becomes non responsive. The sharp increase in server response time may also cause user connections to time out, creating the perception that the server is down or crashed. In mission-critical applications, such as e-commerce, perceived server downtime
- One feature of the present invention is to improve performance of a web server.
- Another feature of the present invention is to increase performance of a web server by serving user access requests for the web server from less resource intensive contents in the web server when the web server is to be overloaded.
- a further feature of the present invention is to provide an adaptive web server that, when in overload situation, routes the user access requests to the less resource intensive contents in the web server such that the web server does not become overloaded or non-functional.
- a data service system in a data access network system includes a content server that stores content files for access by external access requests. Each of the content files is stored in a full content format and any number of adapted or degraded content formats which are less resource-intensive to serve than the full content format.
- the data service system also includes an adaptive load control system which is coupled to the content server to pass the access requests to the content server. The adaptive load control system modifies an access request to access the corresponding content file in one of the adapted content formats when the content server is in an overload condition such that the content server can be maintained at safe load conditions.
- a method of maintaining the content server at safe load conditions includes the step of determining load condition of the content server when the data service system receives an access request to access one of the content files stored in the content server. This step can be performed internally in the content server or external to the content server. If the content server is determined to be in an overload condition, then the access request is modified to access the corresponding content file in an adapted content format which is less resource-intensive to serve than the same file in a full content format such that the content server is maintained at the safe load conditions.
- FIG. 1 shows a data access network system that includes a number of data service systems.
- FIG. 2 shows the structure of one of the adaptive data service systems of FIG. 1 in accordance with one embodiment of the present invention.
- FIG. 3 shows the control and feedback loop of the adaptive data service system of FIG. 2 for partial degradation.
- FIG. 4 schematically shows the structure of the content server of the adaptive data service system of FIG. 2 .
- FIGS. 5 and 6 show various comparisons between the adaptive data service system of FIG. 2 and a prior art data service system.
- FIG. 7 shows a Gap Estimation method employed by a load monitor of FIGS. 2-3 .
- FIG. 1 shows a data access network system 10 .
- the network system 10 is an Internet system.
- the network system 10 is an Intranet system.
- the data access network system 10 may be any other known network system that employs a known communication protocol.
- the data service system 20 is connected to a number of user terminals 11 a through 11 n via an interconnect network 12 .
- Users at each of the user terminals 11 a - 11 n can access the data service system 20 for the services provided by the data service system 20 .
- the users at each of the user terminals 11 a - 11 n can also access the data service system 15 for the services provided by the data service system 15 via the data service system 20 and the Internet 13 .
- the user terminals 11 a - 11 n can also be referred to as client devices.
- Each of the user terminals 11 a - 11 n may be located at a residence, a school, or an office of the user.
- Each of the user terminals 11 a - 11 n in includes a network access application program (e.g., a web browser application program such as Netscape's Navigator) that allows the user to access the data services offered by the data service systems 15 and 20 .
- a network access application program e.g., a web browser application program such as Netscape's Navigator
- Each of the user terminals 11 a - 11 n can be a computer system or other electronic device with data processing capabilities (e.g., a web TV).
- the interconnect network 12 can be any known network, such as Ethernet, ISDN (Integrated Services Digital Network), T-1 or T-3 link, FDDI (Fiber Distributed Data Network), cable or wireless LMDS network or telephone line network.
- FIG. 1 shows only two data service systems 15 and 20 for the data access network system 10 for illustration purposes only.
- the data access network system 10 may include many more data service systems.
- the Internet 13 is formed by a number of data service systems connected together via a network. Data communications among all data service systems are conducted using a predetermined communication protocol for Internet/Intranet communications.
- the communication protocol is the Hyper Text Transport Protocol (HTTP).
- HTTP Hyper Text Transport Protocol
- other known communication protocols for Internet/Intranet communications can also be used.
- Each of the data service systems 15 and 20 has substantially the same functional structure, which will be described in more detail below.
- Each of the data service systems 15 and 20 can be employed by an Internet/Intranet Service Provider (ISP) to offer data services (e.g., web, news, or advertisement) and other services (e.g., e-commerce, e-mail) to users or subscribers connected to the data access system 10 via one of the data service systems 15 and 20 .
- ISP Internet/Intranet Service Provider
- Each of the data service systems 15 and 20 includes a number of servers (e.g., web servers, e-mail servers, news servers, e-commerce servers, domain name servers, address assignment servers, proxy servers, advertisement servers, and session manager servers).
- the web servers, e-mail servers, news servers, e-commerce servers, and advertisement servers can be collectively referred to as local service servers or content servers.
- a content server typically stores a number of content files that include Hyper-Text Markup Language (HTML) web pages, gif images, video clips, etc.
- HTML Hyper-Text Markup Language
- Data transfers to and from the content servers are enabled by transport protocols such as Transport Control Protocol (TCP) and the User Datagram Protocol (UDP).
- TCP Transport Control Protocol
- UDP User Datagram Protocol
- the content servers support a variety of Internet applications to provide services such as access to the World Wide Web, electronic mail, bulletin boards, chat rooms, news groups, and e-commerce.
- content files e.g., web pages, news, images, e-mails,
- the access to the content servers is structured around the HTTP protocol which is a request-and-response protocol.
- HTTP protocol is a request-and-response protocol.
- a user at a user terminal desires to access a content file stored in a content server of one of the data service systems 15 and 20 , at least one request is generated and sent to the content server.
- the content server can handle multiple requests at the same time.
- the content server has limits on the number of requests it can handle per second.
- the content server will be in an overload condition, resulting in at least unbearably longer response time and poor performance for users. This means when the request rate increases beyond the server capacity, server performance deteriorates dramatically, potentially causing a service outage.
- each of the data service systems 15 and 20 includes an adaptive load control mechanism to control the load condition of its content servers such that each content server does not become overloaded.
- the load control mechanism of the present invention can also be applied to other servers of each of the data service systems 15 and 20 .
- the adaptive load control mechanism is described in more detail below, also in conjunction with FIGS. 2-6 .
- a data service system. 30 in accordance with one embodiment of the present invention is shown to include a content server 31 .
- the data service system 30 also includes an adaptive load control system 40 that controls the load condition of the content server 31 in accordance with one embodiment of the present invention.
- the data service system 30 can also be referred to as adaptive data service system or adaptive web server.
- the data service system 30 can be any one of the data service systems (e.g., the data service system 15 or 20 ) of FIG. 1 .
- the content server 31 stores content files or dynamic executable code/program for access by access requests.
- the content files hereinafter refer to (1) static content files, (2) dynamic content files, and (3) executable programs/codes.
- Each of the content files is stored in a full content format and an adapted or degraded content format.
- a content file in the adapted content format is much smaller in size and less resource intensive than the same file in the full content format.
- the adaptive load control system 40 passes the access requests to the content server 31 when receiving the requests. When the content server 31 is in an overload condition or about to be in the overload condition, the adaptive load control system 40 modifies the access requests to access their corresponding content files in the adapted content format.
- the adaptive load control system 40 When the adaptive load control system 40 determines that the content server 31 is not in the overload condition, the adaptive load control system 40 modifies the access requests to access their corresponding content files in the full content format. This allows the content server 31 to be maintained at safe load conditions. Alternatively, the adaptive load control system 40 can adapt the content server 31 based not on the load condition of the content server 31 , but on the client's abilities (e.g., network connection of the client device, latency, or display capability of the client device). The data service system 30 will be described in more detail below, also in connection with FIGS. 2-6 .
- the data service system 30 can be implemented in a computer system or other data processing system,
- the computer system that implements the data service system 30 can be server computer system, a workstation computer system, a personal computer system, or a mainframe computer system, a notebook computer system, or any other computer system.
- the data service system 30 includes a network interface 35 that interfaces the adaptive load control system 40 .
- the adaptive load control system 40 then interfaces with the content server 31 .
- the network interface 35 serves as the interface of the data service system 30 with external network (not shown).
- the interface 35 can be implemented by any known network interface technology, and thus will not be described in more detail below.
- the network interface 35 receives external requests for the content server 31 and passes the requests to the content server 31 via the adaptive load control system 40 .
- the content server 31 can be any kind of content server that stores a number of content files. Each of the content files can be accessed by an access request.
- the content server 31 may also include a number of content sites, each storing a number of content files for access by multiple access requests. The multiple content sites may belong to different content providers or customers with potential conflicting interests.
- the adaptive load control system 40 causes the content server 31 to be a number of performance-isolated virtual content servers, which will be described in more detail below.
- Each of the servers has a capability to provide independent performance guarantees regardless of the load on the other virtual servers.
- the content server 31 can be a static server or dynamic server.
- the content server 31 is a static content server that stores static content files only.
- the content server 31 may store both static and dynamic content files.
- web content is generally classified as static, such as a file, or dynamic, such as cgi-scripts.
- Dynamic content may be generated at run-time by a back-end engine (e.g., a database engine) that is separate from the server itself. Any adaption at the back-end engine is not handled by the adaptive load control system 40 .
- each of the content files stored in the content server 31 is stored in multiple formats or versions.
- a content file can be stored in its original full content format or version.
- the same content file is also stored in a degraded or adapted format or version in the content server 31 .
- FIG. 4 shows an example of such arrangement.
- the content file in the degraded format 72 is smaller in size and worse in viewing quality than the same file in the full content format or version 71 . This means that the degraded format of the content file requires less resource (e.g., bandwidth) to serve than the counterpart full content format of the same content file.
- a content file can be shown, for example, at 64 ⁇ 64 pixel resolution, at 128 ⁇ 128 pixel resolution, at 256 ⁇ 256 pixel resolution, or at higher or lower resolutions.
- the 256 ⁇ 256 pixel resolution version of a content file clearly has more image data and requires more resource to transfer than the 128 ⁇ 128 pixel resolution version of the same content file.
- the degraded version of the same content file can be at 128 ⁇ 128 pixel resolution, or at 64 ⁇ 64 pixel resolution.
- the degraded format of a content file may include images of less resolution, fewer embedded images, simplified or simpler pages without backgrounds or fewer hyper-links than the full content format of the same content file.
- each of the content files stored in the content server 31 has several degraded versions of the same content file.
- the content server 31 will be described in more detail below with only two versions (i.e., the full content version and one degraded version), for example, for each of the content files stored in the content server 31 for illustration purposes only.
- each access request specifies a content file stored in the content server 31 using the Universal Resource Locator (URL).
- the URL path for a content file stored in the content server 31 does not specify which version or format from which the content file will be accessed. This means that multiple content trees contain different versions of the same URL.
- the path to a particular URL in a given content tree will be the concatenation of the content tree name and the URL name, prefixed by the name of the root service directory of the content server 31 . For example, two content trees may be created in the root service directory “/root”, one being “/full_content” and the other “/degraded_content”.
- the “/full_content” content tree name directs the request to access the corresponding content file in the full content version or format.
- the “/degraded_content” content tree name directs the request to access the corresponding content file in the degraded content version or format.
- a URL of “http://www.hpl.hp.com/my_picture.jpg” may be served from either the directory “root/full_content/my_picture.jpg”, or from the directory “root/degraded_content/my_picture.jpg”.
- the scheme applies even if the content file is dynamic content, such as the content generated by cgi-scripts.
- Multiple content trees may contain different versions or formats of the named cgi-scripts (e.g., “my_script.cgi”).
- the “/degraded_content/” version or format is a less resource-intensive search script that, for example, looks only for the first five matches, instead of the “/full_content” version or format that looks for one hundred matches.
- the URL of the script is then prefixed by the right tree name to determine which version of the script to execute under given load conditions.
- the URL of “http://www.hpl.hp.com/my_script.cgi” may be modified to become either “http://www.hpl.hp.com/root/full_content/cgi-bin/my_script.cgi” or “http://www.hpl.hp.com/root/degraded_content/cgi-bi/my_script.cgi”.
- the less resource-intensive “/degraded_content” search script can be replaced or substituted with a static version or format, This can be done by switching to a different content tree.
- an on-line vendor using the content server 31 can use a dynamically generated version of their product catalog that interacts with a stock database to display the items currently in stock.
- the content server 31 is overloaded, its statically pre-stored catalog version is then accessed and displayed. The reason for doing this is that a dynamic script, even degraded or simplified, still consumes more resources to serve than the static content.
- the adaptive load control system 40 is connected to the content server 31 .
- the adaptive load control system 40 detects overload condition of the content server 31 and causes the access requests to access the corresponding file from one of several alternative content qualities for delivery in accordance with the load conditions.
- the adaptive load control system 40 causes the incoming access requests to the content server 31 to access the corresponding content files in the degraded format which is less resource-intensive to serve.
- the degraded format reduces the size of responses.
- the degraded format also reduces the total number of requests as the simplified content includes fewer embedded images and/or other objects.
- the degraded format also removes or reduces links embedded in the requested content file. Thus, the degraded format is less resource intensive to serve than the full content format.
- the adaptive load control system 40 is also able to reject access requests if needed.
- the adaptive load control system 40 supports co-existence of multiple performance-isolated virtual servers such that performance levels and load adaption decisions of virtual server do not affect that of another virtual server. This is required when the content server 31 hosts multiple content sites that possibly belong to different customers or content providers with potentially conflicting interests. The total load on the server 31 may not be evenly distributed among the hosted sites.
- the adaptive load control system 40 implements a performance isolation mechanism that does not penalize a site for excessive load incurred by another site. Using this mechanism, the content server 31 is viewed as a collection of one or more performance-isolated virtual servers, each with a capability to provide independent performance guarantees regardless of the load on other virtual servers, if any. Guarantees are declared in a service level agreement associated with each virtual server and are used to configure the adaptive load control system 40 for appropriate resource capacity allocation.
- the load adaptive control system 40 makes adaption decisions within each virtual server of the content server 31 on its load and resource allocation.
- the adaptive load control system 40 also gives preferential treatment to guaranteed service classes, while allowing non-guaranteed service classes to consume the excess capability when available.
- the guaranteed service classes are governed by service level agreements.
- the non-guaranteed classes are typically served on best effort basis with available resources.
- the adaptive load control system 40 should direct non-guaranteed requests to their degraded content files first.
- An arbitrary policy can govern the relative degradation of client classes with different importance levels. For example, under overload conditions, the class A requests should be sent to the degraded content files before class B requests. But if resources remain scarce, class B requests are then degraded before rejecting class A requests.
- the adaptive load control system 40 can be implemented at any point in the data service system 30 where the server's requests and responses can be accessed. This means that the adaptive load control system 40 can be implemented in the web server software, in the UNIX socket library, or in the operating system of a computer system that embodies the data service system 30 . Moreover, the adaptive load control system 40 can also be implemented in the gateway that is external to the server, as well as transparent to the server.
- the adaptive load control system 40 requires three entry points. They are (1) initialization point, (2) request pre-processing point, and (3) request post-processing point.
- the initialization point initializes internally used data structures, and creates a separate process that implements various components of the adaptive load control system 40 if desired.
- the request pre-processing point decides the quality of content level of the given request.
- the content adaptor 41 of the adaptive load control system 40 is called from this point.
- the content adaptor 41 may communicate with the other components via shared memory.
- the request post-processing point is an optional point which monitors response size.
- the adaptive load control system 40 can be implemented in a middleware (e.g., a dynamically linked socket library) used by the server 31 .
- a middleware e.g., a dynamically linked socket library
- the server socket is created and a listen( ) call is made to the socket library to listen for incoming requests.
- This library routine can be modified to call the initialization point of the middleware. The modification can be transparent to the content server 31 .
- the functional structure of the adaptive load control system 40 is described in more detail below, also in conjunction with FIGS. 2-7 .
- the adaptive load control system 40 includes a request classifier 33 connected to the network interface 35 .
- the adaptive load control system 40 also includes a load monitor 32 connected to the request classifier 33 .
- a content adapter 41 is then connected to the load monitor 32 and the content server 31 .
- An adaption controller 50 is connected to the content adapter 41 and the load monitor 32 .
- the components 32 - 33 , 41 , and 50 together form the adaptive load control system 40 .
- the adaptive load control system 40 may include more or fewer components than above mentioned.
- the adaptive load control system 40 may function without the request classifier 33 .
- the request classifier 33 receives the access requests from the network interface 35 , The request classifier 33 is used to classify the received access requests into a number of classes based on different classification metrics.
- the request classifier 33 allows preferential treatment to a subset of content sites or clients.
- the request classifier 33 is used in an installation where quality of service differentiation is required among different classes of clients (e.g., guaranteed and non-guaranteed clients, or clients accessing different virtual servers). In this situation, the action taken by the content adapter 41 later may depend on the identity of the client or the requested content.
- the request classifier 33 allows multiple classes of users to share the same content file or site and yet receive different treatments or performance.
- Class-based service is a mechanism for differentiating services given to individual classes. Thus, service performance can be priced based on performance or service agreements. A higher class with greater guarantee can be priced higher than a lower class that may offer less guarantee and more “best effort” services.
- the request classifier 33 can be implemented using any known technology.
- the request classifier 33 then sends the classified access requests to the load monitor 32 .
- the load monitor 32 is used to monitor the load condition of the content server 31 in order to detect overload condition.
- the load monitor 32 employs a number of monitoring mechanisms to monitor the load condition of the content server 31 .
- One way of monitoring the load condition of the content server 31 is to monitor the response time of the content server 31 . This is because response time is proportional to the length of the input request queue of the content server 31 (e.g., the server's socket listen queue in a UNIX implementation). When the server is wider-loaded, the queue tends to be short (or empty), resulting in short response time. When the content server 31 is overloaded, the request queue within the content server 31 overflows, making the response time grow an order of magnitude. This serves as a clear overload indicator.
- the advantage of this approach i.e., estimating queue length by monitoring response time
- the mechanism can be implemented outside the server process and requires no modification to the code of the content server 31 .
- Another way of monitoring whether the content server 31 is overloaded is to monitor utilization of the server machine, as well as the response time. This allows the load monitor 32 to determine how under-utilized the server is. When utilization decreases below a predetermined level, load monitor 32 may determine the content server 31 is no longer in the overload condition.
- utilization measurement should be indicative of the load attributed to useful work (i.e., serving client requests), rather than the aggregated utilization. For example, the existence of a low priority process or thread that implements a busy-waiting loop on some event will render the utilization identically 100% even in the absence of load on the server. Another concern arises when all server threads or processes have been committed to current requests and are blocked in I/O. The resulting idle time cannot be taken advantage of since all resources are already utilized and therefore should not be measured towards available capacity.
- the load monitor 32 employs a linear function of measured request rate R and delivered bandwidth BW to determine the system utilization U, consumed by processing client requests.
- the linear function is shown as follows.
- constants a and b can be computed by either on-line or off-line priori profiling as will be described below.
- the function is good for estimating offered load as long as it does not exceed server capacity.
- the linear approximation method is invalid.
- the linear approximation method is combined with the response time monitoring to determine the actual load condition of the server 31 .
- each server process or thread Tj records independently its observed request rate Rj, delivered bandwidth BWj, and utilization Uj.
- a separate monitor process or thread then reads and sums up the recorded values to compute the aggregate request rate R, bandwidth BW, and utilization U. Synchronization of access to shared data structures is not required since there is only one writer to any piece of data.
- the method allows server capacity planning for quality of service guarantees, and it provides means for converting the desired request rate and bandwidth guarantee into a corresponding resource capacity allocation (i.e., allocated utilization). This method, however, requires computing the constants a and b using the priori profiling.
- R and BW can be measured online by counting requests seen and bytes delivered within a given period. Utilization U can be measured using a Gap Estimation method which will be described later. Given the measured values of R, BW, and U, at successive time intervals, estimation theory provides a way to find a linear fit that minimizes the error. It provides the necessary formulae for computing and updating the parameters a and b online in view of successive new measurements.
- the parameters a and b can be determined by testing the content server 31 with a pre-specified workload. Testing can proceed by requesting a URL of a given size at an increasing rate until client connections start timing out, indicating server overload. The maximum attained request rate and bandwidth are then recorded. For our purposes, server utilization can be assumed to be 100% at this load condition. The experiment is repeated for different sizes of the requested URL, giving a different rate and bandwidth combination that saturates the server 31 .
- the second mechanism for the load monitor 32 is referred to as the Gap Estimation Method.
- This method estimates the fraction of time that the content server 31 spends serving requests.
- the load monitor 32 employs a global counter to track the requests.
- the load monitor 32 increments the counter when a request is received and decrements the counter when a response departs the content server 31 .
- the counter will return to its initial value only when a gap is present (i.e., when all current requests have been served and no additional requests have arrived yet).
- a monitor process or thread can compute the total “idle time” G within that time period T.
- the idle time is the time where no requests are pending.
- the Gap Estimation method is illustrated in FIG. 7 .
- concurrent processing of request bursts are separated with idle time.
- the method does not require a prior profiling, and does not require monitoring response time.
- the global counter may be updated by multiple writers and thus requires some form of locking or access synchronization.
- the load monitor 32 sends the load condition information of the content server 31 to the content adapter 41 and the adaption controller 50 .
- the adaption controller 50 also receives a predetermined desired load value which indicates the threshold of the overload condition. Based on the comparison of the desired load value and the monitored load value, the adaptive controller 50 determines whether the content server 31 is in the overload condition in which case the adaptive controller 50 causes the content adapter 41 to modify the URLs of all the incoming access requests to access their corresponding content files in the degraded content format or version. If the adaptive controller 50 determines that the content server 31 is not in the overload condition in which case the adaptive controller 50 causes the content adapter 41 to modify the URLs of the incoming access requests to access their corresponding content files in the full content format or version.
- the desired load value for the adaption controller 50 can be a predetermined threshold T of the server response time.
- This threshold T can be set equal to (or slightly smaller than) the maximum server response time specified in the service level agreement, if any, thereby causing adaption when the agreement is about to be violated.
- the content adapter 41 effects adaption by changing directory links.
- the content adapter 41 modifies the URLs of the incoming access requests to access the corresponding content files in either the full content format or one of the degraded content formats.
- the content adapter 41 can modify the URL of “http://www.hpl.hp.com/my_picture.jpg” as either “http://www.hpl.hp.com/full_content/my_picture.jpg” to access the content file “my_picture.jpg” in the full content format, or “http://www.hpl.hp.com/degraded_content/my_picture.jpg” to access the content file “my_picture.jpg” in the degraded content format.
- the content adapter 41 can modify the URL as either “http://www.hpl.hp.com/full_content/cgi-bin/my_script.cgi” to access the full script file, or “http:www.hpl.hp.com/degraded_content/cgi-bin/my_script.cgi” to access the less resource-intensive script.
- the adaption controller 50 determines that the content server 31 is in the overload condition, the adaption controller 50 causes the content adapter 41 to modify the URLs of the incoming access requests to access the corresponding content files in the degraded or adapted content format. As described above, the degraded or adapted content format requires less resources to serve than the full content format. When the adaption controller 50 determines that the content server 31 is not in the overload condition, the adaption controller 50 causes the content adapter 41 to modify the URLs of the incoming access requests to access the corresponding content files in the full content format.
- the content adapter 41 switches transparently between the high quality service tree 73 ( FIG. 4 ) and the degraded quality service tree 74 ( FIG. 4 ) in the root service directory 70 ( FIG. 4 ) based on the load condition of the content server 31 .
- the switching operation is completely transparent to the content server 31 .
- FIG. 5 shows the result of the adaption process by the adaptive load control system 40 .
- the curve 81 shows the percentage of connection failures by a prior art server without the adaptive load control system 40 .
- the curve 80 shows the percentage of connection failures by the content server 31 with the adaptive load control system 40 .
- the curve 80 shows significantly fewer connection failures than the prior art curve 81 as the request rate increases.
- the traditional server suffers an increasing error rate (i.e., connection failures) when offered load exceeds capacity of the server.
- the controller 50 determines that the content server 31 is overloaded, the controller 50 causes the content adapter 41 to adapt all incoming access requests to the degraded content format. This potentially makes the content server 31 under-loaded, thus not achieving adequate resource utilization when the adaption takes place.
- FIG. 6 the solid line curve 85 represents the delivered bandwidth of a prior art content server without the adaption mechanism in accordance with one embodiment of the present invention and the broken line curve 86 represents the delivered bandwidth of the content server 31 with the adaptive load control system 40 .
- a sharp drop in delivered bandwidth for the curve 86 is shown for the content server 31 when adaption takes place by switching to less resource-intensive content format. This leaves server bandwidth underutilized.
- the fraction can be between nothing and the all of the incoming access requests.
- This algorithm provides a continuous partial degradation spectrum that ranges between serving all requests from the highest quality content tree and rejecting all requests.
- a self-regulating technique will be described, in conjunction with FIG. 3 , that uses monitoring and feedback to automatically select the best value of G such that the overload is prevented and a specified target utilization is maintained.
- FIG. 3 shows the control and feedback loop of the adaptive load control system 40 of FIG. 2 for self-regulating partial degradation in accordance with another embodiment of the present invention.
- a partial degradation module 60 is used to control the partial degradation.
- the module 60 includes a integral controller 60 and an adder 61 .
- the achieved server utilization from the load monitor 32 is fed to the adder 61 of the module 60 .
- the desired load utilization is also fed to the adder 61 .
- the end result E i.e., the difference between the two utilizations
- the integral controller 62 then adjusts the control parameter G to control the content adapter 41 to adjust the number of access requests for adaption.
- the content adapter 41 can quickly and automatically adjust the number of degraded requests to be around the target or desired utilization.
Abstract
A data service system in a data access network system is described. The data service system includes a content server that stores content files for access by external access requests. Each of the content files is stored in a full content format and an adapted or degraded content format which is less resource-intensive to serve than the full content format. The data service system also includes an adaptive load control system which is coupled to the content server to pass the access requests to the content server. The adaptive load control system modifies an access request to access the corresponding content file in the adapted content format when the content server is in an overload condition such that the content server can be maintained at safe load conditions. A method of maintaining the content server at safe load conditions is also described.
Description
- 1. Field of the Invention
- The present invention pertains to Internet/Intranet systems. More particularly, this invention relates to an adaptive web server that serves user access requests for the web server to less resource intensive contents in the web server when the web server is overloaded.
- 2. Description of the Related Art
- An example of a data access network system is the Internet or Intranet network system. An Internet/Intranet network system typically includes a number of data service systems connected together via interconnect networks. The data service systems typically include web and/or content servers that host contents for various customers or applications. The customers are the owners of the contents hosted in the data service systems such that subscribers or users can access the contents via their computer terminals. The content servers typically utilize Internet applications, such as electronic mail, bulletin boards, news groups, and World Wide Web access. The hosted contents are arranged in the form of content sites within the content servers. Each site may include a number of pages (e.g., world wide web pages). A content site is typically for one customer of the server while a particular customer may own a number of content sites. The content servers may also be referred to as web servers.
- Often, a web page is formed from a number of data and/or executable program files, such as text files, graphics files, audio files, video files, and application program files. Each of the files is referred to as an “object”. The web pages can be accessed by a user at a user terminal (e.g., a personal computer system or a web access device) connected to any one of the data service systems.
- As is known, access to the web pages via Internet is typically structured around the HTTP (Hyper Text Transfer Protocol) protocol. The HTTP protocol is a request-and-response protocol. When a user at a client device designates a particular web page, at least one request is generated. The number of requests is dependent upon the sophistication of the designated web page. As described above, a web page may include one or more “objects”. A multi-object page is aesthetically pleasing, but each object requires a separate request and a separate response. Therefore, the time for each request-and-response round trip plays a role in determining the total time a user must wait to view the complete web page.
- A web server can be accessed by multiple users at the same time. The web server typically handles the user access requests in the first-come-first-served (FIFO) fashion. One disadvantage of this type of prior art web server structure is that the web server is not equipped with protection mechanism against excessive load conditions. As is known, a web server typically has limits on the number of requests it can handle per second. In addition, the web server also has limits on the response time and bandwidth that can be served. Under high loads, the web server can be overwhelmed with requests, resulting in longer response time and poor performance for users. Thus, when the number of user access requests to the web server greatly exceeds the processing capacity of the web server (i.e., overloaded), the web server either handles a request in an unbearably long period of time or simply becomes non responsive. The sharp increase in server response time may also cause user connections to time out, creating the perception that the server is down or crashed. In mission-critical applications, such as e-commerce, perceived server downtime may result in loss of sales and/or other financial losses.
- One feature of the present invention is to improve performance of a web server.
- Another feature of the present invention is to increase performance of a web server by serving user access requests for the web server from less resource intensive contents in the web server when the web server is to be overloaded.
- A further feature of the present invention is to provide an adaptive web server that, when in overload situation, routes the user access requests to the less resource intensive contents in the web server such that the web server does not become overloaded or non-functional.
- A data service system in a data access network system includes a content server that stores content files for access by external access requests. Each of the content files is stored in a full content format and any number of adapted or degraded content formats which are less resource-intensive to serve than the full content format. The data service system also includes an adaptive load control system which is coupled to the content server to pass the access requests to the content server. The adaptive load control system modifies an access request to access the corresponding content file in one of the adapted content formats when the content server is in an overload condition such that the content server can be maintained at safe load conditions.
- In a data service system of a data access network system having a content server that stores content files for access by external access requests, a method of maintaining the content server at safe load conditions includes the step of determining load condition of the content server when the data service system receives an access request to access one of the content files stored in the content server. This step can be performed internally in the content server or external to the content server. If the content server is determined to be in an overload condition, then the access request is modified to access the corresponding content file in an adapted content format which is less resource-intensive to serve than the same file in a full content format such that the content server is maintained at the safe load conditions.
- Other features and advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, illustrating by way of example the principles of the invention.
-
FIG. 1 shows a data access network system that includes a number of data service systems. -
FIG. 2 shows the structure of one of the adaptive data service systems ofFIG. 1 in accordance with one embodiment of the present invention. -
FIG. 3 shows the control and feedback loop of the adaptive data service system ofFIG. 2 for partial degradation. -
FIG. 4 schematically shows the structure of the content server of the adaptive data service system ofFIG. 2 . -
FIGS. 5 and 6 show various comparisons between the adaptive data service system ofFIG. 2 and a prior art data service system. -
FIG. 7 shows a Gap Estimation method employed by a load monitor ofFIGS. 2-3 . -
FIG. 1 shows a dataaccess network system 10. In one embodiment, thenetwork system 10 is an Internet system. In another embodiment, thenetwork system 10 is an Intranet system. Alternatively, the dataaccess network system 10 may be any other known network system that employs a known communication protocol. - As can be seen from
FIG. 1 , thedata service system 20 is connected to a number ofuser terminals 11 a through 11 n via aninterconnect network 12. Users at each of the user terminals 11 a-11 n can access thedata service system 20 for the services provided by thedata service system 20. The users at each of the user terminals 11 a-11 n can also access thedata service system 15 for the services provided by thedata service system 15 via thedata service system 20 and the Internet 13. The user terminals 11 a-11 n can also be referred to as client devices. - Each of the user terminals 11 a-11 n may be located at a residence, a school, or an office of the user. Each of the user terminals 11 a-11 n in includes a network access application program (e.g., a web browser application program such as Netscape's Navigator) that allows the user to access the data services offered by the
data service systems - Each of the user terminals 11 a-11 n can be a computer system or other electronic device with data processing capabilities (e.g., a web TV). The
interconnect network 12 can be any known network, such as Ethernet, ISDN (Integrated Services Digital Network), T-1 or T-3 link, FDDI (Fiber Distributed Data Network), cable or wireless LMDS network or telephone line network. -
FIG. 1 shows only twodata service systems access network system 10 for illustration purposes only. In practice, the dataaccess network system 10 may include many more data service systems. In addition, theInternet 13 is formed by a number of data service systems connected together via a network. Data communications among all data service systems are conducted using a predetermined communication protocol for Internet/Intranet communications. In one embodiment, the communication protocol is the Hyper Text Transport Protocol (HTTP). Alternatively, other known communication protocols for Internet/Intranet communications can also be used. - Each of the
data service systems data service systems data access system 10 via one of thedata service systems - Each of the
data service systems - As described above, the access to the content servers is structured around the HTTP protocol which is a request-and-response protocol. When a user at a user terminal desires to access a content file stored in a content server of one of the
data service systems data service systems data service systems FIGS. 2-6 . - As can be seen from
FIG. 2 , a data service system. 30 in accordance with one embodiment of the present invention is shown to include acontent server 31. Thedata service system 30 also includes an adaptiveload control system 40 that controls the load condition of thecontent server 31 in accordance with one embodiment of the present invention. Thus, thedata service system 30 can also be referred to as adaptive data service system or adaptive web server. Thedata service system 30 can be any one of the data service systems (e.g., thedata service system 15 or 20) ofFIG. 1 . - As will be described in more detail below, the
content server 31 stores content files or dynamic executable code/program for access by access requests. Thus, the content files hereinafter refer to (1) static content files, (2) dynamic content files, and (3) executable programs/codes. Each of the content files is stored in a full content format and an adapted or degraded content format. A content file in the adapted content format is much smaller in size and less resource intensive than the same file in the full content format. The adaptiveload control system 40 passes the access requests to thecontent server 31 when receiving the requests. When thecontent server 31 is in an overload condition or about to be in the overload condition, the adaptiveload control system 40 modifies the access requests to access their corresponding content files in the adapted content format. When the adaptiveload control system 40 determines that thecontent server 31 is not in the overload condition, the adaptiveload control system 40 modifies the access requests to access their corresponding content files in the full content format. This allows thecontent server 31 to be maintained at safe load conditions. Alternatively, the adaptiveload control system 40 can adapt thecontent server 31 based not on the load condition of thecontent server 31, but on the client's abilities (e.g., network connection of the client device, latency, or display capability of the client device). Thedata service system 30 will be described in more detail below, also in connection withFIGS. 2-6 . - The
data service system 30 can be implemented in a computer system or other data processing system, The computer system that implements thedata service system 30 can be server computer system, a workstation computer system, a personal computer system, or a mainframe computer system, a notebook computer system, or any other computer system. - The
data service system 30 includes anetwork interface 35 that interfaces the adaptiveload control system 40. The adaptiveload control system 40 then interfaces with thecontent server 31. Thenetwork interface 35 serves as the interface of thedata service system 30 with external network (not shown). Theinterface 35 can be implemented by any known network interface technology, and thus will not be described in more detail below. Thenetwork interface 35 receives external requests for thecontent server 31 and passes the requests to thecontent server 31 via the adaptiveload control system 40. - The
content server 31 can be any kind of content server that stores a number of content files. Each of the content files can be accessed by an access request. Thecontent server 31 may also include a number of content sites, each storing a number of content files for access by multiple access requests. The multiple content sites may belong to different content providers or customers with potential conflicting interests. In this case, the adaptiveload control system 40 causes thecontent server 31 to be a number of performance-isolated virtual content servers, which will be described in more detail below. Each of the servers has a capability to provide independent performance guarantees regardless of the load on the other virtual servers. - The
content server 31 can be a static server or dynamic server. In one embodiment, thecontent server 31 is a static content server that stores static content files only. In another embodiment, thecontent server 31 may store both static and dynamic content files. As is known, web content is generally classified as static, such as a file, or dynamic, such as cgi-scripts. Dynamic content may be generated at run-time by a back-end engine (e.g., a database engine) that is separate from the server itself. Any adaption at the back-end engine is not handled by the adaptiveload control system 40. - In accordance with one embodiment of the present invention, each of the content files stored in the
content server 31 is stored in multiple formats or versions. For example, a content file can be stored in its original full content format or version. The same content file is also stored in a degraded or adapted format or version in thecontent server 31.FIG. 4 shows an example of such arrangement. As can be seen fromFIG. 4 , the content file in thedegraded format 72 is smaller in size and worse in viewing quality than the same file in the full content format orversion 71. This means that the degraded format of the content file requires less resource (e.g., bandwidth) to serve than the counterpart full content format of the same content file. This is possible because a content file can be shown, for example, at 64×64 pixel resolution, at 128×128 pixel resolution, at 256×256 pixel resolution, or at higher or lower resolutions. The 256×256 pixel resolution version of a content file clearly has more image data and requires more resource to transfer than the 128×128 pixel resolution version of the same content file. Thus, if the full content version of a content file is at the 256×256 pixel resolution, the degraded version of the same content file can be at 128×128 pixel resolution, or at 64×64 pixel resolution. In summary, the degraded format of a content file may include images of less resolution, fewer embedded images, simplified or simpler pages without backgrounds or fewer hyper-links than the full content format of the same content file. Alternatively, each of the content files stored in thecontent server 31 has several degraded versions of the same content file. Thecontent server 31 will be described in more detail below with only two versions (i.e., the full content version and one degraded version), for example, for each of the content files stored in thecontent server 31 for illustration purposes only. - Referring again to
FIG. 2 , each access request specifies a content file stored in thecontent server 31 using the Universal Resource Locator (URL). The URL path for a content file stored in thecontent server 31, however, does not specify which version or format from which the content file will be accessed. This means that multiple content trees contain different versions of the same URL. The path to a particular URL in a given content tree will be the concatenation of the content tree name and the URL name, prefixed by the name of the root service directory of thecontent server 31. For example, two content trees may be created in the root service directory “/root”, one being “/full_content” and the other “/degraded_content”. The “/full_content” content tree name directs the request to access the corresponding content file in the full content version or format. The “/degraded_content” content tree name directs the request to access the corresponding content file in the degraded content version or format. In this case, a URL of “http://www.hpl.hp.com/my_picture.jpg” may be served from either the directory “root/full_content/my_picture.jpg”, or from the directory “root/degraded_content/my_picture.jpg”. - The scheme applies even if the content file is dynamic content, such as the content generated by cgi-scripts. Multiple content trees may contain different versions or formats of the named cgi-scripts (e.g., “my_script.cgi”). The “/degraded_content/” version or format is a less resource-intensive search script that, for example, looks only for the first five matches, instead of the “/full_content” version or format that looks for one hundred matches. The URL of the script is then prefixed by the right tree name to determine which version of the script to execute under given load conditions. For example, the URL of “http://www.hpl.hp.com/my_script.cgi” may be modified to become either “http://www.hpl.hp.com/root/full_content/cgi-bin/my_script.cgi” or “http://www.hpl.hp.com/root/degraded_content/cgi-bi/my_script.cgi”.
- Alternatively, the less resource-intensive “/degraded_content” search script can be replaced or substituted with a static version or format, This can be done by switching to a different content tree. For example, an on-line vendor using the
content server 31 can use a dynamically generated version of their product catalog that interacts with a stock database to display the items currently in stock. When thecontent server 31 is overloaded, its statically pre-stored catalog version is then accessed and displayed. The reason for doing this is that a dynamic script, even degraded or simplified, still consumes more resources to serve than the static content. - The adaptive
load control system 40 is connected to thecontent server 31. In accordance with one embodiment of the present invention, the adaptiveload control system 40 detects overload condition of thecontent server 31 and causes the access requests to access the corresponding file from one of several alternative content qualities for delivery in accordance with the load conditions. When the adaptiveload control system 40 detects that thecontent server 31 is in the overload conditions, the adaptiveload control system 40 causes the incoming access requests to thecontent server 31 to access the corresponding content files in the degraded format which is less resource-intensive to serve. The degraded format reduces the size of responses. The degraded format also reduces the total number of requests as the simplified content includes fewer embedded images and/or other objects. The degraded format also removes or reduces links embedded in the requested content file. Thus, the degraded format is less resource intensive to serve than the full content format. - The adaptive
load control system 40 is also able to reject access requests if needed. In addition, the adaptiveload control system 40 supports co-existence of multiple performance-isolated virtual servers such that performance levels and load adaption decisions of virtual server do not affect that of another virtual server. This is required when thecontent server 31 hosts multiple content sites that possibly belong to different customers or content providers with potentially conflicting interests. The total load on theserver 31 may not be evenly distributed among the hosted sites. In this case, the adaptiveload control system 40 implements a performance isolation mechanism that does not penalize a site for excessive load incurred by another site. Using this mechanism, thecontent server 31 is viewed as a collection of one or more performance-isolated virtual servers, each with a capability to provide independent performance guarantees regardless of the load on other virtual servers, if any. Guarantees are declared in a service level agreement associated with each virtual server and are used to configure the adaptiveload control system 40 for appropriate resource capacity allocation. The loadadaptive control system 40 makes adaption decisions within each virtual server of thecontent server 31 on its load and resource allocation. - Moreover, the adaptive
load control system 40 also gives preferential treatment to guaranteed service classes, while allowing non-guaranteed service classes to consume the excess capability when available. The guaranteed service classes are governed by service level agreements. The non-guaranteed classes are typically served on best effort basis with available resources. Under overload conditions, the adaptiveload control system 40 should direct non-guaranteed requests to their degraded content files first. An arbitrary policy can govern the relative degradation of client classes with different importance levels. For example, under overload conditions, the class A requests should be sent to the degraded content files before class B requests. But if resources remain scarce, class B requests are then degraded before rejecting class A requests. - The adaptive
load control system 40 can be implemented at any point in thedata service system 30 where the server's requests and responses can be accessed. This means that the adaptiveload control system 40 can be implemented in the web server software, in the UNIX socket library, or in the operating system of a computer system that embodies thedata service system 30. Moreover, the adaptiveload control system 40 can also be implemented in the gateway that is external to the server, as well as transparent to the server. - Basically, the adaptive
load control system 40 requires three entry points. They are (1) initialization point, (2) request pre-processing point, and (3) request post-processing point. The initialization point initializes internally used data structures, and creates a separate process that implements various components of the adaptiveload control system 40 if desired. The request pre-processing point decides the quality of content level of the given request. Thecontent adaptor 41 of the adaptiveload control system 40 is called from this point. Thecontent adaptor 41 may communicate with the other components via shared memory. The request post-processing point is an optional point which monitors response size. - If server code is not available, the adaptive
load control system 40 can be implemented in a middleware (e.g., a dynamically linked socket library) used by theserver 31. in a UNIX environment, as part of server initialization, the server socket is created and a listen( ) call is made to the socket library to listen for incoming requests. This library routine can be modified to call the initialization point of the middleware. The modification can be transparent to thecontent server 31. The functional structure of the adaptiveload control system 40 is described in more detail below, also in conjunction withFIGS. 2-7 . - As can be seen from
FIG. 2 , the adaptiveload control system 40 includes arequest classifier 33 connected to thenetwork interface 35. The adaptiveload control system 40 also includes aload monitor 32 connected to therequest classifier 33. Acontent adapter 41 is then connected to theload monitor 32 and thecontent server 31. Anadaption controller 50 is connected to thecontent adapter 41 and theload monitor 32. The components 32-33, 41, and 50 together form the adaptiveload control system 40. Alternatively, the adaptiveload control system 40 may include more or fewer components than above mentioned. For example, the adaptiveload control system 40 may function without therequest classifier 33. - The
request classifier 33 receives the access requests from thenetwork interface 35, Therequest classifier 33 is used to classify the received access requests into a number of classes based on different classification metrics. Therequest classifier 33 allows preferential treatment to a subset of content sites or clients. Therequest classifier 33 is used in an installation where quality of service differentiation is required among different classes of clients (e.g., guaranteed and non-guaranteed clients, or clients accessing different virtual servers). In this situation, the action taken by thecontent adapter 41 later may depend on the identity of the client or the requested content. Therequest classifier 33 allows multiple classes of users to share the same content file or site and yet receive different treatments or performance. Class-based service is a mechanism for differentiating services given to individual classes. Thus, service performance can be priced based on performance or service agreements. A higher class with greater guarantee can be priced higher than a lower class that may offer less guarantee and more “best effort” services. Therequest classifier 33 can be implemented using any known technology. - The
request classifier 33 then sends the classified access requests to theload monitor 32. The load monitor 32 is used to monitor the load condition of thecontent server 31 in order to detect overload condition. The load monitor 32 employs a number of monitoring mechanisms to monitor the load condition of thecontent server 31. One way of monitoring the load condition of thecontent server 31 is to monitor the response time of thecontent server 31. This is because response time is proportional to the length of the input request queue of the content server 31 (e.g., the server's socket listen queue in a UNIX implementation). When the server is wider-loaded, the queue tends to be short (or empty), resulting in short response time. When thecontent server 31 is overloaded, the request queue within thecontent server 31 overflows, making the response time grow an order of magnitude. This serves as a clear overload indicator. The advantage of this approach (i.e., estimating queue length by monitoring response time) is that the mechanism can be implemented outside the server process and requires no modification to the code of thecontent server 31. - Another way of monitoring whether the
content server 31 is overloaded is to monitor utilization of the server machine, as well as the response time. This allows the load monitor 32 to determine how under-utilized the server is. When utilization decreases below a predetermined level, load monitor 32 may determine thecontent server 31 is no longer in the overload condition. - In this case, in order to provide a faithful server load estimate, utilization measurement should be indicative of the load attributed to useful work (i.e., serving client requests), rather than the aggregated utilization. For example, the existence of a low priority process or thread that implements a busy-waiting loop on some event will render the utilization identically 100% even in the absence of load on the server. Another concern arises when all server threads or processes have been committed to current requests and are blocked in I/O. The resulting idle time cannot be taken advantage of since all resources are already utilized and therefore should not be measured towards available capacity.
- To account for the above mentioned concerns, two different mechanisms are developed for the
load monitor 32. The first is referred to as the Linear Approximation Method. Using this method, the load monitor 32 employs a linear function of measured request rate R and delivered bandwidth BW to determine the system utilization U, consumed by processing client requests. The linear function is shown as follows. -
U=aR+bBW - wherein constants a and b can be computed by either on-line or off-line priori profiling as will be described below. The function is good for estimating offered load as long as it does not exceed server capacity. When measuring in the overload condition, the linear approximation method is invalid. Thus, the linear approximation method is combined with the response time monitoring to determine the actual load condition of the
server 31. This combined load monitoring function means that if the measured response time is above the predetermined overload threshold, U=100%. If the measured response time is below the predetermined overload threshold, U=aR+bBW. - One advantage of this linear approximation method is that it is easy to implement. Using this method, each server process or thread Tj records independently its observed request rate Rj, delivered bandwidth BWj, and utilization Uj. A separate monitor process or thread then reads and sums up the recorded values to compute the aggregate request rate R, bandwidth BW, and utilization U. Synchronization of access to shared data structures is not required since there is only one writer to any piece of data. The method allows server capacity planning for quality of service guarantees, and it provides means for converting the desired request rate and bandwidth guarantee into a corresponding resource capacity allocation (i.e., allocated utilization). This method, however, requires computing the constants a and b using the priori profiling.
- A simple way of computing the constants a and b using the priori profiling is to obtain several measurements of U and the corresponding R and BW. Then a linear regression technique is employed to determine the constants a and b that best fit the equation U=aR+bBW. Note that R and BW can be measured online by counting requests seen and bytes delivered within a given period. Utilization U can be measured using a Gap Estimation method which will be described later. Given the measured values of R, BW, and U, at successive time intervals, estimation theory provides a way to find a linear fit that minimizes the error. It provides the necessary formulae for computing and updating the parameters a and b online in view of successive new measurements.
- Alternatively, the parameters a and b can be determined by testing the
content server 31 with a pre-specified workload. Testing can proceed by requesting a URL of a given size at an increasing rate until client connections start timing out, indicating server overload. The maximum attained request rate and bandwidth are then recorded. For our purposes, server utilization can be assumed to be 100% at this load condition. The experiment is repeated for different sizes of the requested URL, giving a different rate and bandwidth combination that saturates theserver 31. The set of R, BW, and U=100% points are used to construct the line aR+bBW=100 on an R, BW, plane. The line intersects the R and BW axes at 100/a and 100/b respectively, from which a and b are found. - The second mechanism for the load monitor 32 is referred to as the Gap Estimation Method. This method estimates the fraction of time that the
content server 31 spends serving requests. Using this method, the load monitor 32 employs a global counter to track the requests. The load monitor 32 increments the counter when a request is received and decrements the counter when a response departs thecontent server 31. The counter will return to its initial value only when a gap is present (i.e., when all current requests have been served and no additional requests have arrived yet). By summing up the gaps over a period of time T, a monitor process or thread can compute the total “idle time” G within that time period T. The idle time is the time where no requests are pending. The utilization is then estimated as U=(T−G)/T. - The Gap Estimation method is illustrated in
FIG. 7 . As can be seen fromFIG. 7 , concurrent processing of request bursts are separated with idle time. The method does not require a prior profiling, and does not require monitoring response time. The global counter may be updated by multiple writers and thus requires some form of locking or access synchronization. - Referring back to
FIG. 2 , the load monitor 32 sends the load condition information of thecontent server 31 to thecontent adapter 41 and theadaption controller 50. Theadaption controller 50 also receives a predetermined desired load value which indicates the threshold of the overload condition. Based on the comparison of the desired load value and the monitored load value, theadaptive controller 50 determines whether thecontent server 31 is in the overload condition in which case theadaptive controller 50 causes thecontent adapter 41 to modify the URLs of all the incoming access requests to access their corresponding content files in the degraded content format or version. If theadaptive controller 50 determines that thecontent server 31 is not in the overload condition in which case theadaptive controller 50 causes thecontent adapter 41 to modify the URLs of the incoming access requests to access their corresponding content files in the full content format or version. - The desired load value for the
adaption controller 50 can be a predetermined threshold T of the server response time. This threshold T can be set equal to (or slightly smaller than) the maximum server response time specified in the service level agreement, if any, thereby causing adaption when the agreement is about to be violated. In the absence of such specification, the threshold can be derived from the pre-configured maximum input request queue length Q and the average service time S. For example, if we consider a 90% full queue to be an overload indication, the threshold T can be set to T=0.9Q×S. - The
content adapter 41 effects adaption by changing directory links. Under the control of theadaptive controller 50, thecontent adapter 41 modifies the URLs of the incoming access requests to access the corresponding content files in either the full content format or one of the degraded content formats. For example, thecontent adapter 41 can modify the URL of “http://www.hpl.hp.com/my_picture.jpg” as either “http://www.hpl.hp.com/full_content/my_picture.jpg” to access the content file “my_picture.jpg” in the full content format, or “http://www.hpl.hp.com/degraded_content/my_picture.jpg” to access the content file “my_picture.jpg” in the degraded content format. As another example, when the content is a dynamic content (e.g., http://www.hpl.hp.com/my_script.cgi), thecontent adapter 41 can modify the URL as either “http://www.hpl.hp.com/full_content/cgi-bin/my_script.cgi” to access the full script file, or “http:www.hpl.hp.com/degraded_content/cgi-bin/my_script.cgi” to access the less resource-intensive script. - When the
adaption controller 50 determines that thecontent server 31 is in the overload condition, theadaption controller 50 causes thecontent adapter 41 to modify the URLs of the incoming access requests to access the corresponding content files in the degraded or adapted content format. As described above, the degraded or adapted content format requires less resources to serve than the full content format. When theadaption controller 50 determines that thecontent server 31 is not in the overload condition, theadaption controller 50 causes thecontent adapter 41 to modify the URLs of the incoming access requests to access the corresponding content files in the full content format. - As can be seen from
FIGS. 2 and 4 , once the adaption trigger is fired from theadaption controller 50, thecontent adapter 41 switches transparently between the high quality service tree 73 (FIG. 4 ) and the degraded quality service tree 74 (FIG. 4 ) in the root service directory 70 (FIG. 4 ) based on the load condition of thecontent server 31. The switching operation is completely transparent to thecontent server 31. -
FIG. 5 shows the result of the adaption process by the adaptiveload control system 40. Thecurve 81 shows the percentage of connection failures by a prior art server without the adaptiveload control system 40. Thecurve 80 shows the percentage of connection failures by thecontent server 31 with the adaptiveload control system 40. As can be seen fromFIG. 5 , thecurve 80 shows significantly fewer connection failures than theprior art curve 81 as the request rate increases. The traditional server suffers an increasing error rate (i.e., connection failures) when offered load exceeds capacity of the server. - As described above, when the
adaption controller 50 determines that thecontent server 31 is overloaded, thecontroller 50 causes thecontent adapter 41 to adapt all incoming access requests to the degraded content format. This potentially makes thecontent server 31 under-loaded, thus not achieving adequate resource utilization when the adaption takes place. This effect is shown inFIG. 6 . InFIG. 6 , thesolid line curve 85 represents the delivered bandwidth of a prior art content server without the adaption mechanism in accordance with one embodiment of the present invention and thebroken line curve 86 represents the delivered bandwidth of thecontent server 31 with the adaptiveload control system 40. As can be seen fromFIG. 6 , a sharp drop in delivered bandwidth for thecurve 86 is shown for thecontent server 31 when adaption takes place by switching to less resource-intensive content format. This leaves server bandwidth underutilized. - This can be overcome by only causing the
content adapter 41 to degrade a fraction of the incoming access requests when thecontent server 31 is in the overload condition. The fraction can be between nothing and the all of the incoming access requests. This mechanism is implemented in thecontent adapter 41. It uses an input control parameter G which can be thought of as an external control knob that tunes the extent of partial degradation. The extremes are G=Max, where no request is degraded, and G=0, where all incoming requests are degraded. Thus, decreasing G will decrease server load. - Assuming the total number of available content trees are Max and the content trees are numbered from 1 to Max in increasing order of quality (
tree 0 is a special tree that stands for request rejection). Let I be the integral part of G and let F be the fractional part. The following rule is used: -
- If G is an integer (i.e., G=I, F=0), the
content adapter 41 lets every request be served from the tree I. - If G is not an integer, the
content adapter 41 computes a pseudo-random number N (in the range [0,1]) upon the receipt of each request. If N<F, the request is served from tree I+1. Otherwise it is served from tree I.
- If G is an integer (i.e., G=I, F=0), the
- This algorithm provides a continuous partial degradation spectrum that ranges between serving all requests from the highest quality content tree and rejecting all requests. Next, a self-regulating technique will be described, in conjunction with
FIG. 3 , that uses monitoring and feedback to automatically select the best value of G such that the overload is prevented and a specified target utilization is maintained. -
FIG. 3 shows the control and feedback loop of the adaptiveload control system 40 ofFIG. 2 for self-regulating partial degradation in accordance with another embodiment of the present invention. In this embodiment as shown inFIG. 3 , apartial degradation module 60 is used to control the partial degradation. Themodule 60 includes aintegral controller 60 and anadder 61. The achieved server utilization from the load monitor 32 is fed to theadder 61 of themodule 60. The desired load utilization is also fed to theadder 61. The end result E (i.e., the difference between the two utilizations) from theadder 61 is then fed to theintegral controller 62. Theintegral controller 62 then adjusts the control parameter G to control thecontent adapter 41 to adjust the number of access requests for adaption. Using the feedback, thecontent adapter 41 can quickly and automatically adjust the number of degraded requests to be around the target or desired utilization. - In the foregoing specification, the invention has been described with reference to specific embodiments thereof. It will, however, be evident to those skilled in the art that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
Claims (19)
1-15. (canceled)
16. A data service system, comprising:
a content server to receive a request from a client for a content file stored in the content server; and
an adaptive load control system to modify the request for the content file into a request for an adapted content file according to the client.
17. The data service system of claim 16 , wherein the adapted content file is less resource-intensive to serve than the content file.
18. The data service system of claim 16 , wherein the content file is stored along a plurality of adapted content files that are degraded versions of the content file.
19. The data service system of claim 16 , wherein the request for the content file comprises a URL addressing the content file.
20. The data service system of claim 19 , wherein the request for an adapted content file comprises a modified URL addressing the adapted content file instead of the URL addressing the content file.
21. The data service system of claim 19 , wherein to modify the request for the content file into a request for an adapted content file according to the client comprises to modify the request for the content file based on a network connection associated with the client.
22. The data service system of claim 19 , wherein to modify the request for the content file into a request for an adapted content file according to the client comprises to modify the request for the content file based on a latency associated with the client.
23. The data service system of claim 19 , wherein to modify the request for the content file into a request for an adapted content file according to the client comprises to modify the request for the content file based on a display capability of the client.
24. A method implemented in a data service system having a content server, the method comprising:
receiving a request from a client for a content file stored in the content server;
determining how to serve the request based on the client and a service class associated with the client; and
modifying the request for the content file into a request for an adapted content file stored in the content server based on the client and the service class associated with the client.
25. The method of claim 24 , wherein the adapted content file is less resource-intensive to serve than the content file.
26. The method of claim 24 , wherein the content file is stored along a plurality of adapted content files that are degraded versions of the content file.
27. The method of claim 24 , wherein receiving a request from the client for the content file comprises receiving a URL addressing the content file.
28. The method of claim 27 , wherein the request for an adapted content file comprises a modified URL addressing the adapted content file instead of the URL addressing the content file.
29. The method of claim 24 , wherein determining how to serve the request based on the client comprises determining a network connection associated with the client.
30. The method of claim 24 , wherein determining how to serve the request based on the client comprises determining a latency associated with the client.
31. The method of claim 24 , wherein determining how to serve the request based on the client comprises determining a display capability of the client.
32. The method of claim 24 , wherein determining how to serve the request based on a service class associated with the client comprises giving preferential treatment to a guaranteed service class.
33. The method of claim 24 , wherein determining how to serve the request based on a service class associated with the client comprises serving a non-guaranteed service class on a best effort basis with available resources.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/473,417 US20120233294A1 (en) | 1999-04-26 | 2012-05-16 | Adaptive web server |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US09/299,684 US8200837B1 (en) | 1999-04-26 | 1999-04-26 | Method and system for maintaining a content server at safe load conditions |
US13/473,417 US20120233294A1 (en) | 1999-04-26 | 2012-05-16 | Adaptive web server |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/299,684 Continuation US8200837B1 (en) | 1999-04-26 | 1999-04-26 | Method and system for maintaining a content server at safe load conditions |
Publications (1)
Publication Number | Publication Date |
---|---|
US20120233294A1 true US20120233294A1 (en) | 2012-09-13 |
Family
ID=23155811
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/299,684 Expired - Fee Related US8200837B1 (en) | 1999-04-26 | 1999-04-26 | Method and system for maintaining a content server at safe load conditions |
US13/473,417 Abandoned US20120233294A1 (en) | 1999-04-26 | 2012-05-16 | Adaptive web server |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/299,684 Expired - Fee Related US8200837B1 (en) | 1999-04-26 | 1999-04-26 | Method and system for maintaining a content server at safe load conditions |
Country Status (4)
Country | Link |
---|---|
US (2) | US8200837B1 (en) |
EP (1) | EP1049031B1 (en) |
JP (1) | JP2000322366A (en) |
DE (1) | DE69939460D1 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015175174A1 (en) * | 2014-05-13 | 2015-11-19 | Pubmatic, Inc. | Intelligent ad auction and sla compliance techniques |
US9699032B2 (en) * | 2013-10-29 | 2017-07-04 | Virtual Instruments Corporation | Storage area network queue depth profiler |
WO2018004644A1 (en) * | 2016-07-01 | 2018-01-04 | Intel Corporation | Adaptive synching |
US20180146033A1 (en) * | 2016-11-23 | 2018-05-24 | Level 3 Communications, Llc | Overload handling in a content delivery network (cdn) |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7822860B2 (en) | 2001-12-11 | 2010-10-26 | International Business Machines Corporation | Method and apparatus for dynamic reconfiguration of web services infrastructure |
US8312131B2 (en) * | 2002-12-31 | 2012-11-13 | Motorola Mobility Llc | Method and apparatus for linking multimedia content rendered via multiple devices |
US8819181B2 (en) * | 2006-03-17 | 2014-08-26 | Apple Inc. | Adaptable network service access through dynamic request routing |
EP2034730B1 (en) * | 2007-09-06 | 2011-11-30 | Nokia Siemens Networks Oy | Load control for a television distribution system |
JP2010123015A (en) * | 2008-11-21 | 2010-06-03 | Nomura Research Institute Ltd | Web server system |
DE102009009040A1 (en) | 2009-02-16 | 2010-09-02 | Siemens Medical Instruments Pte. Ltd. | Apparatus and method for noise estimation in a binaural hearing aid supply |
US9009296B1 (en) | 2009-08-07 | 2015-04-14 | Google Inc. | System and method of determining latency |
US9363312B2 (en) * | 2010-07-28 | 2016-06-07 | International Business Machines Corporation | Transparent header modification for reducing serving load based on current and projected usage |
US9009251B1 (en) * | 2010-07-28 | 2015-04-14 | Cellco Partnership | Server content adjusted based on automated report of available client bandwidth |
US8392558B1 (en) * | 2011-03-22 | 2013-03-05 | Amazon Technologies, Inc. | System and method for determining overload state for service requests |
WO2013071144A1 (en) * | 2011-11-09 | 2013-05-16 | Movable Ink | Management of dynamic email content |
JP6024296B2 (en) * | 2012-08-30 | 2016-11-16 | 富士通株式会社 | Information processing apparatus, copy control program, and copy control method |
US10389792B2 (en) | 2013-02-12 | 2019-08-20 | Nippon Telegraph And Telephone Corporation | Output function dividing system |
US20150271044A1 (en) * | 2014-03-24 | 2015-09-24 | International Business Machines Corporation | Browser response optimization |
JP6062511B1 (en) * | 2015-09-07 | 2017-01-18 | ヤフー株式会社 | CHANGE DEVICE, CHANGE METHOD, DISTRIBUTION METHOD, CHANGE PROGRAM, AND DISTRIBUTION SYSTEM |
US10244046B2 (en) * | 2016-11-28 | 2019-03-26 | T-Mobile Usa, Inc. | Managing traffic at a node within a wireless communication network upon the node entering service |
CN112579310A (en) * | 2019-09-30 | 2021-03-30 | 北京车和家信息技术有限公司 | Service calling method and device |
US11711274B2 (en) * | 2020-02-11 | 2023-07-25 | International Business Machines Corporation | Request response based on a performance value of a server |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5926649A (en) * | 1996-10-23 | 1999-07-20 | Industrial Technology Research Institute | Media server for storage and retrieval of voluminous multimedia data |
US6038598A (en) * | 1998-02-23 | 2000-03-14 | Intel Corporation | Method of providing one of a plurality of web pages mapped to a single uniform resource locator (URL) based on evaluation of a condition |
US6243761B1 (en) * | 1998-03-26 | 2001-06-05 | Digital Equipment Corporation | Method for dynamically adjusting multimedia content of a web page by a server in accordance to network path characteristics between client and server |
Family Cites Families (45)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5917822A (en) * | 1995-11-15 | 1999-06-29 | Xerox Corporation | Method for providing integrated packet services over a shared-media network |
US5802299A (en) * | 1996-02-13 | 1998-09-01 | Microtouch Systems, Inc. | Interactive system for authoring hypertext document collections |
US6553410B2 (en) * | 1996-02-27 | 2003-04-22 | Inpro Licensing Sarl | Tailoring data and transmission protocol for efficient interactive data transactions over wide-area networks |
US6076109A (en) * | 1996-04-10 | 2000-06-13 | Lextron, Systems, Inc. | Simplified-file hyper text protocol |
US5764235A (en) * | 1996-03-25 | 1998-06-09 | Insight Development Corporation | Computer implemented method and system for transmitting graphical images from server to client at user selectable resolution |
US5852673A (en) * | 1996-03-27 | 1998-12-22 | Chroma Graphics, Inc. | Method for general image manipulation and composition |
US5995943A (en) * | 1996-04-01 | 1999-11-30 | Sabre Inc. | Information aggregation and synthesization system |
US5918013A (en) * | 1996-06-03 | 1999-06-29 | Webtv Networks, Inc. | Method of transcoding documents in a network environment using a proxy server |
JP2982698B2 (en) * | 1996-07-08 | 1999-11-29 | 日本電気株式会社 | Distributed information service system |
JP3183181B2 (en) * | 1996-08-28 | 2001-07-03 | トヨタ自動車株式会社 | Information transmission method |
US6067557A (en) * | 1996-09-06 | 2000-05-23 | Cabletron Systems, Inc. | Method and system for allocating CPU bandwidth by prioritizing competing processes |
JPH10133973A (en) * | 1996-10-31 | 1998-05-22 | Hitachi Ltd | Html information providing method |
US5953506A (en) * | 1996-12-17 | 1999-09-14 | Adaptive Media Technologies | Method and apparatus that provides a scalable media delivery system |
US6185625B1 (en) * | 1996-12-20 | 2001-02-06 | Intel Corporation | Scaling proxy server sending to the client a graphical user interface for establishing object encoding preferences after receiving the client's request for the object |
US6028842A (en) * | 1996-12-23 | 2000-02-22 | Nortel Networks Corporation | Dynamic traffic conditioning |
US6052718A (en) * | 1997-01-07 | 2000-04-18 | Sightpath, Inc | Replica routing |
JPH10242997A (en) * | 1997-03-03 | 1998-09-11 | Nippon Telegr & Teleph Corp <Ntt> | Information providing device and information providing method |
WO1998044733A1 (en) * | 1997-03-31 | 1998-10-08 | Broadband Associates | Method and system for providing a presentation on a network |
US6269078B1 (en) * | 1997-04-04 | 2001-07-31 | T. V. Lakshman | Method and apparatus for supporting compressed video with explicit rate congestion control |
US6041041A (en) * | 1997-04-15 | 2000-03-21 | Ramanathan; Srinivas | Method and system for managing data service systems |
US6317795B1 (en) * | 1997-07-22 | 2001-11-13 | International Business Machines Corporation | Dynamic modification of multimedia content |
US6021426A (en) * | 1997-07-31 | 2000-02-01 | At&T Corp | Method and apparatus for dynamic data transfer on a web page |
US6167085A (en) * | 1997-07-31 | 2000-12-26 | Sony Corporation | Image data compression |
US6263114B1 (en) * | 1997-07-31 | 2001-07-17 | Sony Corporation | Apparatus and method for compressing image data by determining a quantization factor which provides more efficient use of an available bit rate |
US6006264A (en) * | 1997-08-01 | 1999-12-21 | Arrowpoint Communications, Inc. | Method and system for directing a flow between a client and a server |
US6061700A (en) * | 1997-08-08 | 2000-05-09 | International Business Machines Corporation | Apparatus and method for formatting a web page |
US6564250B1 (en) * | 1997-08-21 | 2003-05-13 | Planetweb, Inc. | Miniclient for internet appliance |
US6148005A (en) * | 1997-10-09 | 2000-11-14 | Lucent Technologies Inc | Layered video multicast transmission system with retransmission-based error recovery |
US6594699B1 (en) * | 1997-10-10 | 2003-07-15 | Kasenna, Inc. | System for capability based multimedia streaming over a network |
US6157614A (en) * | 1997-10-22 | 2000-12-05 | Netro Corporation | Wireless ATM network with high quality of service scheduling |
US6128668A (en) * | 1997-11-07 | 2000-10-03 | International Business Machines Corporation | Selective transformation of multimedia objects |
US6216157B1 (en) * | 1997-11-14 | 2001-04-10 | Yahoo! Inc. | Method and apparatus for a client-server system with heterogeneous clients |
US6343350B1 (en) * | 1998-02-13 | 2002-01-29 | International Business Machines Corporation | Conserving storage space by means of low resolution objects |
US6105042A (en) * | 1998-02-13 | 2000-08-15 | Cylex Systems, Inc. | Multi-user information management system adapted for efficient, remote, on-demand document management, storage and retrieval |
US6314463B1 (en) * | 1998-05-29 | 2001-11-06 | Webspective Software, Inc. | Method and system for measuring queue length and delay |
GB9817909D0 (en) * | 1998-08-17 | 1998-10-14 | Zeneca Ltd | DNA constructs |
US6182125B1 (en) * | 1998-10-13 | 2001-01-30 | 3Com Corporation | Methods for determining sendable information content based on a determined network latency |
US6185221B1 (en) * | 1998-11-09 | 2001-02-06 | Cabletron Systems, Inc. | Method and apparatus for fair and efficient scheduling of variable-size data packets in an input-buffered multipoint switch |
US6389462B1 (en) * | 1998-12-16 | 2002-05-14 | Lucent Technologies Inc. | Method and apparatus for transparently directing requests for web objects to proxy caches |
US6587109B1 (en) * | 1998-12-24 | 2003-07-01 | B3D, Inc. | System and method for real-time scalability of 3D graphics based on internet bandwidth and CPU speed |
US6505256B1 (en) * | 1999-01-15 | 2003-01-07 | Compaq Information Technologies Group, L.P. | Automatic synchronization of state colors across a web-based system |
US6484212B1 (en) * | 1999-04-20 | 2002-11-19 | At&T Corp. | Proxy apparatus and method for streaming media information |
US6581109B1 (en) * | 1999-06-24 | 2003-06-17 | International Business Machines Corporation | System to dynamically adjust image colors in client/server environment to assure accurate color reproduction |
US6441831B1 (en) * | 2000-04-04 | 2002-08-27 | Learningaction, Inc. | Choosing a multimedia presentation |
US20020069296A1 (en) * | 2000-12-06 | 2002-06-06 | Bernie Aua | Internet content reformatting apparatus and method |
-
1999
- 1999-04-26 US US09/299,684 patent/US8200837B1/en not_active Expired - Fee Related
- 1999-11-18 DE DE69939460T patent/DE69939460D1/en not_active Expired - Lifetime
- 1999-11-18 EP EP99122965A patent/EP1049031B1/en not_active Expired - Lifetime
-
2000
- 2000-03-22 JP JP2000079687A patent/JP2000322366A/en active Pending
-
2012
- 2012-05-16 US US13/473,417 patent/US20120233294A1/en not_active Abandoned
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5926649A (en) * | 1996-10-23 | 1999-07-20 | Industrial Technology Research Institute | Media server for storage and retrieval of voluminous multimedia data |
US6038598A (en) * | 1998-02-23 | 2000-03-14 | Intel Corporation | Method of providing one of a plurality of web pages mapped to a single uniform resource locator (URL) based on evaluation of a condition |
US6243761B1 (en) * | 1998-03-26 | 2001-06-05 | Digital Equipment Corporation | Method for dynamically adjusting multimedia content of a web page by a server in accordance to network path characteristics between client and server |
Non-Patent Citations (1)
Title |
---|
Engelschall, Ralf S. "Apache HTTP Server Version 1.3: Module mod_rewrite". Archived on January 11, 1998 by the Internet Archive: . Pages 1 to 18. * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9699032B2 (en) * | 2013-10-29 | 2017-07-04 | Virtual Instruments Corporation | Storage area network queue depth profiler |
WO2015175174A1 (en) * | 2014-05-13 | 2015-11-19 | Pubmatic, Inc. | Intelligent ad auction and sla compliance techniques |
WO2018004644A1 (en) * | 2016-07-01 | 2018-01-04 | Intel Corporation | Adaptive synching |
US11106621B2 (en) | 2016-07-01 | 2021-08-31 | Intel Corporation | Adaptive synching |
US20180146033A1 (en) * | 2016-11-23 | 2018-05-24 | Level 3 Communications, Llc | Overload handling in a content delivery network (cdn) |
US10924583B2 (en) * | 2016-11-23 | 2021-02-16 | Level 3 Communications, Llc | Overload handling in a content delivery network (CDN) |
Also Published As
Publication number | Publication date |
---|---|
US8200837B1 (en) | 2012-06-12 |
DE69939460D1 (en) | 2008-10-16 |
EP1049031B1 (en) | 2008-09-03 |
EP1049031A3 (en) | 2001-03-21 |
JP2000322366A (en) | 2000-11-24 |
EP1049031A2 (en) | 2000-11-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20120233294A1 (en) | Adaptive web server | |
Abdelzaher et al. | Web server QoS management by adaptive content delivery | |
US7400633B2 (en) | Adaptive bandwidth throttling for network services | |
US8799502B2 (en) | Systems and methods for controlling the number of connections established with a server | |
Abdelzaher et al. | Web content adaptation to improve server overload behavior | |
US6308214B1 (en) | Self-tuning dataflow I/O core | |
US6308216B1 (en) | Service request routing using quality-of-service data and network resource information | |
US7388839B2 (en) | Methods, apparatus and computer programs for managing performance and resource utilization within cluster-based systems | |
US6189033B1 (en) | Method and system for providing performance guarantees for a data service system of a data access network system | |
US9350795B2 (en) | Methods, systems and computer program products for proactively offering a network turbo boost service to end users | |
US20160255145A1 (en) | Automated server replication | |
US6742016B1 (en) | Request acceptor for a network application system and a method thereof | |
US6941370B2 (en) | Dynamic proxy reconfiguration system and method to support sharing of extra capacity | |
US20090077233A1 (en) | Load Control Device and Method Thereof | |
Bochmann et al. | Introducing QoS to electronic commerce applications | |
US7062557B1 (en) | Web server request classification system that classifies requests based on user's behaviors and expectations | |
Pradhan et al. | Adaptive multimedia content delivery for scalable web servers | |
Loukopoulos et al. | Optimizing download time of embedded multimedia objects for web browsing | |
Goldszmidt et al. | Scaling Internet Services by Dynamic | |
Lee et al. | Efficient server selection system for widely distributed multiserver networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |