Fast restoration strategy with availability aware in optical burst switching networks

WANG Xianliang, LIANG Jian, XIE Yi，F(xiàn)U Yu，ZHANG Yan

(1. Chongqing Electric Power Company, Chongqing 400015, P.R. China; 2. Chongqing Institude of Telecommunications,Chongqing 401336, P.R. China)

?

Fast restoration strategy with availability aware in optical burst switching networks

WANG Xianliang1, LIANG Jian1, XIE Yi1，F(xiàn)U Yu1，ZHANG Yan2

(1. Chongqing Electric Power Company, Chongqing 400015, P.R. China; 2. Chongqing Institude of Telecommunications,Chongqing 401336, P.R. China)

Abstract:In order to solve the congestion of backup path caused by the local deflection under the general restoration scheme, a fast restoration scheme based on availability of perception for optical burst switching networks is proposed. The proposed restoration scheme apperceive the availability of back link according to the feedback of the node status message, then the restoration nodes can make the retransmission of damaged service adjust the traffic according to the situation of link congestion. And it can effectively avoid the node congestion caused by simple retransmission. Through detailed descriptive analysis and a comprehensive simulation study, the scheme based on the availability of perception has better burst loss performance, a good service-distinction performance, and enhances the capabilities of high-priority service.

Keywords：optical burst switch (OBS); packet loss probability; restoration; availability; traffic priority

Article ID：1673-825X(2016)02-0207-06

0Introduction

Optical burst switching (OBS) has been proposed as a promising technique to support high-bandwidth, bursty data traffic in the next-generation optical internet[1-3]. Due to the huge capacity of an optical fiber, a large amount of bursts could be lost when a network failure, such as a link failure occurs. It is critical important to ensure the normal operation of OBS networks, so the research on restoration scheme for OBS is significant.

In order to improve the speed of restoration, active restoration scheme with pre-computed routing is generally used, and the link capacity will set to be enough[4]. Reference[5] applies optimization techniques for static burst traffic to pre-compute explicit backup routes for various failure scenarios. According to the characteristics of the OBS network, the multi-protocol label switch (MPLS) 1+1 protection scheme is introduced into OBS network, but this protection scheme could only suit traffic with large data burst length[6]. Facing the link fault, a new restoration scheme is proposed to achieve the quick restoration speed[7]. In the scheme, the fault node deflect traffic to backup path, but it cause node congestion. In order to reduce the block rate, a restoration scheme based on load balance is proposed[8-9]. The scheme calculate work paths and backup paths by the routing algorithm based on load balance, but it needs a long restoration time. In these above restoration schemes, traffic in failure path will be removed to the backup path simply when a fault occurs, which lead to congestion of the backup path and decline of network throughput.

In order to avoid the congestion in backup link, network requires not only the operation of load balancing, but also the strengthening of perception ability to link availability. In view of these, a scheme based on availability of perception was recommended into traditional restoration scheme with the supporting of business distinguish. In this scheme, network balances the allocation for damage business according to the state of restoration paths’ congestion, its restoration capability to high priority business was improved.

1Quick restoration strategy

Shown in Fig.1, the network uses just in time(JIT) protocol[10]. Each core node in the network has the forward information to the next node of working path and backup path which can lead to the destination node. Running the 2-shortest-path routing algorithm, as one of the path is working path, the other is backup path. With the centralized management, all nodes in the network renew their forward information according to link state information from central network management system.

Fig.1　Fast restoration strategy

If a failure occur in a link belong to the path between nodeXand nodeD, nodeXfirstly forwards business to the backup pathL0with local restoration, then it produces and broadcasts alarm information. When nodeYreceives the alarm information, it shields the failure link firstly. Then, nodeYupdates and broadcasts the alarm information and selects the optimal path fromL1andL0as the restoration path. Lastly, the affected business will be forward. When alarm information reach source node, failure link will be shielded by source node and channel restoration will be started. Affected business will be forward by the restoration pathL2.

Fast restoration strategy takes different restoration strategy ant different restoration stages, so it has both advantages of the two different restoration strategies. In order to accelerate the speed of restoration, the failure front-end node takes local restoration and radio and alarm information. With the broadcast of the alarm information, each node can take the appropriate restoration mechanism based on the utilization of resources. So, the resource utilization is improved.

As one-way wavelength channel reservation is used in OBS network, it can easily lead to the competition of wavelength channel resources, which will result in data loss. Especially in the case of network failure, the business forwarding from failure path to backup path is more likely to cause congestion on the backup path, which result in the loss of a large amounts of data and the rapid decline of the core nodes business throughput. If the congestion situation can not be solved in time, it will further increase OBS network congestion and result in that the link is unavailable. Also, the network performance will deteriorate.

2Restoration scheme based on the availability of perception

The simple forwarding of node restoration for damaged business in can lead to congestion of backup path and reduce the availability of network. In order to solve these problems, the availability of perception scheme is introduced into fast restoration scheme. Restoration node make balance load distribution according to restoration path’s congestion information that reflected by restoration node. This new scheme can improve business restoration speed and improved network performance.

2.1Restoration program description

As show in Fig.2, when a link between nodeXand nodeDfails, the nodeXwill take local restoration and produce the control management packet (CMP). CMP will be send to destination nodeDalong the restoration pathL0. At the same time, nodeXwill produce and broadcast alarm information. As the CMP arrived, destination nodeDwill process and generate the confirmation packet Control Management Packet-Acknowledge (CMP-ACK). CMP-ACK return to nodeXalong pathL0periodically, the node on the way will fill their congestion status in CMP-ACK. NodeXadjust the forward traffic according to the node congestion information contained in CMP-ACK, the traffic can’t be forward successfully will be discarded.

Fig.2　Fast restoration scheme basedon availability of perception

When nodeYreceives alarm information, it will shield failure link and fresh the alarm information. After this, the node broadcasts the alarm information. Then, nodeYselects the optimal pathL0as default restoration path and sends CMP1and CMP2to destination nodeDthrough restoration pathL1andL0before the forwarding of fault business. Traffic load to theses two path are adjusted by nodeYaccording to cyclical CMP1and CMP2. If one path has congestion, nodeYreduces the window of this path and increases the window of the other link. If both the two paths are congested, the node discards burst data. When the congestion removed, nodeYincreases traffic load to the default restoration path. As the same, when alarm information arrived, source node shield fault path and start restoration path. After that, source node adjusts the traffic load toL0andL1andL2according to CMP-ACK which come from these paths.

2.2Availability determine and traffic shaping

In this paper, as a condition to determine the link availability is that whether the loss rate of the BDP on all paths which through the port of the node in a link is greater than the path available threshold or not. Here, only link unavailability caused by traffic congestion is considered. So, link availability threshold can be the node congestion threshold. Assume that the core node congestion threshold isDand the total loss rate of the node isPnode; assume that there areNpaths through node; assume that the total number of burst data which nodeireceived per unit time isDiand the number of lost burst data isDLi; assume that congestion relief threshold isD-δ;δis an arbitrary decimal. So

(1)

IfPnode>D, node congestion, path unavailability, set node congestion alarm indicationCFin CMP-ACK to be 1; IfPnode≤D-δ, congestion relief, restoration path availability, set node congestion indicationCFto be 0.

Core nodes on the restoration path send CMP-ACK including their congestion information to restoration node. Restoration node adjusts the traffic flow with corresponding measures according to link availability. When the value ofCFis 0, select the optimal restoration path to forward all affected business. When the value ofCFis 1, reduce the traffic flow with the adjustment strategy.

At present, there is no high-speed optical random access memory in OBS network and BDP does not have cache process, only fiber optic delay line (FDL) can be the buffer of optical data unit, which can only obtain limited delay. In core nodes, it is difficult to do random access to a large number of businesses[11]. When BDP arrived at core node, it will be discarded if there is no enough wavelengths. So, the traditional method to adjust node business sending rate is no longer applicable to OBS network.

In the paper，we prioritize the traffic and then adjust the traffic load forwarded by restoration node. Note that we regards the traffic with different priorities as adjusting object and the same priority to all affected BDP as the adjusting granularity .The flow adjustment mechanism is as follows:

1)Do not receive congestion alarm,default to the restore path without congestion,forward all affected traffic to destination node through optimal restoration path;

2)Receive congestion alarm,CF=0, restoration path without congestion, forward all affected traffic to destination node through optimal restoration path;

3)Receive congestion alarm,CF=1, restoration path congestion, inquire the priority of forwarding business in the current congestion path, forward low priority traffic on current congestion path to destination node through other restoration path. If there is no any other restoration path for this node, or other restoration paths do not have enough bandwidth resource, discard these low priority businesses. Then, forward the higher priority traffic to destination node through optimal restoration path, until all traffics are forwarded to destination node.

4)Receive congestion alarm,CF=0, restoration path congestion removed, forward the higher priority traffic to destination node through optimal restoration path, inquire the priority of forwarding business in the other congestion paths, forward the higher priority traffic in other restoration paths to destination node through the optimal restoration path, until all traffics are forwarded to destination node.

2.3Restoration program implementation process

According to whether restoration node receives alarm information or not, the implementation process of restoration program can be divided into two stages: restoration start-up phase and adaptive adjustment phase. Assume there are 3 kinds of traffic: Class0, Class1and Class2. Class0 has the highest priority, Class1 followed by, Class2 has the lowest priority. The implementation processes for every restoration stage are as follows:

In the restoration start-up phase, node generates alarm information and broadcast it as failure occurred. Then, this node inquires the restoration path. Before the forwarding of affected traffic, the node generates CMP and sends it to destination node through the control channel in restoration path. Meanwhile, the node receives CMP_ACK from destination node. CMP_ACK has congestion information of the restoration path. If node do not receive CMP_ACK, it forward all affected traffic to restoration path. If there are no enough resource in restoration channel, discard these business.

In the adaptive adjustment phase, when CMP_ACK arrived, node adjusts different priority traffic adaptively according to the congestion information carried by CMP_ACK. Fig.3 and Fig.4 ,respectively, are the adjustment process whenCF=0 andCF=1.

3Simulation results

We use OPNET software to evaluate the performance of the above design scheme. The topology used in the simulations was shown in Fig.5. Traffic source adopt ON / OFF model, and the ON∶OFF=1∶1. During ON, the generate interval of IP packet is Poisson, The burst is divided into three priority levels, in descending order, they are BE, AF, EF，and the intensity ratio of three types of traffic is 1∶2∶2. Each link has eight data channels and one control channel, the transmission rate of the channel is 2.5 Gbit / s, the propagation delay of the link is 0.2 ms.

Fig.3　Operation flowchart for restoration node when CF=0

Fig.4　Operation flowchart for restoration node when CF=1

Fig.5　Simulation network topology

Fig.6 is the relationship between Restoration time and load of different restoration scheme. It can be seen from the figure, Fast restoration scheme based on the availability of perception (Con_QoS) and link-based restoration scheme (Link) has a lower restoration time which close to an average of 0.506 ms. This is because the front-end node of the failure link take restoration operations in both scheme, and this method save the fault alarm transmission delay. While, Sub-path-based restoration scheme (Sub path) and channel-based restoration scheme (Path) take the restoration actions after the fault alarm information has been passed to restoration operation nodes and source nodes, so compared with Fast restoration scheme based on the availability of perception (Con_QoS) and link-based restoration scheme (Link), they have a large propagation delay, and resulting in the increased of restoration time.

Fig.6　Restoration time VS load

Fig.7 shows the packet loss probability under different thresholds for two mechanisms, when the load of node 2 is 0.5. NON_QoS stands for the restoration mechanism without availability of perception, CON_QoS and CON_QoS_1 and CON_QoS_2 adopt the restoration mechanism with availability of perception. Unavailability threshold is set at the values of 0.012 and 0.011 5 and 0.011. It can be seen from Fig.7 that NON_QoS has a better performance on packet loss probability than restoration mechanisms without availability of perception. This is because when the backup link is congested, packet loss probability is higher than unavailability threshold, NON_QoS adjust traffic load in time and discard lower priority traffic in order to guarantee link availability. For these reasons, packet loss probability is near the unavailability threshold.

Fig.8 shows packet loss probability under Con_QoS which collected by node 2 in different alarm cycle. Set parameterLoad=0.5,D=0.011. The Fig.7 illustrate that the shorter the operation at restoration node is, the smaller the swing of packet loss probability is. This is because when backup link congested, Con_QoS can adjust traffic in time. It discards lower priority traffic to guarantee link availability. When alarm cycle is lower, change rate to the number of lower priority packets forwarded by restoration node is lower; also the swing of packet loss probability at unavailability threshold in backup link is lower.

Fig.9 shows the total packet loss for different restoration mechanisms at node 2. NON_QoS stands for traditional restoration mechanism without availability of perception, CON_QoS stands for the restoration mechanism with availability perception. Availability thresholdDcan be set at the value 0.02 and 0.015 and 0.011. When load is less than 0.51, both of the two mechanisms have an increasing packet loss with traffic load increased. Also, the two mechanisms have the same packet loss. The reason for this is that packet losses of these two mechanisms have not yet reached the available threshold. However, when packet loss is higher than the available threshold, as the load increase, CON_QoS has a packet loss remained at the available threshold while NON_QoS has a packet loss increased. This feature proved that whenLoad>0.51, CON_QoS has a better performance on packet loss than NON_QoS. This feature results from the periodically traffic adjustment CON_QoS used, which effectively avoid link failure caused by congestion and promote success restoration probability. In addition, the lower availability threshold is, the lower packet loss is.

Fig.8　Packet loss probability at node 2 VS simulation time (Load=0.5，D=0.011)

Fig.9　Packet loss probability node 2 VS load

4Conclusions

In view of the problems existed in fast restoration mechanism, a fast restoration scheme based on availability of perception for optical burst switching networks is proposed in this paper. With the availability of perception for backup link, restoration node adjusts traffic adaptively according to link congestion stages. Traffic forwarding takes priority as the adjustment granularity, which achieve well service differentiation and enhance network restoration ability to high-priority business.

Through a comprehensive simulation study, the scheme based on the availability of perception has a better burst loss performance and a good service-distinction performance. It obviously reduces packet loss of high-priority business, improved network performance. Also, this mechanism promotes the restoration ability to damaged traffic, especially to high-priority business.

References

[1]WU G, ZHANG T, CHEN J, et al. An index-based parallel scheduler for optical burst switching networks[J]. Lightwave Technology, 2011, 29(18): 2766-2773.

[2]BELBEKKOUCHE A, HAFID A, GENDREAU M, et al. Path-based QoS provisioning for optical burst switching networks[J]. Journal of Lightwave Technology, 2011, 29(13): 2048-2063.

[3]AKAR N, KARASAN E, VLACHOS K G, et al. A survey of quality of service differentiation mechanisms for optical burst switching networks[J]. Optical Switching and Networking, 2010, 7(1): 1-11.

[4]XIN Y, TENG J, KARMOUS-EDWARDS G, et al. Fault management with fast restoration for optical burst switched networks[C]//Broadband Networks, 2004. San Jose, USA:IEEE Press, 2004: 34-42.

[5]ZHANG I J, WANG S, ZHU K, et al. Pre-planned global rerouting for fault management in labeled optical burst-switched WDM networks[C]//Global Telecommunications Conference, 2004. Texas, USA: IEEE Press, 2004: 2004-2008.

[6]GRIFFITH D,LEE S K.A1+1 protection architecture for optical burst switched networks[J].Selected Areas in Communications,IEEE Journal on,2003,21(9):1384-1398.

[7]KOZLOVSKI E, BAYVEL P. Link failure restoration in wavelength-routed optical burst switched (WR-OBS) networks[C]//Optical Fiber Communication Conference(OFC2003). Atlanta, USA: Optical Society of America, 2003:222-223.

[8]KIM L Y, LEE S K, SONG J S. A dynamic load-aware congestion control scheme in optical burst switching networks[J]. Photonic Network Communications, 2007, 13(3): 277-287.

[9]HUANG Y G, HERITAGE J P, MUKHERJEE B. A new fault-management method using congestion-avoidance routing for optical burst-switched networks[J]. Photonic Network Communications, 2008, 16(2): 117-125.

[10] KIM Y M, RYU M S, PARK H S. Performance comparisons of restoration techniques in optical burst switching networks[J]. Photonic Network Communications, 2009, 17(2): 171-181.

[11] HUANG Y, HERITAGE J P, MUKHERIEE B. Dynamic Routing with Preplanned Congestion Avoidance for Survivable Optical Burst switched Networks (OBS)[C]//Optical Fiber Communication Conference (OFC2005). Anaheim, USA: Optical Society of America, 2005: 3-7.

Biography:

Wang Xianliang(1975-), Male, Chongqing, M.S., Deputy senior engineer. His research interests include information and communication technologies. E-mail：wangxianliangcq@163.com.

Liang Jian(1974-), Male, Chongqing, M.S., Senior engineer. His research interests include power grid planning and communication technology research and management.

Xie Yi(1991-), Female, B.A.. Her research interests include information and communications operations.

Fu Yu(1984-), Male, B.A.. His research interests include Electric power communication.

(編輯：魏琴芳)

光突發(fā)交換網(wǎng)絡(luò)中帶有可用性感知的快速恢復(fù)策略

王賢亮1，梁健1，謝一1，付渝1，張炎2

(1. 重慶市電力公司, 重慶 400015; 2. 重慶電信研究院，重慶 401336)

摘要：為了解決光突發(fā)交換網(wǎng)絡(luò)中傳統(tǒng)恢復(fù)機(jī)制簡(jiǎn)單轉(zhuǎn)發(fā)造成備用路徑擁塞的問(wèn)題，提出了一種帶有可用性感知的快速恢復(fù)策略。該策略根據(jù)反饋的節(jié)點(diǎn)狀態(tài)信息感知鏈路的可用性，對(duì)轉(zhuǎn)發(fā)的受損業(yè)務(wù)進(jìn)行自適應(yīng)調(diào)整，從而有效避免簡(jiǎn)單轉(zhuǎn)發(fā)而導(dǎo)致的節(jié)點(diǎn)擁塞現(xiàn)象。仿真結(jié)果表明：帶有可用性感知的快速恢復(fù)策略具有較好的丟包性能，很好地實(shí)現(xiàn)了業(yè)務(wù)區(qū)分，增強(qiáng)了網(wǎng)路對(duì)高優(yōu)先級(jí)業(yè)務(wù)的恢復(fù)能力。

關(guān)鍵詞：光突發(fā)交換；丟包率；恢復(fù)；可用性；業(yè)務(wù)優(yōu)先級(jí)

DOI：10.3979/j.issn.1673-825X.2016.02.011

收稿日期：2014-09-24修訂日期：2014-11-27通訊作者：王賢亮 wangxianliangcq@163.com

基金項(xiàng)目：國(guó)家電網(wǎng)公司科技項(xiàng)目(SGTYHT/13-JS-175) The Science and Technology Projects of State Grid(SGTYHT/13-JS-175)

CLC number：TN929.11

Document code：A