CN102420704B - Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) - Google Patents
Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) Download PDFInfo
- Publication number
- CN102420704B CN102420704B CN201110409971.8A CN201110409971A CN102420704B CN 102420704 B CN102420704 B CN 102420704B CN 201110409971 A CN201110409971 A CN 201110409971A CN 102420704 B CN102420704 B CN 102420704B
- Authority
- CN
- China
- Prior art keywords
- bandwidth
- lsp
- lsr
- algorithm
- path
- 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.)
- Expired - Fee Related
Links
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a traffic classes differentiated (TCD) failure recovery method based on a multi-protocol label switching technology (MPLS-TE). The method comprises an off-line processing stage and an on-line processing stage. In the off-line processing stage, by using a TCD algorithm, a label switched path (LSP) state information base is established and maintained in a network initialization stage, wherein the LSP state information base comprises an LSP identifier (ID), a main path, a backup path, a service flow attribute, a maintenance priority attribute and an establishment priority attribute and is used for subsequent flow switching and resource preemption. Two steps of flow switching and resource preemption are executed in the on-line processing stage. By adoption of the TCD algorithm, different failure recovery requirements of different service flows can be differentiated. The TCD algorithm has lower packet loss rate, shorter failure recovery time and higher resource utilization efficiency, and is superior to the conventional failure recovery algorithm.
Description
Technical field
The present invention relates to a kind of flow grade differentiated fault recovery method based on MPLS-TE, belong to communication technical field.
Background technology
Along with the surge of data business volume in network and customer volume, the degree of dependence of people to network increases day by day, and the integrity problem of network highlights day by day, and network survivability technology is arisen at the historic moment thus.The survivability of network mainly refers to that network is after node, link failure, when not needing human intervention, automatically can recover the ability of the business by fault effects.MPLS (Multi-Protocol Label Switching) is as core network switching technology of future generation, pass through label swapping algorithm, not only can provide and ensure and traffic engineering than the more effective QoS (Quality of Service) of traditional IP, and there is very strong network survivability sexuality.The protection algorism of MPLS is after fault occurs, and comparable IP layer provides the reaction time faster.MPLS protection can provide protects flow granularity very flexibly, and can select the type protecting flow.
One of important breakthrough of MPLS technology is exactly the support to QoS, namely possesses the reliably ability of efficient converting flow.Although current routing algorithm relative maturity, recovery time and performance after network failure still cannot be guaranteed.Therefore, although IP route provides heavy-route function, its recovery time longer (several seconds to a few minutes), for the business requiring high reliability, high real-time, such recovery time is obviously unacceptable.In addition, for the business that some needs QoS to ensure, not only need to re-establish new route after a failure and carry out delivery flow rate, and need the network bandwidth, network delay, the packet loss suitable with former transmission path.And lP route is when implementing fault recovery, be but difficult to ensure above requirement.Typical Network is as VOIP (Voice over IP), voice transfer, transmission of video, and for this kind of business, IP heavy-route usually seems unable to do what one wishes.IP heavy-route is when implementing to recover, and routing algorithm possibility convergence time is longer, will cause more data-bag lost, and IP heavy-route cannot guarantee the network bandwidth.Compare to traditional IP heavy-route, as the switching technology of backbone network, MPLS needs to respond faster network failure, can have the reaction time faster after network failure, more reliable bandwidth sum Delay Guarantee, and the reliability of network is improved with this.Any resource in network all likely breaks down, and wants the network providing a high reliability, and this network must can be expected and effectively be solved these possible faults.In actual applications, link failure, node failure or new used equipment replace etc. all can an of short duration interruption, and this interruption is certainly the smaller the better, and as the hot spot technology of next generation network, MPLS can recover fault rapidly within the time of a few tens of milliseconds level.Briefly, the crucial object of repairing fault in a MPLS network is exactly shorten time that network data transmission that fault causes interrupts as much as possible and flow parameter after making reparation still can meet network requirement as much as possible, if possible, the label switched path (may transmit data) set up unbrokenly should be able to carry out fault recovery.This just means that link and cross-coupled equipment should be unable to be subject to the impact of fault, and the data of losing are less.In order to achieve this end, there has been proposed some failure recovering algorithms based on MPLS, and effectively solving the some problem existed in network disaster recovery, but wherein also there are some areas for improvement.
(1) Makam algorithm
Makam algorithm is that a kind of typical global protecting switches recovery algorithms, as shown in Figure 1.When node 2 detects that fault occurs, it sends FIS (Fault Indication Signal) signal to PSL (Path Switch LSR), PSL performs stream and switches after receiving this signal, uses pre-configured backup LSP (Label Switched Path) (entrance-4-5-6-exports) to replace operating path.Makam algorithm fault recovery implementation algorithm is simple, and the recovery operation time is relatively little, sets up in advance because recover LSP.But when if the place of making a mistake distance Ingress node is far away, because FIS signal noticed by needs, the fault notification time in the restore cycle of this algorithm is relatively long, make existing many data stream before learning switching, thus these data flow transmitted will be abandoned by other LSR (Label SwitchRouter) and cause packet loss higher.
(2) Haskin algorithm
As shown in Figure 2.Haskin algorithm is the hybrid protection switching recovery algorithms that this locality adds the overall situation; its restoration path is made up of two parts: " reverse backup path " that on operating path, the upstream node of fault origination point and link are formed, and pre-configured another of PSL and operating path incoherent " top backup path ".Owing to not using special FIS signal to carry out fault notification, but by the packet that reverse backup path returns to notify that PSL carries out stream and switches, and Makam method comparison, Haskin algorithm decreases the fault notification time, and reduces packet loss.
Operating path is that entrance-1-2-3-exports, and when fault occurs, packet is delivered to PSL along reverse backup path 2-1 by LSR3, and stream is switched to top backup path entrance-4-5-6-and exports by PSL again.
(3) Hundessa algorithm
Lemma Hundessa and Jordi domingo Pascual utilizes buffer memory and tagged method to solve the problem of Haskin method packet backward, and the method is called as Hundessa Approach.This algorithm additionally adds that buffer memory buffer carrys out storage data bag at each LSR.
Just the data flow transmitted from upstream is entered in buffer memory buffer after each LSR on operating path receives reciprocal packet, then label tag is added at last packet transmitted toward downstream, at this label from when receiving in the other direction, the data flow of upstream is sent in buffer memory all always and store.Just the data flow handoff buffer memory is transmitted to the backup path toward entrance route direction after the tag returned in the other direction until receive.When representing after ingress router receives tag that the LSR on operating path back transmits data flow and being switched on backup path by ingress router, now ingress router just by the data flow handoff that transmits on operating path on backup path.The method all increases a buffer memory and the burden of LSR is strengthened on each LSR, and cost also increases, and when buffer memory occurs overflowing overflow, still can produce the situation that data-bag lost also has backward.
(4) Das algorithm
Das algorithm is the innovatory algorithm that Santos Kummer Das and P.Venkataram proposes for Makam Approach.Temporarily this algorithm sets up a backup path so that stream compression is dealt into egress router with the adjacent LSR that breaks down.As shown in Figure 3, when fault happens in operational, then adjacent with this fault downstream LSR transmits fault notification message to egress router, when egress router receives fault notification message FIS, just transmits fault notification message and heavy-route request to ingress router.
And before fault notification message does not arrive ingress router, the upstream LSR adjacent with this fault just calculates the interim backup path that arrives egress router immediately, and the data flow handoff transmitted before then ingress router not being received fault notification message FIS is on this interim backup path.After ingress router receives fault notification message FIS and heavy-route request just by data flow handoff on backup path, before fault notification message can not being arrived ingress router like this, the data flow that ingress router transmits continues to be sent to egress router by its interim backup path, to avoid the loss of packet.
(5) Dynamic algorithm
Dynamic algorithm adopts local heavy-route---and restoration path sets up reset mode as required.As shown in Figure 4, when link (2,3) breaks down, the neighboring upstream node LSR2 at fault place is responsible for the restoration path 2-5-6-egress LSR that on-demand computing one meets bandwidth demand, then utilizes signaling protocol to set up and recovers LSP, and performs stream switching.As shown in Figure 4, from resource utilization, the recovery LSP of this recovery algorithms optimizes, but due to needs on-demand computing restoration path after fault occurs, then also need to use signaling protocol to set up and recover LSP, therefore the recovery operation time of Dynamic algorithm is long, and when breaking down, packet loss is high.
There is following problem in existing method:
(1) Makam algorithm has mass data bag and is abandoned by LSR in the process oppositely transmitting FIS, causes packet loss higher.Haskin algorithm adds failure recovery time, thus causes larger time delay; Backup path resource cost is serious; When fault restoration switches back main path well, there is more serious backward situation.In Hundessa algorithm, LSR burden is heavy, and maintenance cost is high, may occur cache overflow.In Das algorithm, FIS transmission path path length Dynamic algorithm needs on-demand computing restoration path after fault occurs, and then also need to use signaling protocol to set up and recover LSP, therefore the recovery operation time is long, and when breaking down, packet loss is high.
(2) when fault occurs, existing Restoration Mechanism does not distinguish different ranks and qos requirement, just utilizes unified strategy to recover.Particularly have in limited time at Internet resources, these mechanism are difficult to the application ensured compared with high survivability requirement can obtain the fastest recovery.
Summary of the invention
In order in network recovery process, distinguish different ranks and qos requirement; Simultaneously when Internet resources are not very sufficient, ensure and can obtain the fastest recovery compared with the application of high survivability requirement.The present invention proposes a kind of fault recovery method (Traffic Classes Differentiatedfailure recovery algorithm, is designated as TCD) that can carry out Differentiated Services to flow grade based on MPLSTE.This algorithm is while the different survivability qos requirements considering different application; also existing several MPLS recovery algorithms is combined; being applicable to the traffic protection under consideration survivability QoS environment and recovery, is a kind of protection recovery algorithms distinguishing different survivability qos requirement.
The algorithm that this invention proposes combines the advantage of existing Makam algorithm, Haskin algorithm, introduce the bandwidth preemption mechanism in Differentiated Services simultaneously, link down and bandwidth will be minimized put in the first place and pay the utmost attention to, the problem that after solving the fault recovery of failure recovering algorithm in the past, high-grade Business Stream bandwidth service declines, fully ensure that the data flow with high priority can take the lead in recovering, ensured that the data flow be preempted can maintain its basic QoS demand to greatest extent simultaneously.The different faults that this algorithm can distinguish different business stream recovers demand, shows lower packet loss, less failure recovery time and has again the higher level of resources utilization simultaneously, being better than traditional failure recovering algorithm.
1. propose a kind of failure recovering algorithm distinguishing different survivability qos requirement newly.
2. new algorithm establishes LSP slip condition database at each LSR, database contain LSPID corresponding to every bar LSP,
Main path, backup path, Business Stream attribute, seizing attribute, be preempted the information such as attribute, for judging flow attribution in time when fault occurs, thus making corresponding judgement according to flow attribution.
3. database is set up by network flood when netinit or when topology changes, and in whole mpls domain, each LSR maintains a identical LSP slip condition database, and upgrades in time along with the requirement of Virtual network operator.
4., in order to alleviate the burden of the nearest upstream LSR near fault point, the flow switch of business higher for requirement of real-time is placed on entrance LSR place by new algorithm.
5. the switching of the flow to the higher business of reliability requirement is placed on and carries out on the LSR of fault point.
6. all very high Business Stream is required for real time and reliability, new algorithm takes the method that secondary switches, flow is first switched on reverse path after fault occurs, in time arriving border LSR, second time is switched on backup path again, so both ensure that the loss reducing packet to greatest extent, the flow delay after simultaneously also making secondary switch is enough little.
7. the new preemption algorithm of new algorithm application, according to the attribute setting up priority and maintenance priority of this LSP in LSP slip condition database, compare with the association attributes of the LSP be preempted, seize according to the principle of the lower Business Stream of the Business Stream preempting priority that priority attribute is higher, thus ensure that the qos requirement of the Business Stream seized and be preempted simultaneously.
Accompanying drawing explanation
Fig. 1 is Makam algorithm schematic diagram in prior art;
Fig. 2 is Haskin algorithm schematic diagram in prior art;
Fig. 3 is Das algorithm schematic diagram in prior art;
Fig. 4 is Dynamic algorithm schematic diagram in prior art;
Fig. 5 is TCD algorithm main flow chart of the present invention;
Fig. 6 is flow switch flow chart of the present invention;
Fig. 7 is preemption algorithm flow chart of the present invention;
Fig. 8 is KL2 network topology
Fig. 9 is the fault interrupting time
Figure 10 is lost packets quantity
Figure 11 is out-of-order packets quantity
Figure 12 is Packet transfer delay after fault recovery
Figure 13 is Makam algorithm occupied bandwidth situation
Figure 14 is Haskin algorithm occupied bandwidth situation
Figure 15 is TCD algorithm occupied bandwidth situation.
Embodiment
Below in conjunction with specific embodiment, the present invention is described in detail.
Failure recovering algorithm (the Traffic Classes Differentiated failurerecovery algorithm of flow grade Differentiated Services, TCD) basic thought the stream of different stage is given differentiation treat, the corresponding a kind of flow switch algorithm of stream of each rank, consider that after switching, high level Business Stream may reduce service bandwidth in addition, low-level Business Stream may completely tie up by high-grade business, TCD algorithm introduces new preemption mechanism.TCD algorithm takes into full account bandwidth resources and network performance, will minimize link down and bandwidth waste and to put in the first place consideration, by taking the principle of approaching layer by layer to improve degree of restraint to bandwidth-hogging, is down to minimum by link down and bandwidth waste.
TCD algorithm comprises processed offline (also i.e. initialization) and two stages of online process, and program flow diagram as shown in Figure 5.
In the processed offline stage, TCD algorithm is set up in the netinit stage and is maintained a LSP state information storehouse, contain LSP ID, main path, backup path, Business Stream attribute in storehouse, keep priority attribute, set up the information such as priority attribute, seize for follow-up flow switch and resource.
Online process successively performs two steps: flow switch and resource are seized.
The work that the flow switch stage completes is: Business Stream is divided into different ranks, the application of delay sensitive class is provided to the transmission service of low time delay, low loss rate, network is with peak data rate (Peak Data Rate, PDR) transmission user load, to guarantee that it has minimum transmission delay, exceed the packet that service level agreement (Service Level Agreement, SLA) configures maximum transfer rate to be dropped, the Business Stream of this rank is defined as SC0 level.Higher for reliability requirement and that requirement of real-time is higher Business Stream is defined as SC1 level and SC2 level respectively, directly carries out third layer routing forwarding for the Business Stream of doing one's best.Flow switch flow chart as shown in Figure 6.
(1) as follows to the treatment step of SC0 level Business Stream:
A1: after fault being detected, at the LSR place nearest from fault point, first discharges label to the SC0 DBMS stream arrived.
A2: according to the main path attribute of LSP slip condition database, reverse path is put into packet header, directly carry out third layer forwarding; Upstream send FIS information simultaneously, inform that upstream entrance LSR somewhere node breaks down, need to carry out flow switch, because FIS is the notification message in Level 2 Forwarding, transmission rate is much larger than the forwarding rate of bag, and therefore FIS signal can arrive entrance LSR place in advance.
A3: entrance LSR after receiving FIS signal, stops the transmission of MPLS layer immediately, performs following three tasks simultaneously:
4) by the data of follow-up arrival stored in buffer memory.
5) bag sent out for the last time is marked r.
6) according to the information of backup path in LSP slip condition database, send label request to downstream, after egress LSR returns label mapping message, wait for the arrival of the IP bag of passback.
A4: after the IP bag arrival of passback, according to tag mapping information, whether performed MPLS layer and forward, detect each bag is that last is labeled as the bag of r simultaneously, if not, then continue to forward; If so, A5 is performed.
A5: if the bag being labeled as r arrives, then search the bag being labeled as r+1 in the buffer, if can not find, then illustrate and abandoned, start to search the bag being labeled as r+2, by that analogy, if find correspondence markings to be the bag of r+n, performs A6.
A6: start by with the bag being labeled as r+n for the flow switch of beginning is on backup path, data flow enters the normal state forwarded.
(2) as follows to the treatment step of SC1 level Business Stream:
B1: whether fullly detect buffering area, if buffering area is full, it is described shared by other flows to be forwarded, then upstream LSR sends a notification message, and LSR place forwards in upstream; As buffering area less than, then forward at local LSR place; After detection, data enter buffering area.
B2: search its backup path in LSP slip condition database, merges the path between the fault point in main path to entrance LSR and backup path, as new backup path, sends label request to egress LSR.
B3: return label mapping message from egress LSR, now backup path is successfully established.
B4: by flow switch on backup path, data enter normal forwarding state.
(3) to the treatment step of SC2 level Business Stream
C1: after fault being detected, sends FIS message to entrance LSR.
C2: entrance LSR after receiving reverse FIS information, sends label request message to egress LSR.
C3: egress LSR returns label mapping message, now backup path is successfully established.
C4: by flow switch on backup path, data enter normal forwarding state.
The work completed in resource preemption phase is: by node LSR on enhancing link to the resources control ability of LSP, LSR is made to possess the ability of management LSP occupied bandwidth resource, LSR selects the LSP that a part of priority is lower, reduces their transmission rate to adapt to the bandwidth demand of new LSP.After initiating terminal LSR selects to reduce speed, algorithm will upgrade each LSR bandwidth on LSP, when there is available bandwidth in network, then recover the bandwidth of LSP.
[definition 1] set L be the set of the low all LSP of maintenance priority ratio p (new LSP sets up priority), and the set to be selected of the LSP that namely may be preempted, wherein comprises n element, be respectively LSPl
1, LSPl
2..., LSPl
i... LSPl
n.
L ', for reducing the set of the LSP of bandwidth, is respectively LSPl '
1, LSPl '
2..., LSPl '
i... LSPl '
n.
Reducible ratio is Δ %.
[definition 2] B is the bandwidth reserved set of the LSP that can be preempted, and wherein comprises n element, is respectively b (l
1), b (l
2) ... b (l
i) ... b (l
n), referred to as b (1), b (2) ... b (i) ... b (n), wherein b (i) represents LSP l
ithe bandwidth reserved of ∈ L.B ' is also bandwidth reserved set, wherein comprises k element, be respectively b ' (l '
1), b ' (l '
2) ... b ' (l '
i) ... b ' (l '
k), referred to as b ' (1), b ' (2) ... b ' (i) ... b ' (k), wherein b ' (i) represents LSP l '
ithe bandwidth reserved of ∈ L '.
[definition 3] r is the actual bandwidth needing to seize, r=b-A
bwl (), wherein b is the demand bandwidth of new LSP, A
bwl () is the available bandwidth on link.
[definition 4] P is priority set, wherein comprises n element, is respectively p (l
1), p (l
2) ... p (l
i) ... p (l
n), referred to as p (1), p (2) ..., p (i) ..., p (n), wherein p (i) represents LSP l
ithe priority of ∈ L.
[definition 5] X allly reduces bandwidth summation, that is,
X＝∑b′(l′
i)×Δ％ (1)
Algorithm takes into full account bandwidth resources and network performance, will minimize link down and bandwidth waste and to put in the first place consideration, by taking the principle of approaching layer by layer to improve degree of restraint to bandwidth-hogging, link down and bandwidth waste is dropped to minimum level.As shown in Figure 7, concrete steps are as follows for preemption algorithm flow process:
D1: LSP to be successively decreased arrangement by bandwidth according to the value of bandwidth reserved b (i) of LSP, if bandwidth is identical, arrangement of according to priority successively decreasing; If arrangement postorder is classified as b (i) (1≤i≤n).Make k=1, m=1.
D2: if need the actual bandwidth r seized to be not less than the bandwidth reserved of kth bar link, i.e. r >=b (k), then seize b (k), upgrades r=r-b (k), k=k+1, performs D2; Otherwise, perform D3.
D3: if r < b (n), performs [step4]; Otherwise, perform D7.
D4: if need the actual bandwidth r seized to be less than all reducible bandwidth synthesis X, i.e. r < X, perform [step5].Otherwise, perform D6.
D5: if the actual bandwidth r that needs are seized is greater than the reducible bandwidth on m article of link, i.e. r > b ' (m) × Δ %, then bandwidth-hogging b ' (m) × Δ %, make r=r-b ' (m) × Δ %, m=m+1, upgrade L ', perform [step5]; Otherwise, make b ' (m)=b ' (m)-r, more new data, seize end.
D6: seize the LSP minimum with b (n) bandwidth equal priority, more new data, seizes end.
D7: establish bandwidth to be just less than the LSP of r (as there being the LSP of multiple same band simultaneously, select the LSP that priority is minimum) be b (j), if after seizing the bandwidth of jth bar link, actual bandwidth-hogging r can be less than all reducible bandwidth summation X, i.e. r-b (j) < X, then seize b (j), make r=r-b (j), perform D5; Otherwise bandwidth-hogging, just greater than the LSP of r, more new data, seizes end.
Emulation experiment
KL2 network is the basic module in catenet, is also the basic network unit of research traffic engineering.The present invention uses this topology to carry out emulation experiment, and divide into groups after carrying out fault interrupting time, lost packets number, out-of-order packets number, fault restoration to Makam algorithm, Haskin algorithm and new algorithm respectively time delay and average occupied bandwidth carry out comparative study.
As shown in Figure 8, this network is by 15 nodes, and 28 bidirectional weighting links form for KL2 network topology.Wherein, the bandwidth capacity on thick limit is the link of 48Mbps, 2-5,2-3,3-6,3-7,6-11,6-10,10-11,7-10,14-15 namely in figure, and the link of to be bandwidth capacity be on thin limit 12Mbps, corresponding 5 sources, destination node are to being (S
i, D
i) i.e. (1,13), (5,9), (4,2), (15,5), (13,2).
By programming automatic generation 200 stream, the settling time of its stream and bandwidth by seed stochastic generation, and flow bandwidth Stochastic choice in 64Kbps, 100Kbps, 128Kbps and 200Kbps, and each packet due in obeys Poisson distribution.Emulation experiment carries out 20 times altogether, and each random seed is not identical, adds up after emulation to stream bandwidth ratio all types of on link.
Experiment service request all produces at random between all ingress-egress nodes, the Business Stream ratio of SC0 level, SC1 level, SC2 level respectively accounts for 1/3rd, has carried out four groups of experiments herein: first group of experiment compares three kinds of algorithms on interrupt service time for the difference of the Business Stream of three kinds of different brackets in KL2 network topology; Second group of experiment compares three kinds of algorithms in lost packets quantitatively for the difference of the Business Stream of three kinds of different brackets; 3rd group of experiment compares three kinds of algorithms in out-of-order packets quantitatively for the difference of the Business Stream of three kinds of different brackets; 4th group of experiment is for comparing the difference for the Business Stream of three kinds of different brackets on the bandwidth occupancy of three kinds of algorithms after fault recovery.
Analysis of simulation result
The checking of 1 fault interrupting time
Fig. 9 reflects the fault interrupting time of three kinds of algorithms.In Makam algorithm, three kinds of Business Streams are all very short for recovery time, because send FIS notification message to entrance LSR after finding fault, flow switches rapidly after receiving message, and its cost has mass data packet loss in the process sending FIS.In Haskin algorithm, because many one section of reverse transfer paths, therefore label request time can be longer, therefore its break period is also longer.In new algorithm, because distinguish the data flow treating different stage, the Business Stream of SC0 level abandons label after a failure, the request of new label is all on the backup path built up in advance with mapping, therefore its path is shorter for settling time, therefore break period is shorter, but still switch fast less than the notification message fast-forwarding based on transport layer of Makam algorithm, flow.
For SC1 level because have employed the algorithm of similar Haskin, therefore recovery time is longer, adopts the event of the method for similar Makam shorter for recovery time, and remain basically stable the break period of Makam method to SC2 level.
The checking of 2 lost packets and out-of-order packets quantity
Figure 10 and Figure 11 reflects lost packets and the out-of-order packets quantity of three kinds of algorithms.Can see, in Makam algorithm, because it ignores the situation of the grouping that period arrives when notification message, cause packet loss phenomenon in the process of whole label request comparatively serious, well switch back afterwards in the process of main path in fault restoration, because backup path all pre-sets, the time delay situation of main path and backup path is substantially similar, therefore, during switching after fault restoration, data flow on main path backup path almost arrives simultaneously, and thus out-of-order packets is less.
In Haskin algorithm, because directly switching from the nearest LSR place of fault, the loss of therefore wrapping is minimum, but because many than main road path length of backup path, therefore after fault restoration, when backup path switches back main path, has more packet loss.In TCD algorithm, because SC0 level Business Stream upstream sends FIS failure indication information after interrupting, meanwhile, grouping does not abandon by fault point the most nearby LSR, but directly oppositely do third layer forwarding, when it arrives entrance LSR, just in time catch up with the label mapping message receiving egress LSR, start to be MPLS to forward, therefore abandon grouping seldom.Similar with Makam algorithm of the situation of out-of-order packets, also makes out-of-order packets little.SC1 level Business Stream is because when switching, adopt the method for similar Haskin, data directly forward, and therefore packet loss is less, and because of main path backup path delay inequality after fault restoration, much thus out-of-order packets is more again.SC2 level Business Stream adopts the method for similar Makam when switching, lost packets is more but out-of-order packets is all less.
To divide into groups after 3 fault restoration the checking of time delay
Figure 12 reflects the Packet transfer delay after fault is restored, and Makam algorithm is the backup path set in advance because of what enable after fault recovery, and therefore time delay can be guaranteed, therefore time delay is shorter.Haskin algorithm is because it is that the reverse path having walked twice adds backup path in actual moving process, and therefore its time delay is longer.In TCD algorithm, for SC0 level Business Stream, transmit along backup path after Business Stream switches, therefore time delay is shorter, for SC1 level Business Stream, because it is based on Haskin pattern, traffic flow is through the reverse path of twice and backup path, and therefore its time delay is longer.For SC2 level Business Stream, directly along backup path transmission after flow switch, therefore its time delay is also less.The checking that 4 tiered band widths take
Figure 13, Figure 14 and Figure 15 are the bandwidth occupancy situation of Makam algorithm, Haskin algorithm and new algorithm respectively:
As can be seen from Figure 13 and in Figure 14, after breaking down at the 6.0th second, Makam algorithm and Haskin algorithm are through recovery after a while, Business Stream is because of the data flow share of bandwidth needed and on backup path, and bandwidth slightly declines, and finally all substantially maintains in this segment of 7000Kbps, but for high-grade Business Stream, only be left about 7000Kbps after recovering, therefore Business Stream bandwidth can not be guaranteed, and service quality declines.
Figure 15 illustrates SC0 level, SC1 level, the occupied bandwidth situation of SC2 level Business Stream under TCD algorithm.Though can see that the Business Stream of each rank slightly declines, but can report its basic traffic requirement of card, three kinds of Business Streams can continue the smooth transport maintaining its Business Stream after recovery, and flow bandwidth declines little, and the service quality of flow is guaranteed.Wherein SC1 level is because adopt the restoration methods of similar Haskin, its backup path in the recovery process of fault is many a reverse path, therefore its flow interrupt time is slightly long.
Should be understood that, for those of ordinary skills, can be improved according to the above description or convert, and all these improve and convert the protection range that all should belong to claims of the present invention.
Claims (1)
1. based on the flow grade differentiated fault recovery method of MPLS-TE (a Multiple Protocol Label Switching, negotiate marking commutation-traffic engineering), it is characterized in that, comprise processed offline and two stages of online process; In the processed offline stage, TCD (Traffic ClassesDifferentiated failure recovery algorithm, the failure recovering algorithm of flow grade Differentiated Services) algorithm is in the foundation of netinit stage and maintain a LSP (Label Switching Path, label switched path) state information storehouse, contain LSP ID, main path, backup path, Business Stream attribute in storehouse, keep priority attribute, set up priority attribute, seize for follow-up flow switch and resource; Online process successively performs two steps: flow switch and resource are seized; Flow switch is used for: Business Stream is divided into different ranks, the application of delay sensitive class is provided to the transmission service of low time delay, low loss rate, network is with the load of peak data rate transmission user, to guarantee that it has minimum transmission delay, the packet exceeding service level agreement configuration maximum transfer rate is dropped, the Business Stream of this rank is defined as SC0 level, higher for reliability requirement and that requirement of real-time is higher Business Stream is defined as SC1 level and SC2 level respectively, directly carries out third layer routing forwarding for the Business Stream of doing one's best; Resource is robbed and is occupied: by strengthening node LSR (Label Swithcing Router on link, LSR) resources control ability to LSP, LSR is made to possess the ability of management LSP occupied bandwidth resource, LSR selects the LSP that a part of priority is lower, reduces their transmission rate to adapt to the bandwidth demand of new LSP; After initiating terminal LSR selects to reduce speed, algorithm will upgrade each LSR bandwidth on LSP, when there is available bandwidth in network, then recover the bandwidth of LSP; The treatment step of described SC0 level Business Stream is as follows:
A1: after fault being detected, at the LSR place nearest from fault point, first discharges label to the SC0 DBMS stream arrived;
A2; According to the main path attribute of LSP slip condition database, reverse path is put into packet header, directly carry out third layer forwarding; Upstream send FIS information simultaneously, inform that upstream entrance LSR somewhere node breaks down, need to carry out flow switch, because FIS is the notification message in Level 2 Forwarding, transmission rate is much larger than the forwarding rate of bag, and therefore FIS signal can arrive entrance LSR place in advance;
A3: entrance LSR after receiving FIS signal, stops the transmission of MPLS layer immediately, performs following three tasks simultaneously:
1) by the data of follow-up arrival stored in buffer memory;
2) bag sent out for the last time is marked r;
3) according to the information of backup path in LSP slip condition database, send label request to downstream, after egress LSR returns label mapping message, wait for the arrival of the IP bag of passback;
A4: after the IP bag arrival of passback, according to tag mapping information, whether performed MPLS (MultipleProtocol Label Switching negotiate marking commutation) layer and forward, detect each bag is that last is labeled as the bag of r simultaneously, if not, then continue to forward; If so, A5 is performed;
A5: if the bag being labeled as r arrives, then search the bag being labeled as r+1 in the buffer, if can not find, then illustrate and abandoned, start to search the bag being labeled as r+2, by that analogy, if find correspondence markings to be the bag of r+n, performs A6;
A6: start by with the bag being labeled as r+n for the flow switch of beginning is on backup path, data flow enters the normal state forwarded;
The treatment step of described SC1 level Business Stream is as follows:
B1: whether fullly detect buffering area, if buffering area is full, it is described shared by other flows to be forwarded, then upstream LSR sends a notification message, and LSR place forwards in upstream; As buffering area less than, then forward at local LSR place; After detection, data enter buffering area;
B2: search its backup path in LSP slip condition database, merges the path between the fault point in main path to entrance LSR and backup path, as new backup path, sends label request to egress LSR;
B3: return label mapping message from egress LSR, now backup path is successfully established;
B4: by flow switch on backup path, data enter normal forwarding state;
The treatment step of described SC2 level Business Stream is as follows:
C1: after fault being detected, sends FIS message to entrance LSR;
C2: entrance LSR after receiving reverse FIS information, sends label request message to egress LSR;
C3: egress LSR returns label mapping message, now backup path is successfully established;
C4: by flow switch on backup path, data enter normal forwarding state;
The concrete steps that described resource is seized are as follows:
D1: LSP to be successively decreased arrangement by bandwidth according to the value of bandwidth reserved b (i) of LSP, if bandwidth is identical, arrangement of according to priority successively decreasing; If arrangement postorder is classified as b (i) (1≤i≤n); Make k=1, m=1;
D2: if need the actual bandwidth r seized to be not less than the bandwidth reserved of kth bar link, i.e. r >=b (k), then seize b (k), upgrades r=r-b (k), k=k+1, performs D2; Otherwise, perform D3;
D3; If r < b (n), perform D4; Otherwise, perform D7;
D4: if need the actual bandwidth r seized to be less than all reducible bandwidth synthesis X, i.e. r < X, perform D5;
Otherwise, perform D6;
D5: if the actual bandwidth r that needs are seized is greater than the reducible bandwidth on m article of link, i.e. r > b ' (m) × Δ %, then bandwidth-hogging b ' (m) × Δ %, make r=r-b ' (m) × Δ %, m=m+1, upgrade L ', perform D5; Otherwise, make b ' (m)=b ' (m)-r, more new data, seize end;
D6: seize the LSP minimum with b (n) bandwidth equal priority, more new data, seizes end;
D7: set bandwidth to be just less than the LSP of r as b (j), as there being the LSP of multiple same band simultaneously, select the LSP that priority is minimum, if after seizing the bandwidth of jth bar link, actual bandwidth-hogging r can be less than all reducible bandwidth summation X, and namely r-b (j) < X, then seize b (j), make r=r-b (j), perform D5; Otherwise bandwidth-hogging, just greater than the LSP of r, more new data, seizes end.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110409971.8A CN102420704B (en) | 2011-12-12 | 2011-12-12 | Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110409971.8A CN102420704B (en) | 2011-12-12 | 2011-12-12 | Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102420704A CN102420704A (en) | 2012-04-18 |
| CN102420704B true CN102420704B (en) | 2015-07-01 |
Family
ID=45944952
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110409971.8A Expired - Fee Related CN102420704B (en) | 2011-12-12 | 2011-12-12 | Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102420704B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103281252B (en) * | 2013-05-14 | 2017-04-26 | 华为技术有限公司 | Message flow control method and device based on multi-path transmission |
| CN103647711B (en) * | 2013-12-20 | 2016-08-17 | 大连大学 | A kind of satellite network heavy route method based on priority mechanism |
| CN106464835B (en) * | 2015-04-30 | 2019-08-13 | 华为技术有限公司 | Distribution method, the device and system of upstream bandwidth |
| CN107547388B (en) * | 2017-08-24 | 2021-03-09 | 新华三技术有限公司 | Message sending method and device |
| CN108933742A (en) * | 2018-06-21 | 2018-12-04 | 重庆邮电大学 | A kind of business stream scheduling method of SD-DCN link failure |
| CN111045845B (en) * | 2019-11-29 | 2021-09-17 | 苏州浪潮智能科技有限公司 | Data returning method, device, equipment and computer readable storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101013995A (en) * | 2007-02-12 | 2007-08-08 | 华为技术有限公司 | Network node equipment, network system and tunnel establishment method |
| CN101136844A (en) * | 2006-08-29 | 2008-03-05 | 华为技术有限公司 | Method and system for implementing differential service flux engineering of multi-protocol label switching network |
| WO2009056053A1 (en) * | 2007-10-24 | 2009-05-07 | Huawei Technologies Co, .Ltd. | A method, device and system of multi-protocol label exchange traffic engineering flow capacity switch |
-
2011
- 2011-12-12 CN CN201110409971.8A patent/CN102420704B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101136844A (en) * | 2006-08-29 | 2008-03-05 | 华为技术有限公司 | Method and system for implementing differential service flux engineering of multi-protocol label switching network |
| CN101013995A (en) * | 2007-02-12 | 2007-08-08 | 华为技术有限公司 | Network node equipment, network system and tunnel establishment method |
| WO2009056053A1 (en) * | 2007-10-24 | 2009-05-07 | Huawei Technologies Co, .Ltd. | A method, device and system of multi-protocol label exchange traffic engineering flow capacity switch |
Non-Patent Citations (1)
| Title |
|---|
| 基于MPLS TE故障恢复的研究;王惠来;《医学教育探索》;20080520(第05期);535-536 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102420704A (en) | 2012-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102420704B (en) | Traffic classes differentiated (TCD) failure recovery method based on multi-protocol label switching technology (MPLS-TE) | |
| CN104954256B (en) | Devices, systems, and methods for reconfiguring point-to-multipoint label switch paths | |
| EP1433287B1 (en) | Protection switching in a communications network employing label switching | |
| CN100558111C (en) | Metro Ethernet provides reliability processing method and the system under the multi-service networking | |
| CN101030939B (en) | Method and apparatus for establishing and restoring label exchange path in automatic-exchange optical network | |
| CN102638388B (en) | Flow label negotiating method, relevant device and system | |
| CN101150587B (en) | A method, device and system for traffic switching of multi-protocol label switching traffic engineering | |
| CN102907051B (en) | Label switching path establishing method, data forwarding method and equipment | |
| EP1843544B1 (en) | A data transmission method and system of label switching network | |
| EP3654664B1 (en) | Method for interconnected ring protection, computer-readable storage medium and computer program product | |
| CN102325037B (en) | A kind of changing method of pseudo-line dual homed network, system and dual-homed Provider Equipment | |
| US8988982B2 (en) | Method and apparatus for protection switching in point-to-multipoint network | |
| CN101159690B (en) | Multi-protocol label switch forwarding method, device and label switching path management module | |
| CN101710877A (en) | Method, device and system for processing traffic flow based on pseudowire | |
| US20080068988A1 (en) | Packet communication method and packet communication device | |
| Hariyawan | Comparison analysis of recovery mechanism at MPLS network | |
| CN100372305C (en) | Service protective method | |
| WO2012097595A1 (en) | Method and system for implementing shared-mesh protection | |
| CN102468973A (en) | Alarm control method and device for aggregation link in packet switching network | |
| CN101902396A (en) | Method and system for protecting tunnel in multi-protocol label switching traffic engineering | |
| CN103081406B (en) | Method and apparatus for restoring a connection through a provider network upon request | |
| KR100353850B1 (en) | Path protection method to support QoS of traffic flows between two routers in MPLS networks | |
| CN113037622B (en) | System and method for preventing BFD from vibrating | |
| Lin et al. | Redirection based recovery for MPLS network systems | |
| KR20020053393A (en) | An Efficient Restoration Mechanism Using Bandwidth Sharing Method In MPLS |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150701 Termination date: 20151212 |
|
| EXPY | Termination of patent right or utility model |