Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberCN1565143 A
Publication typeApplication
Application numberCN 02819755
PCT numberPCT/DE2002/003750
Publication date12 Jan 2005
Filing date4 Oct 2002
Priority date4 Oct 2001
Also published asCA2462793A1, CA2462793C, DE50214634D1, EP1433352A1, EP1433352B1, US7251242, US20040264450, WO2003032676A1
Publication number02819755.0, CN 02819755, CN 1565143 A, CN 1565143A, CN-A-1565143, CN02819755, CN02819755.0, CN1565143 A, CN1565143A, PCT/2002/3750, PCT/DE/2/003750, PCT/DE/2/03750, PCT/DE/2002/003750, PCT/DE/2002/03750, PCT/DE2/003750, PCT/DE2/03750, PCT/DE2002/003750, PCT/DE2002/03750, PCT/DE2002003750, PCT/DE200203750, PCT/DE2003750, PCT/DE203750
InventorsK施罗迪
Applicant西门子公司
Export CitationBiBTeX, EndNote, RefMan
External Links: SIPO, Espacenet
Distributed transmission of traffic flows in communication networks
CN 1565143 A
Abstract  translated from Chinese
在一个网络节点中,给每个现实分支图形预先规定至少各一个等候者长队。 In a network node to each reality each branch at least a predetermined graphics were waiting in long lines. 对此,能将一个等候者长队完全为多个分配扇共同使用,这些分配扇在这个网络节点中映像到相同的分支图形上。 In this regard, can a person waiting in long lines exclusively for the use of multiple distribution sectors together, these allocations fan in this image to the same network node branch graph. 将到来的包按照其对分配扇的从属(即不是按照在分配扇内对一个一定的流的从属)区分。 The arrival of the package in accordance with its subordinate allocation fan (ie not in accordance with the fan in the allocation of a certain slave stream) Division. 相应地将它们登记到一个所属的等候者长队中。 Accordingly, they will be registered to the waiting a long queue belongs. 等候者长队由属于各自分支图形的端口操作。 The waiting in long lines belonging to each branch pattern port operations. 一旦有一个端口空闲出,它就选出下一个由它操作的等候者长队。 Once a port is idle out, it is elected by the next person waiting in long lines it operates. 选择开始的如此及时,致使在出去的线路上没有空隙出现。 Select Start is so timely, so that there are no gaps appear in the line out. 如果一个端口操作多个(在这种情况整个或部分重叠的)分支图形,则在这个选择时首先经过所有这些分支图形寻找不是空白的等候者长队。 If a port to operate multiple (in this case, in whole or in part overlapping) branch graph, then this selection first graph Looking through all these branches were not blank wait in long lines. 如果此时从不同的分支图形中找到多个不是空白的等候者长队,则按照一个预先规定的准则决定,将哪个等候者长队要作为下一个来操作。 If at this time to find more than one person is not blank wait in long lines from different branches of the drawing, in accordance with the guidelines of a predetermined decision to which the waiting in long lines to operate as the next. 由这个等候者长队输出的包按照另一个准则去确定。 From this the waiting in long lines to determine the output of the package according to another criterion.
Claims(29)  translated from Chinese
1.面向包的通信网络(100,100a,200)的网络节点(1-15),-其节点按照网络的拓扑学彼此间连接在一起,使路径的多数存在于网络节点之间,-其中网络节点包括在至少一个分配扇(VF)中,它是为了一个具体的通信关系在一个作为发送节点(A,C)形成的网络节点与一个作为接收节点(B,D)形成的网络节点之间由网络拓扑学推导出并包括所有的网络节点和路径,它们能在通信网络中为了一个分配的传送被分配给通信关系的业务流(VS)合理地利用,-由此至少为了网络节点一个分支图形,在其上映像分配扇,是确定的,通过它确定了从网络节点出去的端口(P)的一个束(B),将分配给分配扇的业务流分配到这些端口上,此时分配给端口的路径导向至少两个不同的、相邻的网络节点,-其中网络节点为分支图形包括至少一个要包括至少一个等候者长队(Q11)的等候者长队系统(Q11-Q1n),如果通过网络节点相应于分支图形要至少现实分配传送一个业务流,等候者长队至少就是预先规定的。 1. The packet-oriented communication network (100,100a, 200) of the network nodes (1-15), - in accordance with its node topology network with each other are connected together, so that the majority is present in the path between network nodes, - wherein The network node comprises at least one distribution fan (VF), which is specific to a network node in a communication relationship is formed as a sending node (A, C) with a network node as a receiving node (B, D) formed of inter deduced by the network topology and includes all network nodes and paths which can be transferred to a dispensing assigned to the communication traffic relationship (VS) rational use in the communication network, - whereby at least one network node to branch graph on which the image distribution sector, is determined by which to determine the port (P) is a bundle (B) out from the network node, is assigned to the assignment allocated fan of traffic to these ports at this time assigned to the route guidance port of at least two different adjacent network nodes - which is a branch network node comprising at least one graphic to include at least one person waiting in a long queue (Q11) of the waiting in a long queue system (Q11-Q1n) , if the corresponding branch is patterned by a network node to allocate at least convey a real traffic, the waiting in long lines at least is predetermined.
2.如权利要求1的网络节点,具有用于每个分支图形的一个单独的等候者长队系统(Q21-Q2n,Q31-Q3n),分支图形的束要与其余分支图形的束在至少一个端口有区别。 2. The network node according to claim 1, having a single person waiting in long lines for each branch graphics system (Q21-Q2n, Q31-Q3n), bundle branch graphic pattern to be with the rest of the branch at least one beam ports are different.
3.如上述权利要求之一的网络节点,具有至少一组等候者长队系统,该组为了一个分支图形(i),在该分支图形上将映像多个不同的分配扇,包括每个映像的分配扇一个等候者长队系统。 3. The network node to one of the preceding claims, having at least one set of the waiting in a long queue system, the set for a branch graph (i), in the branch graphic image will be assigned a number of different sectors, including every image The allocation fan a long queue waiting by the system.
4.如上述权利要求之一的网络节点,具有至少一个单个的等候者长队系统,该系统分配给另一个分支图形(i),将多个不同的分配扇反映到该分支图形上。 The network node of the preceding claims, with at least a single person waiting in a long queue system, which is assigned to another branch graph (i), will be assigned a number of different sectors to reflect the branch graph.
5.如上述权利要求之一的网络节点,其中至少一个等候者长队系统包括至少两个等候者长队,其中两个等候者长队的至少一个是用一个优先权(401-403)标记的。 5. The network node to one of the preceding claims, wherein at least one of the waiting in a long queue system comprises at least two persons waiting in long lines, two of which were long lines waiting for at least one (401-403) marked with a priority a.
6.如上述权利要求之一的网络节点,至少为了那些在一个优先的、尤其是高优先权的等候者长队中等候的包用一个时间信息(501-503),从中能读出何时将包登记到等候者长队中或它们在里面已经有多长时间。 6. The network node to one of the preceding claims, at least for those in a priority, especially those high-priority waiting in long lines waiting for the pack with a time information (501-503), which can be read when The registration package to the waiting in long lines, or how long they have been in there.
7.如权利要求6的网络节点,具有一个用于时间信息的数值范围,它是在考虑到最短可能的包在最快的端口上传送时间的情况下选择的,使在一个有优先权的包所给出的最大延迟时间内不出现溢出。 7. The network node according to claim 6, having a range of values for the time information, which is considered the shortest possible case of packet transmission time on the fastest port selection, so in a priority It does not occur within the maximum packet delay time given to overflow.
8.如权利要求6或7之一的网络节点包括一个分开的、平行进行的过程,以便在考虑到时间信息的情况下挑出和/或抛弃其继续传送看来已毫无意义的包。 8. 6 or 7 of the network node comprises a separate, parallel process of claim to pick and / or discard it seems to have no sense to continue to transmit packets in consideration of the case where the time information.
9.如上述权利要求之一的网络节点,为至少等候者长队之一包括有第二个时间信息,通过它为分配给等候者长队的束的每个端口,显示其当时等候者长队最后的操作时间点。 9. The network node to one of the preceding claims, at least one of the waiting in long lines includes a second time information through its port waiting to be assigned to each team's long bundle, which was displayed by the long wait Operation time team final.
10.装置,尤其是通信网络,包括至少一个如上述装置权利要求之一的网络节点。 10. The apparatus, in particular a communications network, comprising at least one network node apparatus as claimed in one of the above claim.
11.在一个如上述装置权利要求之一的网络节点中,用下列步骤应用一种方法:-将一个到来的包登记到等候者长队之一中去,和-为了决定从等候者长队之一中输出下一个包,至少由网络节点的出去的端口之一触发一个调度-过程。 11. As described in a device according to one of the requirements of network nodes, the following steps to apply a method of: - an incoming packet registration to one of the waiting in long lines to go, and - for those who decided to wait in long lines One of the next packet output from network nodes out of at least one port to trigger a scheduling - process.
12.如权利要求11的应用,其中调度-过程由端口如此及时在其释放前被触发,使关于下一个要输出的包和下一个包本身如此及时准备好,致使下一个包能由端口直接连接到前面的包上传送。 12. The application of claim 11, wherein the scheduling - so timely process is triggered by the port before it is released, so that on the next package to be output and the next packet itself so ready in time, so that the next packet can be directly by the port connected to the transmission on the front of the pack.
13.如上述应用权利要求之一的应用,其中在多个分配给端口已占满的等候者长队的情况,按照一个第一预先规定的准则决定,将哪个等候者长队作为下一个来操作。 Case the waiting in a long queue of claim 13. The use of one of these applications, which are assigned to more than one port filled, in accordance with the guidelines of a first predetermined decision to which the waiting in a long queue as the next to operation.
14.如上述权利要求之一的应用,其中在等候者长队的一个至少部分标记有优先权的情况,在决定时,一直不考虑具有一个低优先权的、已占满的等候者长队,就像还至少有一个与这些等候者长队相比具有较高优先权的、已占满的等候者长队那样。 14. The application of the preceding claims, wherein the waiting in long lines at least part of a situation marked with priority in determining, has not considered a low priority has been filled by a long queue waiting , as also at least one compared to those who wait in long lines with higher priority, it has been filled by a long queue waiting that.
15.如两个上述权利要求之一的应用,其中第一个准则作为下列准则之一形成:-随机原则,-周期顺序,-束的大小,-最短的等候者长队首先,-最长的等候者长队首先,-按照等候者长队的加权,-自从最后操作这个等候者长队以来过去的时间,尤其是作为FIFO原理形成的,在此原理中将最长时间没有被操作的等候者长队作为下一个来操作,或-对其操作负责的端口数量。 15. The application of one of the above two claim, wherein the first criterion is formed as one of the following criteria: - randomly - cycle sequence - beam size - the shortest length of the waiting team first - the longest The length of the waiting team first - according to those who wait in long lines weighting - since the last operation of this long line past the waiting time, especially as the formation of FIFO principle, this principle will not be operated in the maximum time of waiting in long lines as the next person to operate, or - the number of ports in charge of its operation.
16.如上述应用权利要求之一的应用,其中在一个用一个以上的包占据等候者长队的情况,将下一个包按照一个第二准则来确定。 16. The application of one of the above claim application, which in the case of a package with more than one person to occupy wait in long lines, and the next packet according to a second criterion to determine.
17.如上面权利要求的应用,其中第二准则是作为FIFO形成的。 Application, wherein the second criterion 17. The above claims are formed as FIFO.
18.如上述应用权利要求之一的应用,其中等候者长队只是在需要时才被开辟并将长时间不使用的等候者长队重新交付使用。 18. The use according to claim one of the applications in which the waiting in long lines to be opened only when needed and the waiting in long lines will not be used to re-delivery.
19.在一个面向包的通信网络(100,100a,200)中传送业务流(VS)的方法,应用在一个按照上述装置权利要求之一的网络节点(1-15)中,以下列步骤:-将到来的包按照其对分配扇(VF)之一的从属,登记到一个相应的等候者长队(Q)中,和-经过分配给分配扇(VF)的那个束(B)的端口(P)之一传送,其端口通过分支图形确定,分配扇映像到分支图形上。 19. The method of transmitting traffic in a packet-oriented communication network (100,100a, 200) in (VS), and used in a device according to one of the above claims network node (1-15), the following steps: - The arrival of the package in accordance with its subordinate to (VF) assigned one fan, registered to a corresponding length of the waiting team (Q) in, and - after distribution to distribution fan (VF) of the bundle (B) of the port (P) transferred one of its ports through branch pattern determination, allocation fan image to branch graph.
20.如上述方法权利要求之一的方法,其中要将一个分支组的业务流有目的不均匀地经过分配的束的端口传送。 20. A method as claimed in one of the above methods of claim traffic to a branch of the group which has port transmit beam aim through unevenly distributed.
21.如上述方法权利要求之一的方法,其中尤其是在部分重叠的束时,一个接近于一个所期望的分配目标数值,通过与调度-规则的适应匹配产生。 21. The method as claimed in one of the above claims, wherein especially in partially overlapping beams, one close to a desired target value assigned by the scheduling - adaptation rules match generation.
22.如上面权利要求的方法,其中将要求的匹配事先算出,并在分配调度-规则时考虑。 22. A method as claimed in the above claims, wherein the matching requirement is calculated in advance, and scheduling - consider rules.
23.如权利要求19至21之一的方法,其中调度-规则在通信网络连续的运行中要适应地匹配。 19-21 23. The method of claim, wherein the scheduling - match rules to adapt to the continuous operation of the communication network.
24.如上面权利要求的方法,其中调度-规则的匹配在网络节点中自主和/或通过外部的触发进行。 24. A method as claimed in the above claims, wherein the scheduling - matching rules in network nodes autonomy and / or by an external trigger.
25.如权利要求23或24之一的方法,其中在考虑到一个调节的准则情况下,尤其是在实际达到的分配与所期望的分配目标值之间比较时,要产生匹配结果。 25. The method of 23 or claim 24, wherein in the case of taking into account a criterion adjustment, especially between the distribution actually achieved and the desired target value assigned when compared, to produce a match was found.
26.如上面权利要求的方法,其中为了评价准则,要测量操作不同等候者长队的频繁程度和每个业务等级端口造成的载荷。 26. A method as claimed in the above claims, wherein the evaluation criteria in order to measure the frequency of different operating load the waiting in long lines, and each traffic class port caused.
27.如权利要求23至26之一的方法,其中为了网络范围内的修正,要测量不同等候者长队的平均和/或最大出现的长度。 23-26 27. The method of claim, wherein in order to correct within the network, and to measure the average and / or maximum length of the waiting longer appear different teams.
28.如权利要求26或27之一的方法,其中在包的数量的基础上进行测量。 28. The method of 26 or claim 27, wherein the number of packets on the basis of measurements.
29.如权利要求26或27之一的方法,其中在考虑单个包的单独长度的情况下进行测量。 29. The method of 26 or claim 27, wherein in the case of considering a single packet length is measured separately.
Description  translated from Chinese
通信网络中业务流的分配传输 Allocation of transmission in a communication network traffic flow

本发明的内容涉及在通信网络中业务流分配传输领域。 The present invention relates to the field of distribution of transport traffic in a communication network. 通信网络能是在专利申报DE10161508.6中公开的方式,它的公开涉及到本说明书的内容。 Communications network could be disclosed in the patent application DE10161508.6 way, its disclosure relates to the contents of this manual. 此外,专利申报DE10161547.7的公开也涉及到本说明书的内容。 In addition, patent application DE10161547.7 disclosure also relates to the contents of this manual.

在一个通信网络中,要将业务按照一定的规则尽可能均匀地分配到所有节点和通信网络-也称为“网络”-中的所有连接线路。 In a communication network, to business as evenly distributed according to certain rules to all nodes and communication networks - also known as "network" - all connected lines.

在这种方式的分配时,为每个从一个一定的输入端A到一个一定的输出端B的通信关系产生一个所谓的“分配扇”,它包括所有对这个通信关系有用的节点和连接线路(参见图1及在DE10161547.7中所属的通道)。 In this way the allocation for each generate a so-called "distribution sector" from a certain input terminal A to a predetermined communication relationship between the output terminal B, which includes all of the communication nodes and links useful relationship line (see Figure 1 and in DE10161547.7 belongs in the channel). 在一个相应分支的网络中,不同通信关系的分配扇不可避免地要重叠,使在各个节点上出现不是相同的就是部分重叠的或完全对立的“分支图形”(参见图2)。 In a corresponding branch network, different communication relations distribution sector inevitably overlap, so that is not the same or partially overlapping completely opposite "branch Graphics" (see Fig. 2) appear on each node. 重叠取决于在网络节点中要使用的分配机制。 Overlapping distribution mechanism depends on the network node to be used.

迄今,知道用于单独分配数据包到出去的导线束上的下列机制:1)到一个出去的导线束上没有优先权简单分配到来的业务:(a)在每个端口的单独等候者长队中事先分配业务:一个中央业务分配器将到来的业务分配到各个等候者长队上,从中给导线束的一个出去的端口准确地各分配一个等候者长队。 To date, the following mechanisms alone knows for distributing data packets to go out on the wire harness: 1) to a simple assignment does not have priority on coming out of the wiring harness business: (a) separate the waiting in long lines for each port the prior distribution business: a central business distributor of business will come assigned to each person waiting in long lines, an outgoing port from which to accurately strands are each assigned one person waiting in long lines. 该分配能周期性地或(例如在不同的端口带宽时)按照不同准则加权地进行。 The allocation can be periodically or (e.g., in a different port bandwidth) in accordance with different criteria to be weighted. 此时,例如也能顾及到各个等候者长队(按照包的数量或在以字节为单位的可变包长度时按照真实数据量)的现实填满状态,或也能顾及到现实要分配的数据包的单独长度。 In this case, for example, can take into account each person waiting in long lines (according to the number of packets or at variable packet length in bytes of data in real) reality fill the state, or can take into account the reality to be allocated The individual length of the packet. 端口一般按照FIFO-原理操作等候者长队。 Generally in accordance with the principles of operation of the port FIFO- the waiting in a long queue. 在一个长队里有过一次的东西,就必须由所分配的端口做完。 In a long team had a thing, it must be done by a port allocation.

(b)应用一个多服务器原理的唯一的等候者长队:能将同时最佳利用可供使用端口容量的一个有利的分配,用多服务器原理实现。 (B) applying a multi-server principle the only person waiting in a long queue: can simultaneously optimal use of available capacity is advantageous to use a port assignment, with multi-server principle to achieve. 此时,将所有到来的数据包调整到一个唯一的等候者长队中,总是空闲的或将要空闲的端口,从该等候者长队中通常按照一个FIFO(先入先出)原理取出一个要操作的包。 In this case, all the incoming data packet is adjusted to a single person waiting in long lines, and always will be a free port or free, from which the waiting in a long queue usually follow a FIFO (first in first out) principle to remove a package operation.

2)到一个出去的导线束上用优先权分配到来的业务:(a)在每个端口的单独优先等候者长队中事先分配业务:一个中央业务分配器将到来的业务分配到各个等候者长队去,此时为每个优先权等级的每个出去的端口各准备一个单独的等候者长队。 2) to a harness out with the arrival of business prioritization: (a) the prior distribution operations in a single person waiting in a long queue priority for each port: one central business distributor will come to all the waiting traffic distribution long lines to go, then wait for each individual who prepared a long queue for each outgoing port for each priority class. 按照1(a)的方案可相似地应用。 According to 1 (a) program may be similarly used. 在操作等候者长队时,端口按照相应的规则(“严格的”,“加权的”等)考虑到优先权。 In operation the waiting in long lines, port in accordance with the appropriate rules ("strict" and "weighted", etc.) taking into account the priority.

(b)每个优先权等级各有一个等候者长队的多服务器原理:如1(b),此时优先权高的等候者长队按照优先权规则优先操作。 Multi-server principle (b) for each priority level who have a long line waiting: If 1 (b), a high priority at this time the waiting in a long queue in accordance with the priority rules of priority action.

3)到一个出去的导线束上用优先权控制的按流排队分配到来的业务:上述基本机制的细化在于,为了在不同单独通信关系(流)之间的粒化区分和优先化,也要调整每个流的单独和分开的队列。 Press the flow line distribution services on arrival 3) out of the harness to a control with priority: to refine these basic mechanism is that, in order to communicate the relationship between individual (stream) granulated differentiation and prioritization, and To adjust individually and a separate queue for each stream. 因而用于排队(wg.等候者长队的数量)和调度(wg.从大量等候者长队中选择下一个要输出的包)的花费,要提高数倍,且更取决于业务图形(即同时有效的流的数)。 Thus for queuing (wg. Number who wait in long lines), and scheduling (wg. Choose from a large number of those waiting in long lines to be output in the next package) costs, to increase several times, and more dependent on business graphics (ie At the same time the number of valid stream). 此外,此时还必须注意在相同优先化的流之间的资源公平分配,为此使用像例如“加权的公平排队”(WFQ)或相似的专门的机制,它们的复杂性(特别是在很大数量的队列时)能将一个简单的优先权排队提高数倍。 Also in this case the same attention must also be prioritized in resource flows between the equitable distribution, for use as, for example, "Weighted Fair Queuing" (WFQ) or similar dedicated mechanisms, their complexity (especially in the very When a large number of queues) a simple priority queuing can increase several times. 当然,按流排队既可以端口单独地[上面的图形(a)]也可以与多服务器原理[上面的图形(b)]结合使用。 Of course, according to the flow line either port alone [above graph (a)] can also be used in conjunction with multi-server principle [above graph (b)].

4)ECMP(Equal Cost Multiple Path等价多路径):ECMP到数个端口上预先规定一个分配。 4) ECMP (Equal Cost Multiple Path Equal Cost Multipath): ECMP to a predetermined number of port allocation. 在此只考虑一个传送节点的一些端口,这些端口连接的物理线路要通向同一相邻的传送节点。 Here only consider a transfer node some ports, physical lines connected to these ports leading to the same neighboring transmitting node. 以这种方式(即通过负载分配到数个物理线路),如果一个已有的唯一物理线路的传送能力不能继续提高,两个相邻传送节点之间的传送能力也能提高。 In this manner (ie by load distribution to several physical lines), if an existing transmission capacity is not the only physical lines continue to improve the transmission capacity between the two neighboring transmitting node can be improved.

对于所有列举的方法,在等候者长队中执行时,一般为了识别当时的数据包,在一个通常是共有的数据存储器中只占据一个指示器(地址)。 For all the methods enumerated in the implementation of the waiting in long lines, in general, in order to identify the time of data packets in a data memory is usually shared only occupy a pointer (address). 操作的顺序内隐地从等候者长队(例如按照FIFO原理)中登记的顺序得出,或从为选择作为下一个要操作的等候者长队(例如按照优先权和在相同优先权时例如周期性地,longest queue first最长队列第一,shortest queue first最短队列第一,按照加权如在WFQ时)预先安置的方法得出。 Hidden within the sequence of operations from the waiting in a long queue (for example in accordance with the FIFO principle) registered in order to draw, or to choose from as the next person waiting to be operated long lines (such as for example in accordance with the priority period and at the same priority of land, longest queue first longest queue first, shortest queue first shortest queue first, when the weighted as WFQ) method of pre-arranged results.

为了达到附加的专门效应,能将其它信息纳入到这个调度-决定中。 In order to achieve additional special effects, additional information can be incorporated into the schedule - decisions. 特别是在ATM技术中,经常(在IP范围也偶尔)谈到一个传输流整形(Traffic Shaping)的必要性。 Especially in ATM technology, often (in IP range occasionally), referring to the need for a transfer of traffic shaping (Traffic Shaping) of. 用这个方法,一般要遵照一定的带宽,这些带宽大多通过一个连接(即通信关系)的信元(包)之间相应的距离准则实现。 The distance between the corresponding guidelines to achieve using this method, generally follow a certain bandwidth, which bandwidth is mostly through a connection (ie communication relationship) cell (packet). 为此存储有附加的时间信息,它们为一个等候者长队或一个一定的信元(包)给出一个最早的、最迟的和/或最佳的操作时间点(所谓“日历”)。 Storing the information for this additional time, they were a long queue or waiting a certain cell (packet) gives one of the earliest, at the latest and / or optimal operating point in time (called "Calendar").

机制也可以用相同的作用扩展到多个导线束上去应用,只要这些导线束既根本不重叠(是分离的)或完全重叠(是相同的)就行。 Mechanism can also be extended by the same action applied to a plurality of strands up, as long as these do not overlap both strands (is separated) or completely overlap (the same) on the line. 为了一个公平和高效、以包的方式考虑到优先权到部分重叠的导线束上的业务分配,这样一个解决方案尚不为人知。 For a fair and efficient way to pack assigning a priority to take into account business section on overlapping strands, such a solution is still unknown.

本发明的任务在于,如何能在网络节点中,在考虑一个也许预先规定的各个业务流或单独数据包的优先化的情况下,将业务按照预先规定的分支图形,尽可能最佳地分配到出去的连接线路上,此时每个节点要自主和单独决定每个数据包。 The task of the present invention is that, how can the network nodes, in consideration of the circumstances under priority one might predetermined flow of each business or individual packets, traffic branch in accordance with a predetermined pattern to the best possible allocation the outgoing connection lines, then each node to be independent and separate decisions for each packet.

这个任务通过本发明来解决。 This task is solved by the present invention. 本发明预定,在一个按照本发明的节点中,给每个现实的分支图形至少各预先规定一个等候者长队。 The present invention can book in a node according to the present invention, for each of the realities of each branch graph least a predetermined length of the waiting team. 此时,一个等候者长队完全能为多个分配扇共同使用,这些分配扇在这个节点中反映到同一个分支图形上。 In this case, a person waiting in long lines fully able to use multiple distribution sectors together, these allocations reflect the fan to the same branch in the node graph.

将到来的包按照其归属于某个分配扇(即不是按照归属于分配扇的某个流)来区分。 The arrival of the packet belongs to an assignment in accordance with its fan (ie not attributable to the distribution sector in accordance with a certain stream) to distinguish. 相应地要将它们登记到一个所属的等候者长队中。 Accordingly, those who want to register them to wait in long lines one belongs.

本发明的一个与特别美好的优点连接在一起的方案,由一个样板式的实施得出,此时等候者长队由属于当时分支图形的端口操作。 Program of the present invention with particular advantages of better connected together by a boilerplate implementation drawn, then the waiting in long lines belonging to the then branch pattern port operations. 只要有一个端口空闲,它就选出作为下一个要由它操作的等候者长队。 As long as there is a free port, it is selected as the next person waiting to be operated by its long lines. 选择开始的要及时,使在出去的线路上没有空隙出现。 Select Start to timely, so that no gaps appear in the line out. 如个一个端口操作多个(在这个情况是完全或部分重叠的)分支图形,则在这个选择时首先经过所有这些分支图形寻找不是空白的等候者长队。 As a one port to operate multiple (in this case is entirely or partially overlapping) branch graph, then this selection first graph search through all of these branches were not blank wait in long lines. 如果从不同的分支图形找到多个不是空白的等候者长队,则按照一个预先规定的准则决定,将哪个等候者长队要作为下一个被操作。 If more than one person is not blank wait in long lines from different branches of the graphic, in accordance with the guidelines of a predetermined decision to which the waiting in long lines to be operated as the next. 要从这个等候者长队出去的端口就按照另一个准则来决定。 The team from the waiting long out of port in accordance with another criterion to decide.

有利的是,在这时所需要的队列数量是有限制的,完全取决于网络的拓扑学,即取决于在考虑端口成束的情况下相邻节点的数量,但不取决于业务。 Advantageously, in this case the required number of queues is limited, depends entirely on the topology of the network, ie on a bundle in the case of the port adjacent to the number of nodes, but does not depend on the business.

在混合网络中使用时毫无问题,因为本发明只是局部地在一个给出的网络节点使用。 When used in a hybrid network without any problem, because the network node of the present invention is only partially in a given use.

本发明大为有利的发展,由本发明的下面样板性的实施得出:-为了延迟最佳化,在一个分支图形内,预先规定多个、用不同优先权标志的等候者长队。 Greatly advantageous development of the invention, by the following model of the embodiment of the present invention is drawn: - In order to optimize the delay, in a branch graph, a plurality of predetermined, with different priorities who sign long lines waiting. 一个分支图形内的等候者长队,严格按照其优先权来处理。 A branch of graphics within the waiting in long lines, in strict accordance with its priority to deal with. 如果一个端口操作多个(在这个情况是完全或部分重叠的)分支图形,则首先经过所有这些分支图形寻找最高优先的、不是空白的等候者长队,这个等候者长队要被操作。 If a port to operate multiple (in this case is entirely or partially overlapping) branch graph, first through all these branches graph to find the highest priority, not by a long queue waiting blank, that's a long queue waiting to be operated on. 如果此时从不同的分支图形找到多个相同优先权不是空白的等候者长队,则要在这些等候者长队之间按照一个可自由选择的准则进行裁决(Arbitrierung),例如按照随机原则,按照一个周期性顺序,按照束的大小,按照等候者长队长度(longest/shortest first),按照对其操作负责的端口数量。 If at this from a different branch of the same graphic found more than one priority person is not blank wait in long lines, who will have to wait long between these teams award (Arbitrierung) in accordance with the guidelines of a freely chosen, for example in accordance with the principle of random, according to a cyclical order, according to the size of the beam, in accordance with the waiting in a long queue length (longest / shortest first), is responsible for its operation in accordance with the number of ports.

通过使用优先化的等候者长队,以可比较地较少花费导致有一个足够好的延迟和分配质量。 By using the priority of the waiting in long lines to be comparatively less expensive to have a good enough cause delays and distribution quality.

-为了在每个包的最小可能的延迟上的最佳化,同样在一个分支图形内,预先规定多个、用不同优先权标志的等候者长队,它在一个分支图形内严格按照其优先权来处理。 - In order to minimize the possible delay of each packet optimization, also within a branch graphics, a predetermined plurality of different priority mark the waiting in long lines, it is a branch of graphics in strict accordance with their priorities the right to deal with. 如果一个端口操作多个(在这个情况是完全或部分重叠的)分支图形,则首先经过所有这些分支图形寻找最高优先的、不是空白的等候者长队,这个等候者长队要被操作。 If a port to operate multiple (in this case is entirely or partially overlapping) branch graph, first through all these branches graph to find the highest priority, not by a long queue waiting blank, that's a long queue waiting to be operated on. 如果此时从不同的分支图形找到多个相同优先权不是空白的等候者长队,则在这些之间按照一个时间准则来决定。 If at this from a different branch of the same graphic found more than one priority person is not blank wait in long lines, then between them according to a time criterion to decide. 为此,在等候者长队中,除了一个存储指示器(地址,见上面)外还放置一个附加的,优先是相对的时间信息(“TimeStamp”时间标记),从中能读出包在何时登记到等候者长队中和它在里面已经多长时间(图4)。 For this reason, the waiting in long lines, in addition to a memory pointer (address, see above) outside also place an additional priority is relative time information ("TimeStamp" time stamp), which when read package Registration to the waiting in a long queue and how long it has been on the inside (Fig. 4). 然后选出等候者长队,在它的首位是已经早就等待着的包,由此一个FIFO原理要作为准则实现。 Then select the waiting in long lines, in its first place is already long waiting bag, whereby a FIFO principle to be used as guidelines for implementation.

用这种结构有利地确保,用每个要空闲出来的端口,即使在只是部分重叠的导线束时,越过所有从这个端口操作的分支图形,永远要操作具有最小可能延迟的瞬时正好是最紧迫的要求。 With this arrangement advantageously ensures, with each port to free up even only partially overlap when the harness across all branches of the graphics from the port operation, always want to operate with the smallest possible delay is just the most urgent instantaneous requirements. 如果相互配合的实时服务,例如电话-或电视会议服务,要作为高优先权的业务流传送,这就特别有利。 If the mutual cooperation of real-time services, such as phone - or videoconferencing services, to be used as a high priority streaming traffic, which is particularly advantageous.

-包的与众不同的时间准则,不是在所有等候者长队时应用,而只在高优先权的业务(例如时间关键性的实时应用)时应用,以便将这个业务用最小可能的延迟和相应公平的分配来办理。 - Different time guidelines package, not all the waiting in a long queue application, but only when used in high-priority traffic (such as time-critical real-time applications), so this business with the minimum possible delay and equitable distribution corresponding to handle. 能将服务于没有专门延迟要求或简单只是最佳努力(Best Effort)业务的等候者长队,以较低的优先权且不考虑这个时间信息来处理。 It can serve no special latency requirements or simply just best effort (Best Effort) business person waiting in long lines, at a lower priority and do not consider the time to process the information. 原则上,只将时间信息为相应高优先权的等候者长队存储和处理就够了。 In principle, only the time information is correspondingly high priority by long queues waiting to store and process enough.

时间信息的粒度(Granularitaet)要考虑到一个包的最小传送时间,即在束的最快的线路上最短可能的包的传送时间。 Granularity time information (Granularitaet) to take into account the minimum delivery of a package that bundles on the fastest route shortest possible delivery package. 可能的数值范围要选择的,使在一个相应高优先权的包所期待的最大延迟时间内,不出现溢出(考虑到所属队列的最大长度、端口带宽等),并随时有可能是一个安全的决定。 The possible value range to be selected so that a correspondingly high priority within the expected maximum packet delay time, there are no overflow (taking into account the maximum length of the queue belongs port bandwidth), and at any time there may be a safe decision. 此时,高优先权业务一个可能调整过的极限能在网络中用较短等候者长队的优点考虑到。 In this case, a high priority may be adjusted business can limit the advantages are shorter waiting in long lines with the network considered.

执行能经过一个简单的、有规则提供时钟脉冲的圆形计数器(在溢出时重新起动)进行。 Execution can go through a simple, there are rules to provide round the clock pulse counter (restart on an overflow) performed. 计数器只是局部重要,节点之间的同步是不需要的,节拍精确度没有专门要求,带与不带在相同网络中的这个机制的节点混合随时有可能。 Counter is only a partial importantly, the synchronization between nodes is not required, there is no specific requirements for the accuracy of the beat, the node with this mechanism is not in the same network with mixed may at any time. (机制只涉及一个节点内的个别、局部的工作方式。)-为了将由于某些原因长时间等待的和继续传送(在所属的实时应用范围中)看来毫无意义的包挑出和抛弃,为此也应用时间准则。 (Mechanism involves only the individual, the local way of working within a node.) - To a long wait for some reason and continues to transmit (in a range of real-time applications belongs) It seems pointless to pick out and discard the packet , for this time the guidelines also apply. 这或者能在调度时完成,或者能通过一个分开的、平行进行的过程完成。 This is when scheduling is completed or can, or through a separate, parallel process to complete. 这样一个过程例如能将时间信息在包登记到等候者长队中时置上一个规定的起始值,时间信息有规则地向上或向下数提供时钟脉冲,在达到一个一定的极限-或阈值时触发相应的动作,例如从等候者长队中将包清除。 Such a process such as the time information in the package can register to the initial value on the waiting in a long queue set when a specified time information regularly provided downward the number of clock pulses up or reaching a certain limit - or threshold When the action is triggered, for example, remove the package from the waiting in a long queue will. 可选择的是,也能将登记时间点(作为不变的值)与现实时间(把它也计算在内)之间的差,在一个规则性重复比较中作为定夺准则提出。 Alternatively, it can register time (as a constant value) and real time (it is also included) the difference between a regular repeating in comparison presented as decided criteria.

-端口按照多服务器原理操作等候者长队。 - Port in accordance with the principle of multi-server operations are waiting in long lines. 多服务器原理也保证有个公平的分配和最佳利用可供支配的资源,尤其是在可变包的长度时。 Multi-server principle also ensure a fair distribution of resources and the best use of their disposal, especially when in a variable-length packets. 如果出去的端口空闲,它就正好在此时启动调度过程。 If you go out of the port it is idle, it just starts the scheduling process at this time. 但是这发生的(在了解瞬时被操作包的(剩余-)长度的情况下)如此及时,使决定和下一个要输出的包及时随着前面包的结束为输出作好准备。 But it happened (is operated package (remaining in-the-moment - the case) length) is so timely, so decided and the next package to be output in time with the end of the previous package ready for output. 多服务器原理的另一个大优点在于,在不同带宽的端口时,由于要空闲出的端口可以说是自己从队列中取出新的数据,这些决定在分配时自动被考虑到。 Another big advantage of multi-server concept is that, at different bandwidths port, due to free the port can be said to take out their own new data from the queue in the allocation of these decisions are automatically taken into account.

-作为下一个要由一个端口操作的等候者长队的选择,借助于一个调度-功能完成。 - As the next person waiting to be used by the operation of a port of choice in long lines, by means of a scheduler - function is completed.

-为了考虑到附加的、一旦有必要的话每个分配扇不同的准则,要使用其它分开的等候者长队。 - To take into account additional, if necessary, once each distribution sector different criteria, to be used by other separate waiting in long lines.

-另外的准则是一个FIFO(先入先出)。 - Another criterion is a FIFO (first in first out).

-等候者长队只有在需要时才投入。 - The waiting in long lines only when the need to invest. 要将长时间没有利用的队列再交付使用(time-out)。 To not take advantage of a long queue before delivery (time-out). 经此,将实际需要的等候者长队的数量降下来,因为现实分支图形的数量大多要小于现实通信关系的数量。 After this, the number of those who actually need to wait in long lines down, because the number of the most realistic graphics branch is less than the number of real communication relationships.

-业务流是有目的不均匀地按照相应的预先规定来分配。 - Traffic is unevenly purpose in accordance with the respective predetermined allocated. 为此,将其它准则一道纳入到分配决定中。 To this end, other criteria into a distribution decisions.

-给分配预先规定一个适应近似所期望的目标数值。 - To allocate a predetermined target value approximate the desired adaptation. 在部分重叠的束时有可能的是,努力争取的均匀或预先规定的分配没有最佳地起作用,因为束之间的干扰起了阻碍作用。 Possible when partially overlapping beams is assigned a uniform or predetermined striving not function optimally, because of interference between beams from the impediment.

有一个简单的选择可能性,将它预先计算并在预定调度-规则时予以考虑。 There is a simple possibility to choose, it would be calculated in advance and at a predetermined schedule - be considered rules.

在复杂的情况时(例如在这中间总是在改变着实际网络条件),在其中网络状态和因此本发明组成部分的状态不能可靠地预先计算,调度-参数在操作时能适应地匹配是有优点的。 When complex situations (for example, are always in the middle of changing the actual network conditions), part of the state can not be reliably calculated in advance in which the network status and therefore the composition of the present invention, the scheduling - parameter matching can adapt to when operations are advantages.

这或者在节点主动完成,或者通过来自外部(预先规定新规则)的触发完成。 This is either done automatically in the node, or by a trigger from the outside (a prescribed new rules) is completed.

为此能将实际实现的分配与(从)所期望的分配的比较(各自的偏差)用作为调节的准则。 Comparison of the actual distribution can be achieved and for this (from) a desired distribution of the (respective deviation) is used as a criterion adjustment. 优先要为此在节点中例如测量通过不同端口(或每个业务等级端口产生的载荷)不同等候者长队操作的频繁程度,并可提供要修正的装置使用。 Priority To do this, for example by measuring how often a different port (or ports for each service class load generated) by different waiting in long lines operated in the nodes, and provides means to use to fix.

对于网络范围的修正,有关不同等候者长队出现的平均和最大长度的说明也能有意义。 For correction network-wide average and instructions regarding the maximum length of the waiting in long lines appear different can also be meaningful.

优先的是,将这些测量和方法在包的数量基础上进行。 Priority is the number of these measurements and methods on the basis of the package were. 在这种情况的出发点是,使可变长度包的作用经过相对短的时间去查证。 In this case the starting point is to make the role of variable length packets after a relatively short period of time to verify. 可选择的是,也能将这些观察和测量在考虑到所有各个包的独特长度的情况下布置。 Alternatively, they can also observe and measure the length, taking into account the unique circumstances of each packet of all arrangements.

-为了在一个束内实现一个一定的载荷分配,为每个端口的每个队列保持一个时间信息,并在调度时随之考虑。 - In order to achieve a beam of a certain load distribution, a holding time information for each queue per port, and subsequently when scheduling considerations. 时间信息包括例如通过不同端口队列的最后操作时间点。 Time information includes, for example through a different port queues final operation point in time.

有利的是,就此例如端口独特地能调节两个操作之间一个一定的最小-和/或最大载荷距离,或能将时间相对地互相评定(“如果端口x等待双倍长的时间没有像端口y那样轮到,才再轮到它”)。 Advantageously, in this example, a unique port can be adjusted between the two operations a certain minimum - and / or the maximum load distance or time can be assessed relative to each other ("If the port x double waiting a long time did not like the port y that turn, only then turn it "). 也能将这样的时间信息应用于确定一个下一个目标-操作时间点(最早的,最迟的,最佳的)或立即作为这样的存储起来(日历)。 Such time information can also be used to determine a next target - operating time (the first, at the latest, the best) or stored immediately as this (calendar).

一个可选择的和比较简单的解决办法在于,为每个队列只是去数它被什么样的端口操作多少次,从中推导出决定,是否可以将它现实地随着考虑进去。 An alternative and relatively simple solution is that it is just going to count what kind of port operation many times for each queue, which is deduced to decide whether it can be taken into account as realistically. 能将计数器以间隔方式或滑动地(Leaky Bucket漏桶)回置或缩减。 Can counter at intervals or by sliding (Leaky Bucket Leaky Bucket) return home or downsize.

在这种可能性时,如上面所述的预先规定和规则,也能用测量结果适应地去匹配。 When this possibility, such as rules and regulations previously described above, the results will also be able to adapt to match the measurements.

本发明也用在图中示出的实施例进一步说明。 Embodiments of the present invention is also used in the figures shown further instructions. 这里示出的有:图1一个第一个分配扇,它包括有所有在一个网络中从一个发送节点A到一个接收节点B为一个通信关系可合理使用的网络节点和-边,图1a一个第二个分配扇,它包括有所有在网络中从一个发送节点C到一个接收节点D为一个通信关系按照图1可合理使用的网络节点和-边, Here it is shown: Figure 1 a first dispensing sector, which includes all transmitted from a node in a network A to a node B receiving a communication relationship can be used reasonably network nodes and - side, FIG. 1a a The second distribution sector, which includes all the network from a sending node to a receiving node C and D is a communication relationship network node according to a rational use of graphs and - Edge,

图2两个部分重叠的分配扇,在网络中按照图1共同表示出,图3一个本发明解决办法的基本形式(还没有“时间标记”),对于本发明的最基本的结构用n=1,此时箭头表示端口取下一个出去的包,划出的三角是为了此时(从端口出发)被激活的调度功能。 Figure 2 two partially overlapping distribution fan, in the network shown in Figure 1 together, FIG. 3 of the present invention is a solution to the basic form (not yet "time stamp"), the basic structure of the present invention with n = 1, then the arrow indicates that the port is removed out of a bag, draw the triangle is to this point (from the port of departure) is activated scheduling features.

图4一个本发明的解决办法,用时间标记信息,为了最高优先权队列的包(时间标记原理当然也能为了直到最佳努力为止的低优先权队列使用),最高优先权队列由调度随着评定,此时箭头表示端口取下一个出去的包,划出的三角是为了此时(从端口出发)被激活的调度功能。 Figure 4 a solution of the present invention, a time-stamp information, to the highest priority queue packet (timestamp principle can certainly until the best effort so far to lower priority queues), with the highest priority queue by the scheduler assessment, then the arrow indicates that the port is removed out of a bag, draw the triangle is to this point (from the port of departure) is activated scheduling features.

图5一个本发明的解决办法,用每个队列的其它辅助记录器(用辅助信息在分支组要操作的束中为当时的分支组控制一个预先规定的业务分配),此时箭头表示端口取下一个出去的包,划出的三角是为了此时(从端口出发)被激活的调度功能。 Figure 5 a solution of the present invention, with each of the other secondary recording queue (using auxiliary information in bundle branch group was to be operated as a branch of the group controlling a predetermined service distribution), then the arrow indicates the port to take Next out of the package, in order to draw the triangle at this time (from the port of departure) is activated scheduling features.

在图1中,表示的是在通信网络100中一个分配扇VF1的一个例子。 In Figure 1, it shows an example of a communication network 100 of VF1 distribution sector. 网络100包括有多个网络节点1-15,节点1-7是作为输入-和/或输出节点形成的。 Network 100 includes a plurality of network nodes 1-15, the node 1-7 as input - and / or output node is formed. 从存在着的作为发送节点A形成的网络节点1开始,至少将一个业务流VS1如下面分配地传送到存在着的作为接收节点B形成的网络节点4: Starting from the existence of a network node as a sending node A is formed, at least one assigned traffic VS1 as will be transmitted to the existence of a network node as a receiving node B 4 formed:

容易看出,从节点A与B之间的每个网络节点,有一个以上的剩余路径以它为出发点向接收节点B,该节点向这个发送的业务在至少两个剩余路径上分配发送。 It is easy to see from each network node between nodes A and B, between there is more than one remaining path to it as the starting point to the receiving Node B, an allocation of the node that is sending to the service on at least two remaining paths. 对于节点1、8、9、10、11、13和14,中间那列相当于各自边的内部的分支图形,分配扇VF1到其上形成。 For a node 1,8,9,10,11,13 and 14, the middle column corresponds to the respective branch inside edge graphics, assigned to fan VF1 formed thereon. 对于网络节点3、5、12和15,由于缺少不同的网络节点,不存在能到其上形成分配扇VF1的真实分支图形。 For network nodes 3,5,12 and 15, due to the lack of different network nodes, there is no divergence that can be assigned to a real fan VF1 formed thereon.

在图1a中,表示的是在通信网络100中一个分配扇VF1的第二个例子。 In Figure 1a, it shows a communication network 100 in an assigned fan VF1 second example. 从存在着的作为发送节点C形成的网络节点7开始,至少将一个业务流VS1a如下面分配地传送到存在着的作为接收节点D形成的网络节点3: Starting from the existence of a network node as a transmission node C 7 formed, at least one service stream VS1a dispensing transmitted to the existence of the following network nodes is formed as a receiving node D 3:

又很容易看出,从节点C与D之间的每个网络节点,有一个以上的剩余路径以它为出发点向接收节点D,该节点向这个发送的业务在至少两个剩余路径上分配发送。 And very easy to see from each network node between C and D, there is more than one remaining path to it as the starting point to the receiving node D, the node sends traffic to this assignment sent on at least two remaining paths . 对于节点7、9、10和14,中间那列相当于各自边的内部的分支图形,分配扇VF1a到其上形成。 For a node 7,9,10 and 14, the middle column corresponds to the respective branch inside edge graphics, assigned to fan VF1a formed thereon. 对于网络节点11、12和13,由于缺少不同的网络节点,不存在能到其上形成分配扇VF1a的真实分支图形。 For network nodes 11, 12 and 13, due to the lack of different network nodes can not be assigned to fan VF1a on which branch of the true pattern is present.

但是,在这种情况要注意,在图1、1a和2中的分支图形示出的只是节点之间连成闭合网路的结构。 However, in this case, to note that between FIG. 1,1a and 2 graphically illustrates just branch node connected to a closed configuration networks. 这并不排除在图形的一个边上(即在两个相邻的节点之间)能有多个平行的物理导线(甚至也用不同的带宽),这些导线在分配时(能)像在图3至5那样作为分开的端口看待。 This does not exclude an edge pattern (i.e. between two adjacent nodes) can have a plurality of parallel physical wires (even with different bandwidths), which leads in the distribution (energy) as in FIG. 3-5 as viewed as a separate port. 为了到一个边的数个端口上分配,当然很可能有(不真实的)分配扇存在。 In order to assign a number of ports on one side, of course, is likely to have a (false) is assigned fan presence. 如果也没有因此到不同的网络节点1-15上导致业务的力争的分配,则到这些(不真实的)分配扇上基本上同样能反映出分配扇VF。 So if there is no on different network nodes 1-15 lead to strive distribution business, then these (false) on substantially the same allocation reflect the distribution fan fan VF.

最后,图2中的两个分配扇VF1、VF1a共同表示在通信网络中,这样就容易看出,两个分配扇VF1、VF1a是如何在它们的网络节点9和10反映到相同的分支图形i上,如何在它们的网络节点11、13和14反映到部分相同的分支图形t上和如何在它们的网络节点12反映到对立的分支图形d上,如下面的表说明的那样,在其中将不同的边用黑体字标记: Finally, Figure 2 two dispensing sectors VF1, VF1a together represent in a communication network, so it is easy to see, two distribution fan VF1, VF1a is how they reflect the network node 9 and 10 to the same branch graphics i on how their network nodes 11, 13 and 14, reflected on the part of the same branch graphic t and how they are reflected in the network node 12 d on the opposite branch graph, as explained below in the table, in which the Different side with the bold tag:

这些关系在下面的表中再一次用特别强调不同边的情况表示出: These relationships in the following table once again with particular emphasis on the situation shows different sides:

本发明的一个特别美好的优点在于,在使用重叠的分支图形t时,通过机制的任意组合a)为了延迟最佳化在裁决时每个包的时间准则,b)调整一个预先规定的、需要时“偏斜的”业务分配,c)适当的再调整到力争的分配图形,使一个很灵活的和为几乎每个网络应用网络节点1-15的可最佳化的适配和因此网络100的可最佳化的适配作为整体成为可能。 A particularly good advantage of the invention is in the use of overlapping branches graphic t, through any combination of mechanisms a) In order to optimize the time delay of each packet guidelines when ruling, b) a predetermined adjustment needed When "skewed" distribution business, c) appropriate readjustment to strive assignment pattern to make a very flexible and can be optimized for virtually every network application network node adapter 1-15 and so the network 100 The adapter can be optimized as a whole as possible.

借助于将其它准则纳入到分配决定中,按照相应的预先规定,一个有目的使业务流VS不均匀分配的一个例子在于,在对某些端口P按照相应的准则调度时,偶尔没有考虑到某个等候者长队Q。 By means of other criteria will be incorporated into the allocation decision, in accordance with the corresponding predetermined, a purpose to make an example of the uneven traffic distribution VS is that when certain ports P in accordance with the guidelines scheduling, occasionally not take into account certain a person waiting in a long queue Q. 如果例如在一个分支组的高-或最高优先权业务的四个端口P上分配时,以4∶4∶3∶1的比例分配分配到四个端口P上,则对于第三个端口P调度程序没有考虑到这个分支组的最高优先权的等候者长队,对于第四个端口它只是在每个第四个调度-过程时随着被包括在内。 If for example, a branch of the group in high - the highest priority when allocating or services on four ports P, with proportional allocation 4:4:3:1 assigned to the four ports P, then the third port P scheduling the program does not take into account the highest priority of this branch of the group were waiting in long lines, for it is only in the fourth port every fourth scheduling - as when the process is included. 此时,最高优先权业务的总容积无论如何(明确地)保持在为此准备的束B的总容量(端口1+端口2+3/4端口3+1/4端口4)。 In this case, the total volume of business of the highest priority in any case (clearly) in a total capacity of preparing for this bundle B (port 1 + port 2 + 3/4-port 3 + 1/4-port 4).

当然,此时也能补偿端口带宽的不对称。 Of course, this time also can compensate asymmetric port bandwidth. 如果例如第四个端口P与其它三个端口相比只有半个带宽,但还要得到分配总业务的十二分之一,则为它要考虑到每个第2次的相应等候者长队。 Twelfth fourth port P if, for example, compared with only half the bandwidth of the other three ports, but also to get the total distribution business, which was to take into account each of the corresponding second person waiting in long lines .

要强调的是,对本发明通信网络的重要部分的说明,要理解为基本没有限制。 To emphasize that the communication network of the present invention will be described an important part of, to be understood as basically no restrictions. 对于有关专家尤为明显的是,将所应用的概念从功能上而不是从物理上去理解。 Especially for the experts is that the concept of the application from the function rather than from physical up understanding. 这样,将这些部分也能部分地或完整地在软件中和/或经过多个物理装置实现分配。 Thus, these parts can partially or completely in software and / or through a number of physical devices to achieve distribution.

Classifications
International ClassificationH04L12/54, H04L12/70, H04L12/841, H04L12/851, H04L12/801, H04L12/701, H04L12/803, H04L12/725
Cooperative ClassificationH04L2012/5632, H04L45/00, H04L12/5601, H04L47/2433, H04L47/10, H04L45/302, H04L2012/562, H04L47/125, H04L47/28, H04L47/17, H04L2012/5651
European ClassificationH04L47/24C1, H04L45/302, H04L47/10, H04L47/17, H04L47/12B, H04L45/00, H04L47/28, H04L12/56A
Legal Events
DateCodeEventDescription
12 Jan 2005C06Publication
9 Mar 2005C10Request of examination as to substance
7 May 2008ASSSuccession or assignment of patent right
Owner name: NOKIA SIEMENS COMMUNICATION CO., LTD.
Free format text: FORMER OWNER: SIEMENS AG
Effective date: 20080404
7 May 2008C41Transfer of the right of patent application or the patent right
11 Aug 2010C02Deemed withdrawal of patent application after publication (patent law 2001)