CN1601995A - Exchanging scheduling method of multiple packing collection - Google Patents

Abstract

Using period processing of BHP instead of traditional sequential processing of BHP, the invention provides conditions for BHP queuing, and paroxysmal data in high priority to preempt paroxysmal data in low priority. Different QoS are provided for different types of paroxysmal data through flexible queuing strategy. Within a dispatching cycle, multiple BHP is processed through optimized dispatch algorithm in multiple constraints and multiple targets. BHP dispatch in optimal mode raises utilization efficiency of exchange resources so as to reach purpose of raising switching performance.

Description

The exchange dispatching method is converged in a kind of many control groupings

Technical field

The invention belongs to the exchange control technology field in the communication system, it is particularly related to the exchange control technology of light burst switching system.

Background technology

Communication system generally partly is made of access, transmission and exchange etc., and wherein exchange is main realizes the connection between different user in the communication network, as forwarding professional in establishment of connection, the packet switching system in the circuit switching system etc.Along with the increasing of the increasing of the network user's increase, amount of information, network size etc., traditional technology based on the electricity exchange will reach very high switching performance, and as high-exchange-capacity, high port rate, low switching delay etc., the pressure that is faced is increasing.

The basic method that overcomes electricity exchange bottleneck is to adopt the packet-switch technology of full light, but also there is several difficult points in this technology at present, as full optical storage, light regularly/synchronous etc.For this reason, fully in conjunction with the advantage of also bringing into play existing light, power technology, making up the switching system that photoelectricity mixes, is the effective ways of the big capacity switching system of realization, as OBS (Optical Burst Switching, abbreviation OBS).

In the OBS network, carrying customer service bursty data (Burst) can be regarded as the overlength grouping that is made of mass data grouping, and the packets headers of this overlength grouping be exactly bursty data the control grouping (BurstHeader Packet, BHP).Different with legacy packets exchange is that BHP is separating on the physical channel with Burst: in the DWDM transmission system, can adopt one (or a plurality of) special wavelength to be used to transmit BHP as control channel, and other wavelength as data channel.Corresponding one by one between BHP and Burst, comprise bursty data among the BHP for information about, as shift time, burst length, data channel (wavelength) etc.Before the source end really sends user data, the shift time (offset time, promptly the source end sends the blanking time between BHP and the corresponding bursty data of transmission) between BHP and the bursty data is set, make BHP arrive the intermediate node of OBS prior to Burst.BHP is converted to the signal of telecommunication at intermediate node and handles, and comprises the determining of route, the reservation of resource and the configuration of switching matrix etc., guarantee that corresponding data channel has configured when bursty data arrives, thereby the realization data is in the transparent transmission in light territory.

From the function of finishing, the signaling in BHP and the legacy circuit-switched networks network is closely similar, and just in this sense, BHP is also referred to as signaling message.But different with traditional signaling message is, the signaling of OBS needn't be waited for the feedback acknowledgment of destination, and promptly the resource reservation of OBS is unidirectional.Also this just " unidirectional reservation " mechanism has reduced connection establishment delay, has improved channel utilization.

An OBS network mainly constitutes (seeing accompanying drawing 1) by fringe node, core node and DWDM link, and wherein fringe node is responsible for data are divided into groups to carry out buffer memory and encapsulation, is combined into bursty data, sends to it the most contiguous OBS core node then.Fringe node generates the BHP grouping of describing the bursty data characteristic during encapsulation, sends on specific control channel prior to bursty data.Core node is according to the BHP that receives on the control channel, can learn the time of advent, duration, the destination address control informations such as (and forwarding labels) of bursty data, and finish configuration to light path according to these information, the transparent channel that guarantees data (is seen document Y.Xiong, M Vandenhoute, and H.Cankaya.Control architecture in optical burst-switched WDM networks.IEEE Journal on Selected Areas in Communications, 18:1838-1851, October2000).

Though the OBS system can reach very high performance in theory, as big exchange capacity, low switching delay, high resource utilization etc., but under present optical tech level, the technical issues that need to address are also a lot, as the present invention at the exchange scheduling problem.

Exchange scheduling problem to be solved can simply be described as: under limited exchange resource, provide high as far as possible switching performance.At this, available exchange resource has: optical cross-connect matrix, optical delay line, wavelength shifter, a plurality of operation wavelengths of single port (optical fiber) etc.; The switching performance that needs to satisfy has: switching delay, exchange capacity, exchange Loss Rate etc.

Reach in the relevant performance evaluation in existing dispatching algorithm about the OBS core node, all adopted " service earlier first " (First Come First Service, FCFS) mechanism, be that we are called the method that order arrives scheduling: core node is whenever received a BHP who comes from upstream node, start dispatching algorithm at once, be the available exchange resource of Burst distribution of this BHP correspondence, and this BHP is forwarded to downstream node.

But the subject matter that this class exchange control (scheduling) scheme exists is, any one core node among the OBS is after handling a BHP, if the resource of this BHP reservation can be used, then this BHP is forwarded to next node at once, makes the exchange resource that it is preengage to change; BHP reservation requires and can't satisfy for the present node exchange resource, and this BHP is dropped.The problem that causes like this has:

1, the exchange priority of bursty data can't be abundant, guarantee flexibly (to see document M.Yoo and C.Qiao.Supporting multiple classes of services in IP over WDM networks.In proceeding ofGLOBECOM, volume lb, pages 1023-1027,1999. and document Mei Yang, S.Q.Zheng, and D.Verchere, " A QoS supporting scheduling algorithm for Optical Burst Switching DWDMNetworks; " Proc.of IEEE GLOBECOM 2001, Vol.1, pp.86-91,11/01.);

2, from the time angle, the service efficiency of exchange resource is low, there be a large amount of " fragment ", though promptly the part exchange resource is idle in certain time period, but these resources can't be utilized by the bursty data of follow-up arrival (to see document M.Iizuka, M.Sakuta, Yoshiyuki, " A SchedulingAlgorithm Minimizing Voids Generated by Arriving Bursts in Optical Burst Switched WDMNetwork; " Proceedings, IEEE Globecom 2002, November 2002.);

3, the BHP that can't handle in a scheduling process is dropped, make the exchange Loss Rate height of system (see document Jinhui Xu, C.Qiao, J.Li, and G.Xu. " Efficient Channel Scheduling Algorithmsin Optical Burst Switched Networks ", IEEE INFOCOM 2003,22nd Annual JointConference of the IEEE Computer and Communications Societies, San Francisco, March2003).

Summary of the invention

Defective at core exchange dispatching method in the existing OBS system, the purpose of this invention is to provide a kind of many control groupings and converge the exchange dispatching method, this method under identical exchange resource (as identical optical cross-connect matrix, optical delay line, wavelength shifter etc., and the annexation of these light path related devices), can improve the switching performance of OBS system, as reduce the exchange Loss Rate, improve the support of multipriority business etc.

It is that to establish traffic intensity be ρ that the exchange dispatching method is converged in a kind of many control groupings of the present invention, and the arrival frequency of BHP is λ, and core node does not have untreated BHP when initial, it is characterized in that adopting following step:

Step 1 is determined the length step of BHP binding time: arrival frequency and switching performance according to BHP require (mainly being bursty data Loss Rate and switching delay), determine the length of a BHP binding time, and this time span also is simultaneously one and exchanges dispatching cycle;

Step 2 starts time-out count device step: when first BHP (from the BHP that can't handle fringe node, core node or previous dispatching cycle) arrives current core node, start window time-out count device, and the time of advent of writing down each BHP of follow-up arrival;

Step 3 queue step: according to the information of carrying among the BHP, comprise service quality (Quality of Service, QoS), burst data length and shift time information, with all BHP queuings that arrive in the binding time; The method of described queuing can be to rank the time of advent according to bursty data, can be to rank according to the priority, also can be to rank according to burst data length, can also be that above several different methods is integrated;

Step 4 dequeue step: when window time-out count device value surpasses the binding time that sets, then from a plurality of waiting lists, take out several BHP, send into the exchange scheduler and dispatch processing according to justice service algorithm; Described fair service algorithm can be DRR (Dual Round-Robin), also can be WFQ (Weighted Fair Queue), can also be ERR (Elastic Round-Robin); Each number of taking out BHP can determine according to complexity, the requirement of performance raising degree and the disposal ability of processor that dispatching algorithm realizes;

Step 5 scheduling step: what the BHP of dequeue was entrained in the exchange scheduler extraction step 4 takies relevant information with exchange resource, comprise bursty data time of advent, burst data length, outbound port and wavelength, search for current switching node table of available resources (the idle information that mainly comprises each outbound port wavelength, the available information of wavelength shifter, optical cross-connect matrix and fiber delay line configuration status), adopt the multiple constraint multi-objective optimization algorithm make BHP as much as possible can be in this dispatching cycle successful reserve resource, promptly make the optical channel of Burst as much as possible obtain the reservation configuration;

Step 6 is upgraded the table of available resources step: the table of available resources of the resource updates present node of having preengage with step 5;

Step 7 is transmitted the BHP step: the BHP of successful reserve resource is forwarded to downstream node, and returns step 2 and wait for that new BHP arrives;

Step 8 is returned step: the BHP that will fail reserve resource and total stand-by period also do not surpass shift time returns the service waiting list, returns step 2; Otherwise abandon.

The flow chart of the inventive method as shown in Figure 2.The realization functional block diagram is seen accompanying drawing 3.

Contrast traditional scheme, the novelty of visible scheme that this paper sets forth is: the BHP binding time is set carries out cyclical process; Can utilize queuing flexibly and dequeue strategy to provide QoS to guarantee, and can reduce total bursty data Loss Rate for the bursty data of high priority; Adopt the further optimized dispatching of multiple constraint multiple-objection optimization dispatching algorithm, improve systematic function.Especially, step 3 and 4 is used for the support of realization system to many QoS bursty data, compares with the QoS mechanism in the existing OBS system, and this method improves a lot at aspects such as fairness, flexibilities.Traditional scheduling scheme is based on that BHP arrives in proper order, BHP of the each processing of scheduler, determine by BHP information whether its desired exchange resource is available: can preengage the configuration light path, then take exchange resource, transmit BHP, and when the Burst of its correspondence arrives the core switching node, allow this Burst pass through; Otherwise abandon this BHP, abandon the Burst grouping of its correspondence.

The essence that the exchange dispatching method is converged in many control groupings of the present invention is: the sequential processes of utilizing this innovation of cyclical process to BHP to break traditions to BHP, and seize the low priority bursty data for BHP queuing and high priority bursty data condition is provided; And utilization queuing flexibly provides different QoS with the dequeue strategy for inhomogeneous bursty data; Utilize multiple constraint multiple-objection optimization dispatching algorithm in a dispatching cycle, to handle a plurality of BHP simultaneously simultaneously, BHP is dispatched in the mode of more optimizing.Thereby improved the utilization ratio of exchange resource, reached the purpose that improves switching performance.

The characteristics that the exchange dispatching method is converged in many control groupings of the present invention are to utilize BHP in the OBS system to arrive the characteristics of core switching node prior to bursty data, not according to the information that BHP is entrained when carrying out each the scheduling, but utilize a plurality of BHP that arrived in certain time period, principle based on exchange resource optimization utilization, finish the processing of a plurality of BHP simultaneously, and then reach the purpose that improves switching performance at identical switching fabric (resource) down.

Relate to a large amount of light, power technology in the light burst switching system, though the technological means of electric treatment at present ripe, realize that cost is also lower, the performance of optical device is still not really desirable, cost is also very high, as the high-speed semiconductor optical switch, can coordinate laser etc.Therefore in the OBS system,, can under certain optical exchange structure, bring higher switching performance, have clear and definite engineering using value if adopt comparatively complicated electric treatment technology.

The present invention proposes a kind of many control groupings and converge the exchange dispatching method, is exactly the raising with the electric treatment complexity of limited extent, exchanges the raising of optical switching system overall performance for.Compare OBS core switching system and traditional electrical switching system, can find that the OBS system is subject to the performance of optical device aspect a lot, as can't the time slot operation, the cost prohibitive that adopts inner exchanging to quicken can only make and use up time-delay rather than light buffer memory truly etc.; But OBS also has its inherent advantages, and as the full optical transmission/exchange of teleservice, each port (or every optical fiber) has a plurality of wavelength availables, and control grouping arrives prior to packet and finishes the reservation of exchange resource etc.Scheduling scheme proposed by the invention has just made full use of these advantages of OBS, in the Computer Simulation that we have carried out, adopt many control groupings to converge the OBS system of exchange scheduling scheme, under the configuration of identical optical switching network, identical exchange load factor, the systems exchange Loss Rate has improved the switching performance of OBS system greatly than the system that adopts the sequential scheduling scheme order of magnitude that can descend.

Accompanying drawing and description of drawings

Fig. 1 is the formation schematic diagram of OBS network

Fig. 2 is the flow chart of the inventive method

Fig. 3 controls grouping more and converges exchange scheduling feature block diagram

A kind of basic core node OBS optical exchange structure of Fig. 4

Fig. 5 employing exchanges the schematic diagram of scheduling scheme in proper order

Fig. 6 adopts many control groupings to converge the schematic diagram of exchange scheduling scheme

The emulation contrast of Fig. 7 high priority exchange Loss Rate

Embodiment

Exchange scheduling scheme proposed by the invention can be used for different optical exchange structures.Below provide one embodiment of the present of invention with optical exchange structure shown in Figure 4.

As shown in Figure 4, the structure of OBS core node mainly comprises optical cross-connect matrix, wavelength shifter and optical delay line.In this system configuration, 4 input optical fibres are arranged, every input optical fibre has 9 wavelength, is used to bursty data and BHP channel of 8 tunnel business of carrying.Same have 4 output optical fibres, so system supports 4 groups of exchanges of totally 32 road bursty datas.Every road has only provided 8 data wavelength among the figure, and they are connected to 8 wavelength shifters by channel-splitting filter when arriving optical exchange structure; The output of wavelength change device by 8 independently the controllable light delay line be connected to 32 * 32 public clog-free optical cross-connect matrixes; 4 groups (8 every group) output of matrix, output wavelength is not overlapping in the requirement group, is connected to an output optical fibre by wave multiplexer.

Under this optical exchange structure, suppose that current output optical fibre (port) has only a wavelength to use, so all bursty datas that mail to this optical fiber all must be by a port output of 32 * 32 optical cross-connect matrixes.When adopting this moment existing order to exchange scheduling scheme, be shown in Fig. 5, suppose that wherein the priority of Burst1 (corresponding to BHP1) is lower than Burst2 (corresponding to BHP2) owing to clashing the situation that causes the part burst group to transmit.As seen from Figure 5, BHP1 (supposing from input optical fibre 1) arrives earlier, has therefore preengage exchange resource (output port of switching matrix); When BHP2 (supposing from input optical fibre 2) arrives, search resource table do not have the free time (output port of switching matrix by Burst1 take, the input port optical delay line reaches maximum), so Burst2 will be dropped.This moment, not only the grouping of high priority (Burst2) did not obtain exchange, and had been dropped.

When adopting many control groupings of the present invention to converge the exchange dispatching method, the time queuing that BHP is arrived by its Burst is overlapping if Burst has earlier, then queuing according to priority again, the length queuing of pressing bursty data when priority is identical again.From Fig. 6, can find, because BHP1, BHP2....BHPn are handled simultaneously (in same scheduling window), the BHP1 that arrive first this moment does not preengage the switching matrix outbound port, but BHP2 has preengage this port, makes Burst2 at first obtain exchange at once; The bursty data packet reservation of BHP1 correspondence to be the optical delay line of its inbound port, and delayed Burst1 and Burst2 are eliminated in the competition of switching matrix outbound port, two Burst are exchanged smoothly, have improved switching performance.

Based on optical exchange structure shown in Figure 4, when adopting existing order exchange scheduling scheme and many control groupings of the present invention to converge the exchange scheduling scheme, the exchange Loss Rate simulation result of high-priority service is seen Fig. 7 (emulation tool is a Visual C/C++ development platform).When adopting existing sequential scheduling, if do not adopt other QoS mechanism, then all business all are to handle coequally in fact; After adopting the inventive method, the Loss Rate of high-priority service descends greatly, particularly when traffic carrying capacity is big.In addition, can find out also from Fig. 7 that along with the increase of binding time, Loss Rate also descends thereupon.

Significantly, the scheduling scheme that the present invention proposes is under the bigger situation of system load rate, and the comparable existing order exchange scheduling scheme of exchange Loss Rate descends nearly 10 times.This has fully shown good result of the present invention.

Claims (1)

1, the exchange dispatching method is converged in a kind of many control groupings, it is characterized in that adopting following step: step 1 is determined the length step of BHP binding time: arrival frequency and switching performance according to BHP require (mainly being bursty data Loss Rate and switching delay), determine the length of a BHP binding time, this time span also is simultaneously one and exchanges dispatching cycle;

Step 2 starts time-out count device step: when first BHP (from the BHP that can't handle fringe node, core node or previous dispatching cycle) arrives current core node, start window time-out count device, and the time of advent of writing down each BHP of follow-up arrival;

Step 3 queue step: according to the information of carrying among the BHP, comprise service quality (Quality of Service, QoS), burst data length and shift time information, with all BHP queuings that arrive in the binding time; The method of described queuing can be to rank the time of advent according to bursty data, can be to rank according to the priority, also can be to rank according to burst data length, can also be that above several different methods is integrated;

Step 4 dequeue step: when window time-out count device value surpasses the binding time that sets, then from a plurality of waiting lists, take out several BHP, send into the exchange scheduler and dispatch processing according to justice service algorithm; Described fair service algorithm can be DRR (Dual Round-Robin), also can be WFQ (WeightedFair Queue), can also be ERR (Elastic Round-Robin); Each number of taking out BHP can determine according to complexity, the requirement of performance raising degree and the disposal ability of processor that dispatching algorithm realizes; Step 5 scheduling step: what the BHP of dequeue was entrained in the exchange scheduler extraction step 4 takies relevant information with exchange resource, comprise bursty data time of advent, burst data length, outbound port and wavelength, search for current switching node table of available resources (the idle information that mainly comprises each outbound port wavelength, the available information of wavelength shifter, optical cross-connect matrix and fiber delay line configuration status), adopt the multiple constraint multi-objective optimization algorithm make BHP as much as possible can be in this dispatching cycle successful reserve resource, promptly make the optical channel of Burst as much as possible obtain the reservation configuration;

Step 6 is upgraded the table of available resources step: the table of available resources of the resource updates present node of having preengage with step 5;

Step 7 is transmitted the BHP step: the BHP of successful reserve resource is forwarded to downstream node, and returns

Step 2 waits for that new BHP arrives;

Step 8 is returned step: the BHP that will fail reserve resource and total stand-by period also do not surpass shift time returns the service waiting list, returns step 2; Otherwise abandon.

Images