CN103250140B - Application distribution in data center - Google Patents
Application distribution in data center Download PDFInfo
- Publication number
- CN103250140B CN103250140B CN201080070538.XA CN201080070538A CN103250140B CN 103250140 B CN103250140 B CN 103250140B CN 201080070538 A CN201080070538 A CN 201080070538A CN 103250140 B CN103250140 B CN 103250140B
- Authority
- CN
- China
- Prior art keywords
- processing unit
- application
- load information
- network load
- network
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 claims description 14
- 238000004891 communication Methods 0.000 claims description 8
- 230000001808 coupling Effects 0.000 claims description 3
- 238000010168 coupling process Methods 0.000 claims description 3
- 238000005859 coupling reaction Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 description 6
- 230000000875 corresponding Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000009825 accumulation Methods 0.000 description 2
- 238000011030 bottleneck Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001419 dependent Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Abstract
Provide the allocation entity of data center, wherein, described allocation entity is configured to: receiving the application distribution request for a processing unit distribution application in multiple processing units of described data center, described application allocation request packet includes application network load information;Obtain the path network load information of the offered load indicating at least one of network path leading to the plurality of processing unit;And based on described application network load information and described path network load information, select processing unit to distribute described application from the plurality of processing unit.
Description
Technical field
The present invention relates to the application distribution in data center.In particular it relates to the allocation entity of data center,
The operation method of allocation entity of data center, computer program, computer program and the method for operation data center
With such data center.
Background technology
Recently data center provides the long-range of application and/or distributed execution to attract bigger concern due to it.To the greatest extent
Pipe is applied in the local operation away from user, and the latter can be from using via network (such as, the Internet), visit Anywhere
Ask and operate the application that can run on a virtual machine.
In this context, the term cloud computing form as calculating based on the Internet has been devised, in cloud computing
In, provide a user with shared resource, software and information by the Internet.In other words, can be considered as cloud computing utilizing and selling
The new paragon of the ability of data center.In the latter, apply actual motion, and process relative users data.Cloud is succeeded in reaching an agreement
It is often to use virtual machine image, in virtual machine image, is mounted with set of applications.Open on the server of data center as required
Dynamic this mirror image.If the demand for application-specific changes, can start or remove this application of offer has identical or class
Mirror image like type.Thus, it is thereon virtual that term application can specify such application or one or more application to run on
Machine mirror image (or referred to as mirror image).
Generally organize data center in a hierarchical manner.Going out as discussed in connection with fig. 1, data center 1 ' includes multiple process
Unit (so-called blade server) 10.These processing units 10 are coupled to certain data via at least one switch-level 200
Central access point.Access point can be exterior gateway or router, such as shown gateway entity 30.In shown result, actual
3 switch-level of upper existence 201,202,203, they constitute the hierarchy of data center 1 '.Therefore pin will can be provided
The processing unit 10 of the usual virtual platform of application is organized as frame (racks), and then frame is organized as row, and then will
Row is organized as so-called district.Such as, the classification in district would correspond to switch-level 201, and the classification of row would correspond to switch-level
202, and the classification of frame would correspond to switch-level 203.Data path 40 (can also be referred to as network path) leads to often
Individual processing unit 10.
Whole data center 1 ' is connected to the outside such as the Internet by the such as center router of gateway entity 30 etc
Network, and the data service in data center is distributed to each classification levels.It thus provides the data in distribution data center
The subsequent switch of business or router.In the example shown, data service is distributed to row by district's switch.Row switch is distributed
To frame, and finally, blade server or processing unit are connected to the remainder of data center by frame switch.
Alternatively, the above-mentioned hierarchy in data center can also be by adding directly between data center's each several part
Shortcut realizes, and introduces these and adds direct shortcut to increase internal network ability.Such as, two frame router switch 23
Can also directly communicate with each other, regardless of the respective communication via the row router switch 22 being associated how.Additionally, classification
The number of rank can change according to the size of data center and complexity.
In order to run the application that networked capabilities is had wilderness demand, data center and assembly thereof (classification levels, switch,
Router etc.) sufficient network communications capability should be provided with the satisfied application demand to communication service.This may mean that
The processing unit physically performed the example of application (or mirror image) is coupled to meet by needs via network path
The Internet of the networking requirement of this application, this network path passes through data center's classification and external interconnections (gateway).Depend on
Application and mirror image, these networking requirements can e.g. specific minimum data transmission bandwidth.
When the application distributed in data center, need to consider mirror image and the process demand of application thereof.There iing sufficient process
Capacity can blade server on start application/mirror image.(best effort) strategy of doing one's best the most only is used to consider
Above-mentioned networking requirement: carry out the rough classification to high and low networked capabilities, and use this rough classification in application distribution.This
Scenario described below can be caused: on application of principle, can process its task, but its other parts with data center or outside
The communication of user is likely to occur congested.Result will be the bad use to disposal ability, and ultimately results in the Consumer's Experience of bad luck.
Consumer's Experience and to application or service response be Key Performance Indicator a lot of application scenarios in, this situation is can not
Accept.Such as, this is the situation for telecommunications service.If other parts of data center will have for the need to application
Seek the sufficient networked capabilities providing service, then this situation can cause shortcoming.In this case, the mirror image in data center and should
Different distribution will cause the integrity service being obviously enhanced.
Summary of the invention
The problems referred to above are solved by the theme of independent claims.Define the preferred of the present invention in the dependent claims
Embodiment.
It is an object of the present invention to provide the enhancing concept that the application in a kind of data center distributes.Specifically, the present invention
Target is consideration offered load when selecting suitably processing unit in data center required by application.Therefore, the present invention
Another target is to provide a kind of Enhancement Method operating data center and such a strengthens data center.Additionally, the present invention
Target be to provide a kind of correlation method operating this allocation entity and corresponding computer program and computer program be provided
Product.
According to aspects of the present invention, it is provided that the allocation entity of a kind of data center, described allocation entity is configured to: connect
Receive the application distribution request for a processing unit distribution application in multiple processing units of described data center, described
Application allocation request packet includes application network load information;Acquisition approach network load information, described path network load information refers to
Show the offered load breath of at least one of network path leading to the plurality of processing unit;And based on described application network
Load information and described path network load information, select processing unit to answer so that distribution is described from the plurality of processing unit
With.
According to the second aspect of the invention, it is provided that the allocation entity of a kind of data center, including: receive for described
The application distribution request of a processing unit distribution application in multiple processing units of data center, described application distribution request
Including application network load information;Acquisition approach network load information, the instruction of described path network load information is led to described many
The offered load of at least one of network path of individual processing unit;And based on described application network load information and described
Path network load information, selects processing unit to distribute described application from the plurality of processing unit.
According to the third aspect of the invention we, it is provided that the allocation entity of a kind of data center, including: receive for described
The application distribution request of a processing unit distribution application in multiple processing units, described application allocation request packet includes application net
Network load information;Acquisition approach network load information, the plurality of processing unit is led in the instruction of described path network load information
The offered load of at least one of network path;And bear based on described application network load information and described path network
Information carrying ceases, and selects processing unit to distribute described application from the plurality of processing unit.
According to the fourth aspect of the invention, it is provided that the allocation entity of a kind of data center, including multiple processing units;
Lead at least one of network path of the plurality of processing unit;And the distribution reality of each embodiment according to the present invention
Body.
According to other aspects of the invention, it is provided that a kind of computer program including code, described code can quilt
Being loaded in the code process unit of allocation entity according to an embodiment of the invention, wherein, described code performs such as to combine this
Method described in embodiment.Additionally, it is a further aspect of the present invention to provide a kind of calculating including above computer program
Machine program product.
Accompanying drawing explanation
Will now be described with reference to the attached figures and present to be more fully understood that inventive concept but be not construed as this
The bright embodiments of the invention carrying out limiting, in the accompanying drawings:
Fig. 1 shows the schematic representation of conventional data centers;
Fig. 2 shows the schematic representation with the data center of allocation entity according to an embodiment of the invention;
Fig. 3 shows the mutual schematic representation of allocation entity according to another embodiment of the present invention;
Fig. 4 illustrates in greater detail the signal of the data center with allocation entity according to another embodiment of the present invention
Represent figure;
Fig. 5 shows the mutual schematic representation of allocation entity according to another embodiment of the present invention;
Fig. 6 illustrates in greater detail the signal of the data center with allocation entity according to another embodiment of the present invention
Represent figure;
Fig. 7 shows the schematic representation of allocation entity according to another embodiment of the present invention;
Fig. 8 A shows the schematic flow diagram of the embodiment of the method for the present invention;And
Fig. 8 B shows the schematic flow diagram of the embodiment of the method for the present invention.
Detailed description of the invention
In the present invention, so-called application can identify traditional application, i.e. in order to provide a user with required merit
Can and the application program that performs.This function includes text-processing, spreadsheet calculations, data base and management, image and figure
Process/generation/the management etc. of data.But, term application can also identify the mirror image of virtual machine to be understood to.Virtual machine is
At the upper software assembly providing a virtual processing entities of another entity (i.e. physical machine).Such as, virtual machine can be to adopt
With virtual Windows (TM) computer of actual motion on the computer of operating system based on Linux.But, at virtual machine
And run virtual machine processing unit between operating system can be the most different from each other.In any case, the feature of virtual machine
Be: it provides a package implementation for performing application to one or more users, wherein, the execution of application and because of
The operation of this virtual machine does not disturb other virtual machines operated on same treatment unit (at least in certain given restriction
Under).
In the present invention, processing unit is appreciated that for processing data and/or the one of communication in data center
Individual unit.Aforementioned so-called blade server may be constructed a this processing unit.But, processing unit can also is that independence
PC or server, these independent PC multiple or server are arranged to interconnected amongst one another to form data center.Can be at least by friendship
Level of changing planes is to use described interconnection, but generally switch-level is arranged corresponding to the physics of multiple processing units, i.e. district, row, machine
Frame etc..But processing unit can also by its each ability be limited to data forward, route and/or exchange.And hence it is also possible to will
Processing unit is interpreted as itself not running switch or the router entity of mirror image, but ought run application on a processing unit
Time, application may be constructed router/switch entity.
As explained, data center uses at least one switch-level, and it can reflect the thing of multiple processing unit
Reason is arranged.If be such as arranged in frame by processing unit, and then frame is arranged as row, and plurality of this row is constituted
, at least 3 this switch-level will be there are, such as by district's switch-level, row switch-level and frame switch-level in so-called district
Represent.Between two exchange/route entities of different switch-level, there is at least one connection for data communication.
Connection between 2 differences or 2 this entities of neighboring switch level can be partially shown as the sub-road of so-called network
Footpath, wherein, is defined as so-called network path terminating at a particular processor unit in multiple processing units.
In other words, network path is the possible path of the data going into or from particular processor unit for being scheduled.Net
Network path can be at the other end at another processing unit or at certain Central Access Point (such as, preceding gateway entity) place knot
Bundle.It is thus possible to define the other end of network path with desired data stream as requested.Such as, if will by from go
Route to such as the Internet toward the data of respective handling unit, can between respective handling unit and certain senior access point shape
Become network path.But, network path can also be maintained in a data center well, such as, do not exist together connection 2
During reason unit.If be such as distributed an application between some processing units, or when an application processes with operating in another
When Another Application on unit is mutual, the most this data stream can become essential.Example can be to process list as first
In unit first applies the web server run and as the second video stream server processing upper second application operation: two
Server application can part be alternately: web server utilizes video stream server to provide video flowing based on web to answer
With.
In the arranging of above-mentioned some switch-level, it is usually present one of tree-like path structure, i.e. a switch-level
Switch/router entity receives and sends data with an entity of higher switch-level, otherwise, exist from an exchange
Machine/router entity launches (fanout) some connections to the entity of lower switch-level.So, as also more comprehensively solved below
Release, the selection launching one of possibility is defined the network path leading to particular processor unit by each switch-level.
According to each aspect of the present invention, when adding application/image instance to data center, can allow to consider in data
The networked capabilities of the heart.Specifically, it may be considered that the current traffic load of each several part of data center and local and interim available
Networking capacity.This load distribution that allows preferably to network in data center, and it can avoid using data center to provide
Application and service time may cause the bottleneck situations of bad Consumer's Experience.
Additionally, each aspect of the present invention also allows for reserving (soft) of capacity in whole data center, thus with generally
Comparing of will allowing of the scheme of doing one's best, it is possible to SLA (SLA) is provided in more-controlled fashion.With do not have
The data center of this probability compares, and this can provide some advantages, wherein it is possible to very conservative and careful mode come to
Go out SLA and ensure (if any).
Fig. 2 and 4 shows the schematic diagram of data center according to an embodiment of the invention.Multiple processing units 10 are via extremely
A few switch-level 200 is coupled to gateway entity 30.As already described above, switch-level 200 can actually include
Corresponding more than one switch-level (district's level, row level, chassis level etc.) is set with the actual physics of processing unit 10.Permissible
Switch-level is represented, such as district's switch 21, row switch 22 and frame switch 23 by respective switch entity.
According to the present embodiment, it is provided that the allocation entity 50 in data center 1, it is from gateway entity 30, one or more
Switch-level 200 and processing unit 10 receive data and/or information.By received data, allocation entity 50 can be received
Collect the information relevant to the all-network path 40 that gateway entity 30 is connected to each processing entities 10.More specifically, distribution
Actual, prediction or scheduling the network that entity 50 can be collected between gateway entity 30 and each processing unit 10 is born
Carry relevant information.Described information can relate to current and/or given interval and/or time in the future point.In other words, distribution is real
Body 50 is configured to acquisition approach network load information, and the instruction of this path network load information leads to multiple processing unit 10 extremely
The offered load of at least part of network path.
Additionally, allocation entity 50 is configured to receive for answering to the distribution of one of multiple processing units 10 of data center 1
Application distribution request 500.Application distribution request 500 includes so-called application network load information, application network load letter
Cease in terms of the required offered load required to correctly operate and/or the required offered load caused by application to multiple
The application of a processing unit distribution in processing unit characterizes.Ensureing service quality (QoS) or can also specify
Big response time aspect defines described correct operation.
Additionally, described application distribution request or described application network load information can include that so-called application network loads
Profile, this application network load profiles specifies the required offered load at aspects such as event, particular point in time or time intervals.This
Sample, allocation entity 50 can become the application-specific knowing requested distribution and requires on particular point in time or cause how many networks
Load.
Therefore, allocation entity 50 can select particular processor unit 10 ', and this particular processor unit 10 ' is via satisfied application
One particular network path 40 ' of the requirement that network load information is illustrated is coupled to gateway entity 30.Such as, distribute to many
The distribution of one of individual processing unit 10 can be the video stream server of a large amount of offered load of requirement, because the video data volume is usual
The biggest.Know the real network requirement of this application, allocation entity 50 can select processing unit 10 ' to distribute application because its
Corresponding network path 40 ' can process currently, offered load in the future or required by special time in the future and time interval.
In other words, allocation entity 50 is introduced as new node (that is, business aware applications/mirror image allotter), its
Allow to find in data center application/mirror image is suitably distributed.In doing so, this node is collected and in data center
All networked link on the relevant data of life-span networking load situation.Use router and the traffic measurement interface of switch
Collect this information.If needing to distribute new application/mirror image, allocation entity 50 can perform search pattern, this search pattern
External interconnections node (gateway entity 50) in data center starts, and by choosing the path with minimum networking load, logical
The classification crossing networked node/switch is carried out to blade server (in the example shown, processing unit 10 ').
So, load consciousness management level are also introduced.This layer by the virtual image of data center's structure (thus, in data
The model of the heart) preserve together with instantaneous networking load situation.Additionally, the application/mirror image distributed is with corresponding application network
The networking load profiles of load information or correspondence.This is the description of load request of networking application.When dividing in data center
When joining application, load consciousness management level are that this application reserved networking virtually is born according to load profiles by data center all the time
Carry.If business consciousness mirror image allotter performs its search pattern to virtual data center model rather than real data center,
In order to be found for the optimal location of this application, i.e. processing unit 10 ', can be by the networking load profiles according to other mirror images
Other mirror image is reserved and potential load needs are taken into account.
It is said that in general, no matter data center distributes the networking aspect of the decision of application how, it is of course possible to exist permissible
Affect a lot of other standards of this decision.Entity and method that the present invention proposes provide contribution to overall decision process.
Referring now to Fig. 3 and 4, allocation entity 50 has the friendship led in the classification of data center 1 and tree network structure
Change planes/router 21,22,23 and the interface of processing unit 10.Additionally, allocation entity 50 has leads to data center management system
The interface 66 of 64.By this interface 66, allocation entity 50 can receive can distribute new opplication for search in data center 1
The order of position.The result of this search is transmitted back to data center management, itself so application/mirror image can be distributed accordingly.
In order to obtain the information of requirement about application, allocation entity 50 have lead to application and or mirror image warehouse (i.e.,
Data storage 600) interface 65, application or mirrored storage are in this warehouse or described in this warehouse.Interface 65 can be by
Application/mirror image Administration API (application programming interface) is constituted.Except the information (data relevant to this application being commonly stored
Code, application data and storehouse 71 in memorizer 61) outside, introduce application network load information 72, itself and application/mirror image phase
Association, to provide and to require relevant additional information, specifically, about the additional information of offered load.Described information 72 is permissible
It is stored in the data storage 62 of separation, or application data 71 and information 72 can be stored in a single common data and deposit
In reservoir 600.
The source of information is positioned in the availability of the interface 67 leading to data center network node 63, and this interface 67 allows to fetch
Relevant to all nodes and network link thereof (switch 21,22,23, gateway entity 30 and the processing unit 10 that i.e. relate to)
Current live network load situation.Therefore, by least some of node and link couples to allocation entity 50, to provide this
Information (dotted line in Tu2 &4).
In one embodiment, allocation entity 50 can ask node and the link currently considering with it/observing on one's own initiative
The relevant prompting message of load situation, or in another embodiment, it can be regularly to the poll load of whole data center 1
Information, and store this information to be used in searching algorithm.In another embodiment, can at the network node of data center (i.e.
The switch 21,22,23 related to) place's registration allocation entity 50, to receive the rule renewal relevant to network load information.
Furthermore, it is possible to these concepts fetching and processing load situation are combined, and can differently process in data
The different piece of the heart 1.
When new opplication should be added to data center 1 or adds the virtual image being mounted with application on it, allocation entity
50 participate in make decision: should distribute application/mirror image in data center 1.It finds has the highest idle connection in data center 1
Network capacity amount, and specifically, on the whole network path 40 ' of outside (i.e. gateway entity 30), there is the free time from processing unit 10
And the suitable treatments unit 10 of suitable capacity.The order performing this search manages 64 typically from data center, and to number
Result is reported back to according to centre management 64.This allows data center management system 64 when distributing the new example of application/mirror image by net
Network load is taken into account.
As has already been mentioned above, application/mirror image can also be with describing the profile that its networking load needs.A reality
Execute in example, can the such as individual files 72 in the warehouse (that is, data storage 62) of this load profiles store this
Load profiles.In this case, suitable profile is individually selected for each application/mirror image.In another embodiment, can be by
Application network load information is together with application/mirrored storage, or is even stored in the district of image file or description of image.
The profile of application network load information can comprise such as by the required minimum bandwidth being less than all the time or maximum belt
Wide information.But, this profile can also comprise the information relevant to conditionality or timeliness bandwidth feature (figure).Example
As, single bandwidth requirement can be provided for over skies different in different time or a week in a day.Allocation entity 50 can
To use this load profiles to determine, network link to be selected is if appropriate for supporting to be assumed to be in data center 1 to be deployed in
Being required of application/mirror image.Can expand to described application network load profiles also to comprise and being not related to of application/mirror image
The requirement that other aspects of networking are relevant.Such as, can come together designated treatment capacity requirement with networking requirement herein.
Referring now to Fig. 5 and 6, allocation entity 50 may be provided for data center's load controller 51.Data center
Load controller 51 establishes intermediate layer in the measurement that data center network loads.Data center's load controller 51 can be by
The map of data center topology stores together with the feature of present load situation.For data center each physical node and
Each link in data center's node, data center's load controller preserves the information that load situation of networking to it is relevant.With
Sample ground, described node can be made up of switch 21,22,23, entity 30 and/or unit 10.Described link may be constructed
State subpath.So, data center's load controller 51 safeguards the abstract mirror image of real data central loading situation.Taking of information
Return and the renewal continuously of load characteristic can be carried out in the way of being similar to describe together with other embodiments of the present invention.
Along with this intermediate layer is on appropriate position, allocation entity 50 the most directly will receive load from real data Centroid
Feature, but receive, from data center's load controller 51, the load characteristic obtained its abstract data center map safeguarded.
Except this change in finding the algorithm of position of new opplication/mirror image, other can keep identical.
This embodiment abstract mirror image based on data center rather than data center itself.Based on data center 1 its is taken out
As mirror image, data center's load controller 51 can perform the reserved of network capacity for each node and link.Therefore, number
Can receive or obtain path network according to central loading controller 51 and reserve information.Note that final deployment in data center 1
Application example, and the network path 40 ' found for this application example to its notice.Once deploy the example, and along quilt
Select in all virtual data center nodes on the path 40 ' for providing the Internet services to this application example and virtual data
Storing this offered load profile at heart link, data center's load controller 501 accesses the respective application offered load letter of application
Shelves (or general, application network load information).
If allocation entity 50 is new application example searching position, then the load presented for each node and link
Feature can be life-span (currently) load, or it can be based in data center's load controller 50 load profiles of storage
The potential load that the accumulation of definition requires.In addition it is possible to use the combination that live load characteristic and accumulation require.
As such, it is possible to reach the soft reservation of network capacity, because when distributing new opplication, the application example that will be deployed
Required capacity is taken into account.Although instruction is left capacity by present load situation, but for new deployment, can refuse
There is the path 40 of the high potential demand relevant to networking.Data center's load controller 51 is known and is serviced by this path
Application there is potential high demand.It can thus be avoided bottleneck in the future.
Owing to offered load profile can provide such as in different time and the complex characteristic of distinguishing load request on the date,
Therefore when determining the load whether specific node or link can obtain Another Application, it is also possible to this point is taken into account.No
Only can mate current requirement, it is also possible to coupling requirement in the future.Need by data center's management to data center's load control
Device 51 processed notifies that all of mirror image is disposed and removes.If removed or Mobile solution example, need to remove or rearrange and refer to
The path of the offered load profile of group corresponding storage together.
Fig. 7 shows the internal schematic diagram arranged of allocation entity according to another embodiment of the present invention.This allocation entity
50 include processor 501 and memorizer 502.Processor performs the code in memorizer 502 as unit 503 storage.Thus,
Can be realized or perform appointing of any feature as combined the entity described by embodiments of the invention or method by unit 503
What step, wherein, is performed its respective code by processor 501.
Fig. 8 A shows the schematic flow diagram of the embodiment of the method for the present invention.Therefore, in step sl, receive for number
Application distribution request according to a processing unit distribution application in multiple processing units at center.Application allocation request packet includes should
Use network load information.Additionally, in step s 2, acquisition approach network load information, the instruction of this path network negative information is led to
The offered load of at least one of network path in multiple processing units.Finally, in step s3, based on application network
Load information and path network load information, select processing unit to distribute application from multiple processing units.
Fig. 8 B shows the schematic flow diagram of the other method embodiment of the present invention.This embodiment describes in data
Search for the possible method of suitable processing unit for application-specific to be distributed in the minds of in.According to this embodiment, first
In step S11, search for suitable processing unit and start at the gateway entity of data center.Then, in step s 12, distribution
Entity checks the load situation relevant to all paths to subsequent node/router/switch.Based on this situation, in step
In S15, allocation entity is the most NA, to the subsequent node with the highest idle capacity before selecting in this search procedure
The link (subpath) of (that is, the switch of next one classification levels).
If in S13, can not find in the node/router/switch of current check and there is sufficient idle networking appearance
The link of amount, then the network node before search returns and step S12 of this algorithm.If owing to there is not more high-grade rank
Network node/router/switch before and make this impossible, then search for unsuccessfully and can not distribute in data center
This application (S14).If the subsequent node after selected link is router/switch equally, algorithm continues in step s 12
Continuous.If but the subsequent node after selected link is processing unit, in step S17, checks whether this unit meets want
Being required of the application/mirror image of distribution.This networking load situation being possible not only to check processing unit, it is also possible at such as inciting somebody to action
Reason load or memorizer utilization rate are taken into account.Required, then in step if the processing unit found in step S17 meets
Rapid S18 manages in this place and on unit 18, distributes/dispose this application.
If but the processing unit found in step S17 is unsatisfactory for being required, the network before search returns saves
Point and step S12 of this algorithm.If owing to there is no network node/router/switch before and make this impossible, then
Search for unsuccessfully and this application/mirror image (same, S14) can not be distributed in data center.
Above-mentioned algorithm can be to have the application/mirror image of high networking requirement to find external network (such as, public internet)
Suitably distribution.The algorithm of amendment can be provided for the application/mirror for there being high networking requirement to other parts of data center
As finding suitably distribution.In this case, in step s 11, search will have Gao Lian at application/mirror image to be applied to it
Start at the processing unit of net demand.The algorithm presented comprises following inspection: the active volume of link is the most fully supported to want
The requirement of the application/mirror image of distribution.Application/image load profile as above can be used to determine relevant to these requirements
Information.If this information is unavailable, then algorithm can also only select optimal path, and does not control whether its capacity will fill
Foot.
In the above embodiment of the present invention, at least contemplate following amendment: according to amendment, distribution can be realized accordingly real
Body is to run on the data centre, i.e. as the code performed by one or more processing units of data center.
Although having been described for detailed embodiment, these embodiments are merely provided for the basis being defined by the following claims
That invents is better understood from, and these embodiments should be considered as restrictive by expection.
Claims (17)
1. a distributing equipment for data center, described distributing equipment is configured to:
-receive to apply and distribute request, described application distribution request is used for the virtual image portion by applying or be provided with described application
On the processing unit being deployed in multiple processing units of described data center, described application allocation request packet includes application network
Load information, described application network load information describes logical in described data center of after following operation described application
The demand of communication service: the processing unit that described application or described virtual image are deployed in the plurality of processing unit
On;
-acquisition approach network load information, the instruction of described path network load information leads to the plurality of processing unit at least
The offered load of the network path of a part;And
-based on described application network load information and described path network load information, select from the plurality of processing unit
Processing unit is to dispose described application or described virtual image.
Distributing equipment the most according to claim 1, wherein, described network path passes through at least one switch-level by described
The gateway device of data center is coupled to the plurality of processing unit, and described distributing equipment is also configured to
-obtaining subpath network load information, described subpath network load information indicates described gateway device and the plurality of
The offered load of the network subpath between processing unit;And
-selecting a processing unit to dispose described application from the plurality of processing unit, described selection is based on being coupled to
State the subpath network load information of the subpath of a processing unit.
Distributing equipment the most according to claim 2, is also configured to
-determine whether another subpath of coupling one processing unit can be used based on described application network load information;
And
If-another subpath described is unavailable, select another processing unit described to dispose from the plurality of processing unit
Application or described virtual image.
Distributing equipment the most according to claim 1, is also configured to
-RX path network reservation information;And
-be additionally based upon described path network and reserve information, from the plurality of processing unit, select described processing unit.
Distributing equipment the most according to claim 1, is also configured to based on Another Application network load information and described road
Footpath network load information, selects another processing unit to dispose Another Application or virtual image from the plurality of processing unit.
Distributing equipment the most according to claim 1, is also configured to by relatively described application network load information with logical
The path network load information of the network path of each processing unit in the plurality of processing unit, selects described process single
Unit.
Distributing equipment the most according to claim 6, is also configured at least based on leading in the plurality of processing unit every
The offered load of the network path of individual processing unit selects described processing unit.
Distributing equipment the most according to claim 1, wherein, described application network load information includes application network load letter
Shelves.
Distributing equipment the most according to claim 1, described distributing equipment be implemented as at one of described data center or
The application run on multiple processing units.
10. the method operating the network equipment of data center, described method includes:
-receive to apply and distribute request, described application distribution request is used for the virtual image portion by applying or be provided with described application
On the processing unit being deployed in multiple processing units of described data center, described application allocation request packet includes application network
Load information, described application network load information describes logical in described data center of after following operation described application
The demand of communication service: the processing unit that described application or described virtual image are deployed in the plurality of processing unit
On;
-acquisition approach network load information, the instruction of described path network load information leads to the plurality of processing unit at least
The offered load of the network path of a part;And
-based on described application network load information and described path network load information, select from the plurality of processing unit
Processing unit is to dispose described application or described virtual image.
11. methods according to claim 10, described method is applicable to according to any one of claim 1 to 9
Distributing equipment.
12. 1 kinds of operations include the method for the data center of multiple processing unit, and described method includes:
-receive to apply and distribute request, described application distribution request is used for the virtual image portion by applying or be provided with described application
On the processing unit being deployed in the plurality of processing unit, described application allocation request packet includes application network load information,
Described application network load information describes after following operation described application to the communication service in described data center
Demand: on the processing unit that described application or described virtual image are deployed in the plurality of processing unit;
-acquisition approach network load information, the instruction of described path network load information leads to the plurality of processing unit at least
The offered load of the network path of a part;And
-based on described application network load information and described path network load information, select from the plurality of processing unit
Processing unit is to dispose described application or described virtual image.
13. according to claim 12 operation data centers methods, wherein, described data center include gateway device with
And at least one switch-level, the plurality of processing unit is coupled to via network path by least one switch-level described
Described gateway device, described method also includes:
-obtaining subpath network load information, described subpath network load information indicates described gateway device and the plurality of
The offered load of the network subpath between processing unit;And
-selecting a processing unit to dispose described application from the plurality of processing unit, described selection is based on being coupled to
State the subpath network load information of the subpath of a processing unit.
14. methods according to claim 13, also include:
-determine whether another subpath of coupling one processing unit can be used based on described application network load information;
And
If-another subpath described is unavailable, select another processing unit described to dispose from the plurality of processing unit
Application.
15. methods according to claim 12, also include:
-RX path network reservation information;And
-be additionally based upon described path network and reserve information, from the plurality of processing unit, select described processing unit.
16. methods according to claim 12, also include: based on Another Application network load information and described path network
Load information, selects another processing unit to dispose Another Application or virtual image from the plurality of processing unit.
17. 1 kinds of data centers, including:
-multiple processing units;
-lead at least one of network path of the plurality of processing unit;And
-distributing equipment according to any one of claim 1 to 9.
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/EP2010/064905 WO2012045338A1 (en) | 2010-10-06 | 2010-10-06 | Application allocation in datacenters |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103250140A CN103250140A (en) | 2013-08-14 |
CN103250140B true CN103250140B (en) | 2016-12-14 |
Family
ID=
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6195703B1 (en) * | 1998-06-24 | 2001-02-27 | Emc Corporation | Dynamic routing for performance partitioning in a data processing network |
CN101156138A (en) * | 2005-04-07 | 2008-04-02 | 国际商业机器公司 | Method, system and program product for outsourcing resources in a grid computing environment |
US7725603B1 (en) * | 2008-04-30 | 2010-05-25 | Network Appliance, Inc. | Automatic network cluster path management |
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6195703B1 (en) * | 1998-06-24 | 2001-02-27 | Emc Corporation | Dynamic routing for performance partitioning in a data processing network |
CN101156138A (en) * | 2005-04-07 | 2008-04-02 | 国际商业机器公司 | Method, system and program product for outsourcing resources in a grid computing environment |
US7725603B1 (en) * | 2008-04-30 | 2010-05-25 | Network Appliance, Inc. | Automatic network cluster path management |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111953526B (en) | Hierarchical computational power network arrangement method, device and storage medium | |
KR100754296B1 (en) | Method and apparatus for web farm traffic control | |
Kim et al. | CometCloud: An autonomic cloud engine | |
CN107241384A (en) | A kind of content distribution service priority scheduling of resource method based on many cloud frameworks | |
CN102047226A (en) | Distributed service framework | |
CN110308995A (en) | A kind of edge cloud computing service system edges cloud node deployment device | |
US8468228B2 (en) | System architecture method and computer program product for managing telecommunication networks | |
CN110209492A (en) | A kind of data processing method and device | |
CN114090244B (en) | Service arrangement method, device, system and storage medium | |
CN108494835A (en) | The realization method and system of distributed dynamic routing based on Raft algorithms | |
CN116389491B (en) | Cloud edge computing power resource self-adaptive computing system | |
WO2020134133A1 (en) | Resource allocation method, substation, and computer-readable storage medium | |
Antevski et al. | Resource orchestration of 5G transport networks for vertical industries | |
US9654333B2 (en) | Application allocation in datacenters | |
CN104823418B (en) | For preventing demand deadlock and realizing the traffic engineering system of balanced link utilization | |
De Palma et al. | A declarative approach to topology-aware serverless function-execution scheduling | |
Gomes et al. | A combined energy-bandwidth approach to allocate resilient virtual software defined networks | |
Vijayalakshmi et al. | Investigations on job scheduling algorithms in cloud computing | |
CN103250140B (en) | Application distribution in data center | |
CN112231096B (en) | Method, system, equipment and medium for task balancing of FPGA (field programmable Gate array) pooled resources | |
CN110430236A (en) | A kind of method and dispatching device of deployment business | |
US10193790B2 (en) | Systems and methods for an intelligent, distributed, autonomous, and scalable resource discovery, management, and stitching | |
Fichera et al. | Experimental evaluation of dynamic resource orchestration in multi-layer (packet over flexi-grid optical) networks | |
Büyükşahin et al. | Support service for reciprocal computational resource sharing in wireless community networks | |
Martínez et al. | Resource Abstractions in NFV Management and Orchestration: Experimental Evaluation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |