CN101370314A - Resource reservation method based on intra-domain network state perception in optical burst exchange - Google Patents

Abstract

A resource reservation method based on intra-domain network status sensing comprises collision detection step and dynamic QoS support step. The collision detection step of intra-domain network status sensing comprises: sending burst control grouping to OBS network in broadcasting mode, all edge node of OBS domain using received BCP for maintaining a intra-domain network status model; Introducing a time slot for detecting latent collision, size of time slot equaling to distance between farthest a pair of nodes in optical signal transmission domain, that is, time needed by intra-domain network diameter. Edge node working step: controlling for creating BCP, sending BCP to OBS network, executing collision detection to received BCP, completing burst data transmission, creating corresponding statistical information. Working steps for setting core node: executing collision detection for received BCP, configuring optical switch OXC according to collision detection result.

Description

In the light burst-switched based on the method for obligating resource of intra-domain network state perception

Technical field

The invention belongs to the data switching networks field, specifically relate to the method for obligating resource in the light burst switching system.

Background technology

Usually adopt unidirectional resource reservation mode among the OBS (light burst-switched), BCP is the time slot light wavelength channel that pairing bursty data DB sets up unidirectional reservation by the burst control grouping, realize the dynamic assignment of bandwidth, the two not only separates in time, but also transmits in different light wavelength channel.The technology that the control of this separation has reduced OBS realizes threshold, makes OBS become a kind of light exchanged form of very attractive.

Therefore the resource reservation mode becomes a kind of very crucial technology of OBS network.The resource reservation mode affects directly that bursty data postpones end to end, burst control grouping request blocking rate and bursty data loss ratio etc.

Up to now, the reservation mode of OBS typically can be divided into one the step reservation mode (Tell And Go, TAG-OBS) and two the step reservation mode (Tell And Wait, TAW-OBS).The representative of TAG-OBS has JIT (Just-In-Time) ^[1]And JET (Just-Enough-Time) ^[2]Two kinds, in the TAG-OBS RSVP, owing to need not to wait for the affirmation message of purpose fringe node, therefore postpone lessly end to end, weak point is to detect potential conflict.Therefore when the resource at intercooler core node place is reserved in a plurality of burst control groupings concomitantly, will produce the burst conflict.And the TAW-OBS RSVP ^[3], since the fringe node in the OBS network and if only if when receiving the affirmation message that the purpose fringe node beams back, just bursty data is sent in the OBS network and goes.Therefore the long delay of bursty data experience, but can obtain lower burst loss ratio.

In the method for obligating resource to QoS (Quality-of-Service, service quality) support the JET-QoS scheme is arranged more typically ^[4]With segmentation QoS scheme ^[5]The JET-QoS scheme is given different offset times by the bursty data of giving the different flow type, the bursty data of high priority is given long offset time, the bursty data of low priority is given short offset time, the bursty data of high priority is left to resource in advance with regard to higher probability is arranged like this, the data that a weak point of the method is a high priority tend to experience long delay, and often the data of high priority are responsive to retardation ratio.In addition, this method is to guarantee the burst loss ratio that high-priority data is low with the bursty data of sacrificing low priority.And in segmentation QoS scheme, upper strata data on flows with different priorities, be assembled in the bursty data according to certain order (as order from high to low), when clashing, discard the low part of level earlier of afterbody, though this kind mode has avoided abandoning whole bursty data when the burst conflict takes place.But one of shortcoming is the assembling complexity that increases fringe node.

Summary of the invention

The objective of the invention is to overcome in the prior art, postpone, then have higher burst loss ratio if the reservation schemes of OBS is emphasized to reduce end to end.If emphasize lower burst loss ratio then sacrifice the defective that reduction postpones end to end, a kind of method for obligating resource based on the intra-domain network state perception is provided, this method can be avoided unnecessary bandwidth waste, reduces the burst loss ratio.Further aim of the present invention is, reduces the delay of high-priority data and the burst loss ratio of minimizing low priority bursty data.

The technical scheme that realizes the foregoing invention purpose is: based on the method for obligating resource of intra-domain network state perception, comprise the collision detection step and the dynamic QoS support step of intra-domain network state perception;

The collision detection step of described intra-domain network state perception comprises:

1) in the mode of broadcasting the burst control BCP that divides into groups is sent in the OBS network and goes, fringe nodes all in this OBS territory can divide into groups with received burst control, safeguard an intra-domain network state model (Intra domainNetwork State Model, INSM).Owing to exist the propagation delay of light, so the intra-domain network state model INSM that each fringe node is safeguarded is near synchronous.Fringe node has had INSM, not only can know the operating position of resource in the current network, and can know the role pressure level of other fringe node.

2) introduce a time slot (Tts represents) and be used for detecting potential conflict: the size of this time slot equals the distance between a pair of farthest node in light signal experience territory, i.e. needed time of network diameter in the territory.Its end-to-end delay is shown in following formula:

T _e＝T _o+T _d+T _b；T _o≥n×T _f+T _ts (1)

Here T _oRefer to the offset time between burst control grouping and the bursty data; T _dFor this bursty data experiences needed time of selected route, T _bBe the duration of transmitting, T _fBe meant and transmit needed time of BCP, T _TsIt is the conflict time slot; T _OXCBe the configuration needed time of optical switch; N is the number of intermediate node.Concrete operation principle is described as shown in Figure 1.

3) the fringe node job step is set: control generates burst control grouping BCP, sends burst control grouping BCP to OBS network, carries out collision detection to receiving BCP, finishes bursty data and sends, and generate corresponding statistical information.

4) the core node job step is set: carry out collision detection to receiving BCP, dispose optical switch OXC according to the collision detection result.

The described fringe node job step that is provided with further comprises:

(1) be set the fringe node work period, this cycle be a bursty data DB from finish be assembled into be sent to fully go the OBS network till;

(2) divide three phases with BCP (Burst Control Packet, the burst control grouping) period of fringe node:

I. after finishing burst assembling, network state model INSM in the inquiry field obtains in suitable route and the current OBS territory information such as idle wavelength collection.

Ii. according to resulting result of stage one, generate burst control grouping BCP.

Iii. send in this BCP to OBS network and go, simultaneously this BCP is saved to " BCP Buffer " lining.

(3) fringe node collision detection: constantly carry out collision detection to receiving BCP.

(4) the bursty data DB period: mainly finish bursty data and send, and generate corresponding statistical information, divide three phases:

I. be stored in when clock controller detects " BCP Buffer " when the pairing time slot of the BCP of lining finishes, when promptly offset time arrives,, the source node of this BCP itself will dispose optical switch OXC if being this fringe node, carry out the preparation of transmission;

Ii. lasting transmission bursty data enters in the OBS network and goes;

Iii. with the information that BCP comprised, as the bursty data size, filling rate, and real-time information (as postponing end to end) is sent to statistical module.

Described fringe node collision detection is made up of three phases:

I. fringe node periodically receives the BCP that sent from the OBS core network and the own BCP that is generated, and is stored in corresponding according to priority classification " BCP Buffer " lining.

Ii. after receiving a BCP, fringe node detect this BCP whether be kept at " BCP Buffer " BCP of lining conflicts mutually.If there is conflict, the BCP of fringe node low priority is set to disarmed state, really do not abandon this BCP, and if low priority BCP exist " " this BCP pairing reservation in the network resources model INSM in the territory also will be cancelled in the lining to BCP Buffer.

If iii. do not detect conflict, fringe node will remove to upgrade intra-domain network state model INSM with this BCP, so that fringe node can in time be understood current network state.

The described core node job step that is provided with further comprises:

(1) be set the core node work period, this cycle is for receiving a BCP, till the configuration of finishing the pairing bursty data DB of this BCP.

(2) (Virtual Configuration Table VCT), promptly when receiving a BCP, only registers, and does not carry out real distribution at core node applying virtual allocation list; When the pairing time slot of this BCP finishes, when just offset time arrives, just carry out optical switch OXC configuration.

Described core node disposes optical switch OXC according to received BCP and comprises two periods: collision detection period and configuration period.

Described core node collision detection step is: receive BCP and classification storage, detect conflict by collision detection algorithm again.Specifically comprise following three phases:

I. core node periodically receives the BCP that is sent from the OBS core network, and is stored in corresponding according to priority classification " BCP Buffer " lining.

Ii. when receiving a BCP, the start detection algorithm detect this BCP whether be stored in " BCP Buffer " other BCP of lining conflicts mutually, if detect conflict, it is invalid that the BCP of low priority is set to, if the BCP of this priority registers in virtual configuration Table V CT, also the registration entries at VCT to be removed.The VCT here comprises the form of information (as information such as bursty data time of advent, duration and routes) among the BCP for record.

If iii. do not detect conflict, core node is registered in virtual configuration Table V CT with the relevant information among the BCP.

Described core node configuration step is: be the bursty data DB configuration optical switch that is about to arrive, specifically be made up of three phases:

When i. in clock controller detects VCT, having bursty data to be about to arrive, configuration optical switch OXC;

Ii. the duration of optical switch OXC, pass through OXC fully until whole burst;

Iii. when finishing when once disposing, generate corresponding statistical information: as the throughput of node, the size of average bursty data etc.

Described dynamic QoS is supported algorithm steps, the goal of the invention that further realizes reducing the delay of high-priority data and reduce the burst loss ratio of low priority bursty data, and this algorithm comprises:

A) two parameters are set: the role pressure index (Task Pressure Index, TPI) and pass the time (ElapseTime, ET).

Role pressure index TPI represents the busy extent of a node; With the TPI initial setting is 5 grades, i.e. TPI _I, TPI ₂TPI ₅, the index number of TPI is big more, represents that this fringe node is busy, and vice versa; And another parameter is to pass time ET, represents the time that pairing BCP is experienced in the OBS network.In order to realize the service of differentiation, here be N different class with bursty data flow type definition, C by name respectively ₁, C ₂C _N, and according to descending, i.e. C ₁Represent the bursty data of limit priority, and C _NRepresent the bursty data of lowest priority, it is high more responsive more to the time to set the bursty data flow priority simultaneously.

B) collision detection: when two or more BCP reserved same outbound concomitantly, node had just detected conflict.

C) detect low priority BCP: after detecting conflict, by comparing the BCP that priority obtains one or more low priority, when priority between the afoul BCP equates, compare the role pressure index TPI between the BCP and pass time ET, therefore in this programme, nearly all BCP can tell priority according to top three parameters.

D) regeneration BCP: in case when obtaining low BCP, this node judges whether self is the source node of low priority BCP, if, whether the time of then further judging this BCP and being experienced surpasses TTL (Time To Live), if do not have, then with the BCP of low priority, heighten one-level, i.e. C _iBe adjusted into C _I-1(2＜i＜N).After adjusting priority, regenerate BCP, and be sent in the OBS network and go.

Basis of the present invention is, whether the abandoning of a data burst, and its tangible fringe node delivery time has just determined that to a great extent this is because the configuration on a large scale of the immature and wavelength shifter of the light caching technology in the OBS network.When fringe node is a bursty data selection route and distribution wavelength, do not penetrate partially or carry out wavelength Conversion if having, institute's assigned wavelength and route can keep consistency, so be sent in the burst conflict on the intercooler core node, can treat as and are the conflict between the fringe node; More in addition, though a lot of nodes of burst control grouping experience often only limit to local node when Resources allocation.Among the present invention, raising OBS network edge node divides into groups more effectively to generate burst control to the perception of network state, introduce a collision detection time slot simultaneously and detect potential conflict, use between the time of TAG-OBS and TAW-OBS to reach the burst loss ratio that approaches the TAW mode.By core node having been designed the virtual configuration table, when the burst conflict, do not have or less bandwidth waste.In addition, in order better to support QoS, designed a dynamic QoS support algorithm that bursty data priority is dynamically adjusted, to reduce the burst loss ratio of low priority bursty data.

The present invention uses two measures and handles conflict between the bursty data: the one, and network state model INSM avoids reserving the bandwidth resources that those are being used in the application domain; Another is exactly to introduce a time slot to detect potential conflict, makes the low BCP of level earlier avoid the BCP institute reserved resource of high priority, to reduce the burst loss ratio and better to support QoS.The introducing of these two measures makes that a lot of unnecessary bandwidth waste are avoided, and makes that the burst loss ratio is very low.

This programme is compared with the TAW scheme with TAG, and its advantage has:

Lower delay and low burst loss ratio: the end-to-end delay T of formula (1) and TAG-JET, TAW _e(T _e=T _o+ T _d+ T _b, T _o〉=n * T _s+ T _OXC, T wherein _sBe the needed time overhead of intermediate node handles BCP; T _e=T _o+ T _d+ T _b, T _o=n * T _s+ 2 * T _d) relatively, the end-to-end delay that can find this method is between TAG-JET and TAW, but owing to can avoid existing conflict and detect the conflict that is about to arrival, therefore the loss ratio of burst can be far below the TAG method.

Dynamic priority is adjusted: the priority of the pairing bursty data of all BCP is not unalterable, and when the resource that BCP applied for of low priority was seized or blocked to fall, the priority of this bursty data also can be raised.Therefore, this kind dynamic priority is adjusted mechanism, promptly consider the node pressure level, allow busy node obtain more chance and remove reserved resource, to accelerate the transmission speed of bursty data, also considered passing the time of BCP in addition, this just makes that experience obtains to be left in advance chance too than the BCP of multinode, has further improved the performance of network; And the priority of all bursty datas, except the bursty data of limit priority, the bursty data of other priority is dynamically heightened along with after getting clogged or seizing, and the BCP that so not only gives high priority has more that multimachine can go to be left in advance bandwidth resources; Take into account fairness simultaneously, guarantee that the BCP of low priority can reserve the corresponding bandwidth resource in a limited time.

Limited support QoS in entire domain: a lot of existing QoS schemes can only be accomplished the strict QoS of support of individual node, promptly all fringe nodes all are that information based on this locality generates corresponding BCP information, in case BCP is sent in the OBS network and goes, middle core node equally also is based on the processing that local information is reserved, and therefore supports unfriendly to QoS.And this method has good perception owing to have the intra-domain network state model to current network state, for bursty data generates corresponding BCP, be not based on this locality, but is being conceived to entire domain.Add the introducing time slot in addition, can detect potential conflict, the bandwidth waste phenomenon of having avoided the invalid transmission of bursty data of low priority to be caused.

No buffer memory and do not have the wavelength transducer: all are similar to the selection of collision detection, route and the decisions such as distribution of wavelength all are by carrying out based on the fringe node of intra-domain network state model INSM, and the intercooler core node only need carry out and transmit received BCP apace, and safeguards that virtual configuration Table V CT gets final product.Therefore even without buffer memory and wavelength shifter, also can make the bursty data loss ratio maintain reduced levels, this also must this method have more actual feasibility.

Description of drawings

Fig. 1 is an operation principle schematic diagram of the present invention

Fig. 2 is 1 fringe node work period an of embodiment of the invention flow chart

Fig. 3 is 1 core node work period an of embodiment of the invention flow chart

Fig. 4 is that the embodiment of the invention 1 dynamic QoS is supported algorithm flow chart

Fig. 5 is the embodiment of the invention 1 network topological diagram

Embodiment

Below in conjunction with specific embodiment, be described further.

Embodiment 1

As shown in Figure 1, described among Fig. 1 and how to have used INSM and time slot makes new bursty data avoid applying for using those resources of using, and avoided the resource that those will use.

Suppose at t1 first fringe node, 1 transmission constantly first burst control grouping BCP1, and second fringe node 2 constantly also sends second burst control grouping BCP2 at t2 and enters the OBS network, and the priority of establishing BCP1 is higher than BCP2, and at the concurrent reserved resource in first core node, 1 place.Obviously, two fringe nodes just avoid reserving the resource that those have taken according to the INSM that is safeguarded separately when the BCP that sends separately, do not recognize potential conflict.And when arriving t3 during the moment, second fringe node 2 is recognized potential conflict, and the pairing bursty data of BCP2 this moment does not send, and therefore second fringe node 2 can send once more to the time that re-routes or cushioned conflict of pairing BCP.And at t4 constantly, first fringe node 1 has also been known the generation of conflict, but because the priority of self BCP that sends is higher, therefore continues the offset time transmission bursty data that sets according to BCP1.And middle core node is owing to using identical detection strategy, therefore also near the synchronous potential conflict that detected.(Virtual Configuration Table VCT), when the pairing time slot of this BCP finishes, when just offset time arrives, just carries out optical switch OXC configuration because at intercooler core node applying virtual allocation list.So can not cause the bandwidth waste that there is no need.

As Fig. 5, the network topology of one 6 node, this topology is by three fringe nodes: first fringe node 1, second fringe node 2 and the 3rd fringe node 3 and three core nodes: (fine rule on the link is represented control channel for first core node 4, second core node 5 and the 3rd core node 6 and six links, thick line representative data channel) forms, the length of the digitized representation link on the link, the km of unit.Core node does not have wavelength transducer and fibre delay line FDL.Link is a two-way link, and every link comprises 1 control channel and 8 data channels, the mode of channel allocation pattern for hitting at random, and channel width is 10Gbit/s, light signal transfer rate in optical fiber is 200,000km/s.

Fringe node use to be obeyed ON/OFF model that Pareto distributes and is produced IP grouping (this model IP data service of approaching reality distributes), and the number of ON/OFF model is 200, and burst coefficient Hurst is 0.75.The length obeys index distribution of IP grouping, the average length value is 1,024 byte.The load that is applied in the network is calculated by following formula:

$\mathrm{Network}\_\mathrm{Load}=\mathrm{Actual\; Bandwidth}/\underset{1}{\overset{\mathrm{ET}}{\mathrm{\Σ}}}\underset{1}{\overset{\mathrm{LT}}{\mathrm{\Σ}}}(\mathrm{Bandwidth}\×\mathrm{WT})---\left(2\right)$

Actual Bandwidth representative is used for producing the bandwidth of IP packet traffic; ET is the total number of fringe node; LT is current fringe node general export number of links; Bandwidth is the bandwidth of current wavelength channel; WT is meant the data channel number that current outbound comprises.

Method of the present invention comprises the collision detection method and the dynamic QoS support method of intra-domain network state perception.

Collision detection method comprises the following steps:

1) in the mode of broadcasting the burst control BCP that divides into groups is sent in the OBS network and goes, fringe nodes all in this OBS territory can divide into groups with received burst control, safeguard an intra-domain network state model (Intra domainNetwork State Model, INSM).

2) introduce a time slot (T _TsExpression) be used to detect potential conflict, the size of this time slot equals the distance between a pair of farthest node, i.e. needed time of network diameter in the territory in the light signal experience territory.

3) fringe node and core node are set, make it to carry out work according to the cycle.

The fringe node one-period be meant a bursty data DB from finish be assembled into be sent to fully go the OBS network till; The core node one-period is meant receives a BCP, till the configuration of finishing the pairing bursty data DB of this BCP.

EdgeNode?Period＝T _BCP+T _D+T _DB，CoreNode?Period＝T _D+T _DB (3)

T _BCPDivide into groups the required time for generating burst control, its size is by bursty data built-up time decision (is 4 milliseconds in set time its value of thresholding packing algorithm); T _DWhether conflict the required time for detecting BCP, its big or small default value is 5 microseconds; T _DBFor sending the required time of bursty data, its size is 1 millisecond for default value.

Burst control grouping BCP transmits in the mode of broadcasting, and the number of times that BCP is transmitted is obtained by following formula:

$\mathrm{Times}=1+\underset{1}{\overset{\mathrm{CT}}{\mathrm{\Σ}}}\mathrm{Degree}\left(\mathrm{corenode}\right)---\left(4\right)$

Here " 1 " expression BCP is sent to the OBS network by its source fringe node; CT be the intercooler core node total number (core node CT is a node 4,5 in this example, 6; Totally 3); Degree is the number of degrees of intercooler core node, i.e. the outbound number.

The time overhead Ts of treatments B CP is 5 microseconds; The expense of optical crossbar switch OXC configuration is 15 microseconds; Transmit the time overhead T of BCP _fBe 3 microseconds.

The control channel bandwidth is 10Gb/s, and 1/2 bandwidth is used to the transmission of BCP signal, and the remaining OBS network operation that then is used for is as People Near Me discovery and link failure location etc.Control channel is two-way, and one is used to receive BCP, and another is with sending BCP.The time set that the intercooler core node is used to read and transmit BCP is 5 microseconds.

Although BCP transmits with broadcast mode, but happen suddenly or overweight traffic load still can cause BCP to lose, therefore when a plurality of BCP are received by the intercooler core node, these BCP are stored in the temporary buffer according to priority, in case too much the time spends in etc. when pending, BCP so that BCP has lost effectiveness, promptly takes place and loses in the situation that can cause offset time to run out.Whether therefore detect a BCP with following algorithm lost efficacy:

If Time〉offset time, then:

BCP abandons

Time else if〉offset time deducts the BNT time, then:

BCP abandons

$\mathrm{Time}=\underset{1}{\overset{\mathrm{CT}}{\mathrm{\Σ}}}\mathrm{BT}$

In the algorithm: Time is illustrated in each intercooler core node and waits for the processing time sum; CT is the abbreviation of the total number of intercooler core node; BT is the BCP pending time such as in temporary buffer, and BNT is meant the needed time of shortest path between the source node that shuttles back and forth light signal this BCP comprised and the Egress node; Offset Time (offset time) is T by formula _o=n * T _f+ T _OXCTry to achieve, can not lose the effect of carrying out resource reservation for it prior to its pairing bursty data DB in case inequality is set up expression BCP; And Offset Time-BNT is T by formula _o=n * T _f+ T _StThe offset time that+Priority*25 tries to achieve, in case inequality is set up expression, when BCP source fringe node began to send bursty data, BCP did not arrive the destination node of its bursty data, thereby had lost the effect that detects.

Therefore in a single day the phenomenon that above-mentioned BCP loses taking place, will cause the inconsistent phenomenon of being safeguarded in all fringe nodes of INSM, reduces fringe node to the intra-domain network state perception.Solution: one side as standard, is updated periodically the INSM that is safeguarded in other fringe nodes with a fringe node, is exactly the arrival of waiting for the low discharge load on the other hand, to eliminate the inconsistent phenomenon of INSM gradually.

As shown in Figure 2, it is as follows the fringe node method of work to be set:

(1) fringe node carries out work by the cycle.This cycle be meant a bursty data DB from finish be assembled into be sent to fully go the OBS network till;

(2) BCP of fringe node (Burst Control Packet, the burst control grouping) period is divided three phases:

I. after finishing burst assembling, network state model INSM in the inquiry field obtains in suitable route and the current OBS territory information such as idle wavelength collection.

Ii. according to resulting result of stage one, generate burst control grouping BCP.

Iii. send in this BCP to OBS network and go, simultaneously this BCP is saved to " BCP Buffer " lining.

(3) the collision detection period: constantly carry out collision detection, form by three phases to receiving BCP:

I. fringe node periodically receives the BCP that sent from the OBS core network and the own BCP that is generated, and is stored in corresponding according to priority classification " BCP Buffer " lining.

Ii. after receiving a BCP, fringe node detect this BCP whether be kept at " BCP Buffer " BCP of lining conflicts mutually.If there is conflict, the BCP of fringe node low priority is set to disarmed state, really do not abandon this BCP, and if low priority BCP exist " " this BCP pairing reservation in the network resources model INSM in the territory also will be cancelled in the lining to BCP Buffer.

If iii. do not detect conflict, fringe node will remove to upgrade intra-domain network state model INSM with this BCP, so that fringe node can in time be understood current network state.

(4) the bursty data DB period: mainly finish bursty data and send, and generate corresponding statistical information, divide three phases:

I. be stored in when clock controller detects " BCP Buffer " when the pairing time slot of the BCP of lining finishes, when promptly offset time arrives,, the source node of this BCP itself will dispose optical switch OXC if being this fringe node, carry out the preparation of transmission;

Ii. lasting transmission bursty data enters in the OBS network and goes;

Iii. with the information that BCP comprised, as the bursty data size, filling rate, and real-time information (as postponing end to end) is sent to statistical module.

As shown in Figure 3, it is as follows the core node method of work to be set:

(1) core node carries out work by the cycle.This cycle is meant receives a BCP, till the configuration of finishing the pairing bursty data DB of this BCP.

(2) (Virtual Configuration Table VCT), promptly when receiving a burst control grouping, only registers, and does not carry out real distribution at core node applying virtual allocation list.When the pairing time slot of this BCP finishes, when just offset time arrives, just carry out optical switch OXC configuration.Therefore, core node only need dispose optical switch OXC according to received BCP, comprises two periods: collision detection period and configuration period.

The collision detection period: mainly receive BCP and classification storage, detect conflict by collision detection algorithm again.Mainly formed by three phases:

I. core node periodically receives the BCP that is sent from the OBS core network, and is stored in corresponding " BCP Buffer " lining according to priority classification.

Ii. when receiving a BCP, the start detection algorithm detect this BCP whether be stored in " BCP Buffer " other BCP of lining conflicts mutually, if detect conflict, it is invalid that the BCP of low priority is set to, if the BCP of this priority registers in virtual configuration Table V CT, also the registration entries at VCT to be removed.The VCT here comprises the form of information (as information such as bursty data time of advent, duration and routes) among the BCP for record.

If iii. do not detect conflict, core node is registered in virtual configuration Table V CT with the relevant information among the BCP.

The configuration period is mainly the bursty data DB that is about to arrive and disposes optical switch, specifically is made up of three phases:

When i. in clock controller detects VCT, having bursty data to be about to arrive, configuration optical switch OXC;

Ii. the duration of optical switch OXC, pass through OXC fully until whole burst;

Iii. when finishing when once disposing, generate corresponding statistical information: as the throughput of node, the size of average bursty data etc.

Among Fig. 5, a pair of farthest nodal distance is 60km, and light signal transfer rate in optical fiber is 200, and therefore 000km/s introduces a time slot T who is used to detect potential conflict _TsBe 3.0 microseconds.

This method for obligating resource is preferential by maximum idle wavelength paths down; The burst packing algorithm is the prominent assembling mode based on length threshold, and its length threshold value is 1000Kbytes.

As shown in Figure 4, dynamic QoS support method is:

A) two parameters are set:

The role pressure index (Task Pressure Index, TPI) and pass the time (Elapse Time, ET).Role pressure index TPI represents the busy extent of a node; With the TPI initial setting is 5 grades, i.e. TPI ₁, TPI ₂TPI ₅, the index number of TPI is big more, represents that this fringe node is busy, and vice versa, and it is defined as follows shown in the face algorithm:

Definition: Percentage=has used number of wavelengths/(having used number of wavelengths+wavelength available number)

If: Percentage〉80% then TPI sets TPI ₅

Else if: Percentage〉60% TPI is set at TPI ₄

Else if: Percentage〉40% TPI is set at TPI ₃

Else if: Percentage〉20% TPI is set at TPI ₂

A TPI is set at TPI else if: Percentage 〉=0% ₁

Another parameter is to pass time ET, represents the time that pairing BCP is experienced in the OBS network.Here be N different class with bursty data flow type definition, C by name respectively ₁, C ₂C _N, and according to descending, i.e. C ₁Represent the bursty data of limit priority, and C _NRepresent the bursty data of lowest priority, it is high more responsive more to the time to set the bursty data flow priority simultaneously.

ET＝T _CT-T _BCT (5)

T wherein _CTBe present clock in the OBS network; T _BCTBe the burst control grouping BCP rise time;

The bursty data C of three kinds of discharge patterns of concrete definition ₀, C ₁And C ₂C wherein ₀Priority the highest, corresponding to real-time data on flows; C ₂Priority minimum, corresponding to the data on flows of transmitting as possible; And C ₁Be both compromise.And with three's proportional control between 20%, 20% and 60%.Be concrete setting below;

(1) bursty data Type C ₀Be used for representing real time data to time-sensitive.Because C ₀Responsive to time ratio, add C ₀Priority is the highest, and the bursty data of other type is always avoided C ₀Bursty data or seized by it, therefore, C ₀The bursty data of type is further optimized, and cancellation collision detection time slot is to reduce end-to-end delay.After the cancellation collision detection time slot, C ₀The end-to-end delay formula of bursty data is as follows:

T _e＝T _o+T _d+T _b；T _o＝n×T _f+T _OXC (6)

Here T _oRefer to the offset time between burst control grouping and the bursty data; T _dFor this bursty data experiences needed time of selected route, T _bBe the duration of transmitting; T _OXCBe the configuration needed time of optical switch; N is the number of intermediate node.

(2) bursty data Type C ₂And C ₁Owing to the time is not so good as responsive C ₀, here except that having increased the collision detection time slot, also increased the certain hour amount to increase the effect of collision detection at different priority with the bursty data of avoiding higher priority, increase after the collision detection time slot C ₂And C ₁The end-to-end delay formula of bursty data is as follows:

T _e＝T _o+T _d+T _b；T _o＝n×T _f+T _st+Priority*25 (7)

Here T _StFor introducing the time slot value of collision detection, value is 3.0 microseconds in this example; The same formula of other parameter (6).C ₁And C ₂Priority P riority value be made as 1,2 respectively.

(3) based on the consideration of controlling end-to-end delay, C ₁And C ₂The bursty data of discharge pattern is only dynamically adjusted once, therefore can obtain C ₁And C ₂The maximum end-to-end delay formula of two types of bursty datas:

T _e＝2×T _o+T _d+T _b?T _o＝n×T _f+T _st+Priority*25 (8)

B) collision detection: when two or more BCP reserved same outbound concomitantly, node had just detected conflict.

C) detect low priority BCP: after detecting conflict, by comparing the BCP that priority obtains one or more low priority, when priority between the afoul BCP equates, compare the role pressure index TPI between the BCP and pass time ET, all BCP can tell priority according to top three parameters.

D) regeneration BCP: in case when obtaining low BCP, this node judges whether self is the source node of low priority BCP, if, whether the time of then further judging this BCP and being experienced surpasses TTL (Time-To-Live, computing formula is referring to formula (6)), if do not have, then with the BCP of low priority, heighten one-level, i.e. C _iBe adjusted into C _I-1(2＜i＜N).After adjusting priority, regenerate BCP, and be sent in the OBS network and go.

TTL＝3×(T _d+n×T _s+T _b) (9)

Claims (10)

1. based on the method for obligating resource of intra-domain network state perception, comprise the collision detection step and the dynamic QoS support step of intra-domain network state perception; It is characterized in that the collision detection step of described intra-domain network state perception comprises:

1) mode with broadcasting sends to BCP in the OBS network and goes, and fringe nodes all in this OBS territory are safeguarded an intra-domain network state model with received BCP;

2) introduce a time slot (Tts) and be used for detecting potential conflict: the size of this time slot equals the distance between a pair of farthest node in light signal experience territory, i.e. needed time of network diameter in the territory; Its end-to-end delay is shown in following formula:

T _e＝T _o+T _d+T _b；T _o≥n×T _f+T _ts (1)

T _oRefer to the offset time between burst control grouping and the bursty data; T _dFor this bursty data experiences needed time of selected route, T _bBe the duration of transmitting, T _fBe meant and transmit needed time of BCP, T _TsIt is the conflict time slot; T _OXCBe the configuration needed time of optical switch; N is the number of intermediate node;

3) the fringe node job step is set: control generates BCP, sends BCP to OBS network, carries out collision detection to receiving BCP, finishes bursty data and sends, and generate corresponding statistical information;

4) the core node job step is set: carry out collision detection to receiving BCP, dispose optical switch OXC according to the collision detection result.

2. method for obligating resource according to claim 1 is characterized in that, the described fringe node job step that is provided with further comprises:

(1) be set the fringe node work period, this cycle be a bursty data DB from finish be assembled into be sent to fully go the OBS network till;

(2) divide three phases with the BCP period of fringe node:

I. after finishing burst assembling, network state model INSM in the inquiry field obtains in suitable route and the current OBS territory information such as idle wavelength collection;

Ii. according to resulting result on last stage, generate burst control grouping BCP;

Iii. send in this BCP to OBS network and go, simultaneously this BCP is saved to " BCP Buffer " lining;

(3) fringe node collision detection: constantly carry out collision detection to receiving BCP;

(4) the bursty data DB period: mainly finish bursty data and send, and generate corresponding statistical information.

3. method for obligating resource according to claim 1 is characterized in that, the described core node job step that is provided with further comprises:

(1) be set the core node work period, this cycle is for receiving a BCP, till the configuration of finishing the pairing bursty data DB of this BCP;

(2) at core node applying virtual allocation list, promptly when receiving a BCP, only register, do not carry out real distribution; When the pairing time slot of this BCP finishes, just carry out optical switch OXC configuration.

4. method for obligating resource according to claim 1 is characterized in that, described dynamic QoS supports that algorithm comprises the following steps:

A) two parameters are set: the role pressure exponential sum passes the time;

B) collision detection: when two or more BCP reserved same outbound concomitantly, node had just detected conflict;

C) detect low priority BCP: after detecting conflict, by comparing the BCP that priority obtains one or more low priority, when priority between the afoul BCP equated, relatively the role pressure exponential sum between the BCP passed the time, tells priority according to top three parameters;

D) regeneration BCP: in case when obtaining low BCP, this node judges whether self is the source node of low priority BCP, if whether the time of then further judging this BCP and being experienced surpasses TTL, if nothing then with the BCP of low priority, is heightened one-level; After adjusting priority, regenerate BCP, and be sent in the OBS network and go.

5. method for obligating resource according to claim 2 is characterized in that, described fringe node collision detection is made up of three phases:

I. fringe node periodically receives BCP and the own BCP that is generated that is sent from the OBS core network, and is stored in corresponding " BCP Buffer " lining according to priority classification;

Ii. after receiving a BCP, whether fringe node detects this BCP and conflicts mutually with the BCP that is kept at " BCP Buffer " lining; If there is conflict, the BCP of fringe node low priority is set to disarmed state, does not abandon this BCP, and if low priority BCP have " BCP Buffer " lining, cancel this BCP pairing reservations item in the network resources model INSM in the territory;

If iii. do not detect conflict, fringe node will remove to upgrade intra-domain network state model INSM with this BCP, so that fringe node can in time be understood current network state.

6. method for obligating resource according to claim 2 is characterized in that, the described bursty data DB period is divided three phases:

I. be stored in the pairing time slot of the BCP of " BCP Buffer " lining when clock controller detects when finishing, when promptly offset time arrives, itself will dispose optical switch OXC, carry out the preparation of transmission if the source node of this BCP is this fringe node;

Ii. lasting transmission bursty data enters in the OBS network and goes;

Iii. information that BCP comprised and real-time information are sent to statistical module.

7. method for obligating resource according to claim 3 is characterized in that, described core node disposes optical switch OXC according to received BCP and comprises two periods: collision detection period and configuration period.

8. method for obligating resource according to claim 7 is characterized in that, described core node collision detection step is: receive BCP and classification storage, detect conflict by collision detection algorithm again.

9. method for obligating resource according to claim 7 is characterized in that, described core node collision detection step specifically comprises following three phases:

I. core node periodically receives the BCP that is sent from the OBS core network, and is stored in corresponding " BCP Buffer " lining according to priority classification;

Ii. when receiving a BCP, whether the start detection algorithm detects this BCP and conflicts mutually with other the BCP that is stored in " BCP Buffer " lining, if detect conflict, it is invalid that the BCP of low priority is set to, if the BCP of this priority registers in virtual configuration Table V CT, also the registration entries at VCT to be removed;

If iii. do not detect conflict, core node is registered in virtual configuration Table V CT with the relevant information among the BCP.

10. method for obligating resource according to claim 7 is characterized in that, the described core node configuration period is divided three phases:

When i. in clock controller detects VCT, having bursty data to be about to arrive, configuration optical switch OXC;

Ii. the duration of optical switch OXC, pass through OXC fully until whole burst;

Iii. when finishing when once disposing, generate corresponding statistical information: as the throughput of node, the size of average bursty data etc.

Images