Elsevier

Computer Networks

Volume 53, Issue 5, 9 April 2009, Pages 634-649
Computer Networks

Differentiated survivability with improved fairness in IP/MPLS-over-WDM optical networks

https://doi.org/10.1016/j.comnet.2008.11.001Get rights and content

Abstract

This paper addresses the priority-fairness problem inherent in provisioning differentiated survivability services for sub-lambda connections associated with different protection-classes in IP/MPLS-over-WDM networks. The priority-fairness problem arises because, high-priority connections requiring high quality of protection such as lambda level pre-configured lightpath protection are more likely to be rejected when compared to low-priority connections which may not need such a high quality of protection. A challenging task in addressing this problem is that, while improving the acceptance rate of high-priority connections, low-priority connections should not be over-penalized. We propose two solution-approaches to address this problem. In the first approach, a new inter-class backup resource sharing (ICBS) technique and a differentiated routing scheme (DiffRoute) are adopted. The ICBS is investigated in two methods: partial- and full-ICBS (p-ICBS and f-ICBS) methods. The DiffRoute scheme uses different routing criteria for the traffic classes. In the second approach, two rerouting schemes are developed. The rerouting schemes are applied with the DiffRoute and ICBS. The rerouting schemes employ inter-layer backup resource sharing and inter-layer primary-backup multiplexing for the benefit of high-priority connections, thus improving fairness. Our findings are as follows. (1) The application of p-ICBS and DiffRoute yields improved performance for high-priority connections. However, it shows penalized performance for low-priority connections. On the other hand, the collective application of f-ICBS and DiffRoute yields significantly improved performance for high-priority connections with no penalized performance as the performance of low-priority connections is also improved. (2) The rerouting schemes, when applied with the DiffRoute and ICBS methods, further improve the performance of high-priority traffic without significantly affecting the performance of other traffic.

Introduction

In IP/Multiprotocol label switching (MPLS)-over-WDM networks, a label switched path (LSP), a sub-lambda connection in IP/MPLS layer, traverses a sequence of lightpaths (LPs) or lambda connections in WDM optical layer where optical–electrical–optical (OEO) conversion, buffering, and electronic processing occur at MPLS routers between two consecutive lightpaths. Several LSPs can be multiplexed onto a LP. This kind of sub-lambda multiplexing is also referred to as traffic grooming and the architectures and algorithms for such traffic grooming have been proposed in the literature [1], [2], [3], [4].

Survivability or failure recovery is an important issue in these networks. Failure recovery at the optical layer, where a primary LP (P-LP) is protected by a backup LP (B-LP), is attractive because of fast recovery within a few tens of milliseconds [5]. In this approach, an LSP is routed over a sequence of LPs and each LP is protected by its own backup LP. In the optical layer recovery, a B-LP may be pre-configured (pre-configured LP protection) or it may not be pre-configured (non-pre-configured LP protection) depending on whether OXCs of the B-LP are configured a-priori or not. Note that, a pre-configured B-LP is associated with quick recovery as it is readily available. A non-pre-configured B-LP must be configured first before failure recovery. Recovery can also be done at the IP/MPLS layer by using a backup LSP (B-LSP) for a primary LSP (P-LSP). In this approach, a LP is not protected by its backup LP (LP is unprotected), where protection is done at the LSP level. This method has an advantage of efficient resource usage because of small bandwidth granularity. Provisioning sub-lambda connections with survivability requirements has been considered in [6], [7], [8], [9].

Providing differentiated survivability services based on the quality of fault tolerance has become an important issue. High-priority traffic may require very low recovery time while other traffic may not need such a high quality of fault tolerance. Adopting a multi-layer protection approach is a viable solution for providing differentiated survivability services for requests associated with different protection-classes. In the multi-layer protection approach, optical layer protection is generally preferred for the high-priority traffic because of its fast recovery and IP/MPLS layer protection is suitable for other low priority traffic. In the optical layer recovery, pre-configured LP protection is highly suitable for high-priority requests which are associated with mission-critical applications.

Multi-layer recovery approaches have been considered in works [10], [11], [12], [13]. In [10], coordinated recovery approaches in providing multi-layer protection were illustrated. Survivable traffic grooming problem was considered in [11] where three approaches – protection-at-lightpath level, mixed protection-at-connection level, and separate protection-at-connection level, were proposed. An architecture for providing coordinated way of controlling bandwidth reservations considering differentiated protection services was proposed in [12]. In [13], multi-layer protection methods with and with no backup lightpath sharing were proposed.

An important issue in provisioning differentiated survivability services using multi-layer recovery approaches is the inefficient resource usage. This problem arises because of the lack of resource-management in the backup provisioning of different traffic classes. Because of this, many requests, which require protection, are blocked in a dynamic traffic scenario. Therefore, it is essential that proper reservation of backup resources should be done to increase the acceptance rate of requests. A method to efficiently utilize resources is employing inter-layer backup resource sharing technique, where backup resources can be shared between two layers. This technique is derived from the common pool survivability concept [14]. The inter-layer backup resource sharing has been proposed/considered in [15], [12], [13].

When provisioning multi-layer protection based differentiated survivability services, it is important to address a fairness problem, which has not been considered in earlier research works. We refer this as priority-fairness problem, and we define it as follows. In a scenario where different priority connections are associated with differentiated survivability using multi-layer protection approaches, high-priority connections requiring high quality of protection such as optical layer pre-configured LP protection are more likely to be rejected when compared to low-priority connections which may not need such a high quality of protection (such as optical layer non-pre-configured LP protection or IP/MPLS layer protection). In this paper, the term fairness is used with respect to the opportunities available to admit different priority connections without regarding the amount of resources they require.

The priority-fairness problem stems from the following factors:

  • (1)

    Resource allocation for optical layer protection is difficult when compared with IP/MPLS layer protection because of the difference in protection bandwidth granularity. This problem becomes worse for optical layer pre-configured LP protection, because no resource sharing is allowed.

  • (2)

    Mission-critical applications are generally associated with large bandwidths as stringent delay requirements of these applications are usually translated into large bandwidths [16], [17].

  • (3)

    The intensity of non-mission-critical request arrivals are generally higher than that of mission-critical requests. Therefore, it is more likely that resources are occupied by non-mission-critical (low-priority) requests (Note that, though the intensity of mission critical requests are smaller than non-mission-critical requests, their admission is significant in terms of the importance/priority of associated traffic and the revenue that can be gained from them).

A challenging task in addressing this problem is that, while improving the fairness for high-priority mission-critical requests, low-priority connections should not be over-penalized.

In this paper, we assume single link failures as which are the predominant type of failures. To address the priority-fairness problem, we propose two solution-approaches. In the first approach, a new inter-class backup resource sharing (ICBS) technique and a differentiated routing scheme (DiffRoute) are adopted. In the ICBS, within a layer, backup resources of traffic which are associated with different protection-classes can be shared. This sharing technique is different from the inter-layer backup resource sharing technique, where backup resources can be shared between two layers. The ICBS is investigated in two methods: partial- and full-ICBS (p-ICBS and f-ICBS) methods. Several critical issues are addressed, which arise as a result of applying the ICBS when connections of different classes are allowed to traverse the same lightpath, such as (1) utilizing or modifying an existing protection to satisfy the protection of a new connection, and (2) releasing or updating resources on a release of a connection while preserving the protections of the other connections. The DiffRoute scheme uses different routing criteria for the traffic classes. An effective way of distributing signaling messages for failure recovery is proposed.

In the second approach, two rerouting schemes are developed. The rerouting schemes are applied with the DiffRoute and ICBS. The rerouting technique has been used earlier in [18] (rerouting at LP level), [19], [20] (rerouting at LP and connection levels). Our proposals are different from these works in the sense that the earlier works do not consider differentiated traffic classes and survivability, and ongoing traffic may be interrupted for rerouting, whereas in our work the rerouting is done in the context of multi-layer protection and differentiated-classes and ongoing traffic will not be interrupted in normal working conditions.

In addition to this, the rerouting schemes employ inter-layer backup resource sharing and inter-layer primary-backup multiplexing (which enables backup-LP resources to be shared by primary LSPs). The inter-layer primary-backup multiplexing is employed, such that no connection loses its recoverability in the event of a failure. The primary-backup multiplexing technique has been used in [21], [22], where it is applied in WDM layer only, whereas in our work it is applied in IP/MPLS and WDM layers. Through extensive simulation experiments, we investigate the performance of the proposals and verify their effectiveness.

We organize the paper as follows. Section 2 presents an overview of the differentiated survivability scheme. The first solution-approach is given in Section 3, where proposed backup sharing methods based on the inter-class backup resource sharing and the DiffRoute routing scheme are presented. In Section 4, the second solution-approach is given, where the rerouting schemes are developed. Implementation issues, signaling-distribution, and failure recovery methods are discussed in Section 5. The performance study of the proposals is given in Section 6. We conclude the paper in Section 7.

Section snippets

Problem statement

A network is represented as a weighted, directed graph G=(N,E), where N is a set of nodes and E is the set of links (edges) in the network. A node nN is an optical cross-connect (OXC) attached to a router. An edge eE is a lightpath (logical edge) or a wavelength link (physical edge) and is associated with attributes that carry information such as bandwidth usage and cost. A connection request is specified as s,d,b,c, where sN is the source node, dN is the destination node, b is the

Backup resource sharing methods

Backup sharing is an efficient way of improving resource utilization. For employing backup sharing, the following constraints should be satisfied: (1). backup resources should be link-disjoint with the primary path of the new request, (2). the corresponding primary paths (of the new request, and the existing request whose backup resource is to be shared) should be link-disjoint. We refer these constraints as link-disjoint constraints. In addition to this, it is also necessary to make sure that

Rerouting

We propose two rerouting schemes: REroute BACKup traffic based routing (REBACK) and REroute WORKing traffic on failure based routing (REWORK), to address the priority-fairness problem. The schemes do not cause any interruption for ongoing traffic during normal operation. Precisely, the rerouting operation will not cause any interruption in REBACK. In REWORK, the rerouting operation may cause interruption for non-mission-critical applications in the event of a component failure only. Further,

Implementation issues and failure recovery functionality

We consider integrated (or peer) model of IP-over-WDM networks. A central route server can be used to keep up-to-date knowledge of the status of the network resources including primary traffic information corresponding to backup LP links such as primary traffic classes and physical routes of the primary traffic. These information can be used to determine primary and backup paths of a new connection request which arrives at the central server. The OXCs along a lightpath can keep protection

Performance study

We evaluate the performance of the two solution-approaches through extensive simulation experiments on two network topologies: a randomly-generated network with 18 nodes and 28 bi-directional links (referred to as Random network), and the existing NSFNET with 14 nodes and 21 bi-directional links. We consider 8 wavelengths per fiber in both networks. Request arrivals follow Poisson distribution and holding time of a request follows exponential distribution with unit mean. The percentage of

Conclusions

In this paper, we addressed the priority-fairness problem inherent in provisioning differentiated survivability services based on multi-layer protection in IP/MPLS-over-WDM networks. We proposed two solution-approaches to address this. In the first approach, backup sharing methods based on a new inter-class backup resource sharing (ICBS) technique and a differentiated routing scheme (DiffRoute) were adopted. In the second approach, two rerouting schemes were developed. Through simulation

Krishanthmohan Ratnam (M’08) received the B.Sc.Eng. degree in Computer Engineering from the University of Peradeniya, Sri Lanka, in 2001. He received the Ph.D. degree in Electrical and Computer Engineering from the National University of Singapore in 2008, where he is currently on a research position in the Department of Electrical and Computer Engineering. He has served as a reviewer for many journals and conferences of international repute. His research interests include WDM optical circuit

References (22)

  • Q. Zheng et al.

    Multi-layer protection in IP-over-WDM networks with and with no backup lightpath sharing

    Computer Networks Journal

    (2006)
  • K. Zhu et al.

    A review of traffic grooming in WDM optical networks: architectures and challenges

    SPIE Optical Networks Magazine

    (2003)
  • H. Zhu

    A novel generic graph model for traffic grooming in heterogeneous WDM mesh networks

    IEEE/ACM Transactions on Networking

    (2003)
  • W. Yao et al.

    A link bundled auxiliary graph model for constrained dynamic traffic grooming in WDM mesh networks

    IEEE Journal on Selected Areas in Communications

    (2005)
  • C. Xin

    Logical topology design for dynamic traffic grooming in WDM optical networks

    Journal of Lightwave Technology

    (2006)
  • S. Ramamurthy et al.

    Survivable WDM mesh networks

    Journal of Lightwave Technology

    (2003)
  • S. Thiagarajan et al.

    Traffic grooming for survivable WDM mesh networks

    Optical Networks Magazine

    (2002)
  • Q. Zheng et al.

    Protection approaches for dynamic traffic in IP/MPLS-over-WDM networks

    IEEE Communications Magazine

    (2003)
  • K. Kar, M. Kodialam, T.V. Lakshman, Routing restorable bandwidth guaranteed connections using maximum 2-route flows,...
  • Sunggy Koo, G. Sahin, S. Subramaniam, “Cost efficient LSP protection in IP/MPLS-over-WDM overlay networks,” in:...
  • D. Colle

    Data-centric optical networks and their survivability

    IEEE Journal on Selected Areas in Communications

    (2002)
  • Cited by (4)

    • Bifurcation multi-path grooming algorithm in multi-granularity transport network

      2016, Xitong Fangzhen Xuebao / Journal of System Simulation
    • A state-based availability model to shared mesh protection in MPLS-TP networks with preemption support

      2012, Proceedings of the 2012 IEEE Network Operations and Management Symposium, NOMS 2012
    • Service-oriented resource multi-backupable method for multi-layer networks

      2011, Proceedings of 2011 1st International Symposium on Access Spaces, ISAS 2011

    Krishanthmohan Ratnam (M’08) received the B.Sc.Eng. degree in Computer Engineering from the University of Peradeniya, Sri Lanka, in 2001. He received the Ph.D. degree in Electrical and Computer Engineering from the National University of Singapore in 2008, where he is currently on a research position in the Department of Electrical and Computer Engineering. He has served as a reviewer for many journals and conferences of international repute. His research interests include WDM optical circuit switching networks, burst switching networks, and MPLS networks.

    Mohan Gurusamy (M’00-SM’07) received the Ph.D. degree in Computer Science and Engineering from the Indian Institute of Technology, Madras in 2000. He joined the National University of Singapore in June 2000, where he is currently an Associate Professor in the Department of Electrical and Computer Engineering. He has held a visiting position at Iowa State University, USA, from January–June 1999. He served as the lead guest editor for two special issues on Optical Networking Testbeds of the IEEE Communications Magazine (OCS), August 2005 and November 2005. His research interests are in the areas of high speed multi-wavelength optical circuit and burst switching networks, wireless sensor networks, Metro Ethernet, and grid networks. He has over 110 research publications to his credit and co-authored two books in the area of optical networks. Dr. Gurusamy has been a member of the IEEE since 2000 and senior member since 2007.

    Luying Zhou (M’01) received the B.S. and M.S. degrees in Automatic Control in 1982 and 1985, respectively, from South China University of Technology, Guangzhou, China, and the Ph.D. degree in Systems Engineering in 1990 from Xi’an Jiaotong University, Xi’an, China. From 1990 to 1995, he was a faculty member at South China University of Technology. From 1995 to 1998, he held research positions at SUNY in Buffalo and at Syracuse University, Syracuse, New York. He joined Kent Ridge Digital Labs, Singapore, in 1998. He is currently a Research Scientist at the Institute for Infocomm Research, A*STAR, Singapore. His research interests include optical WDM and access networks, wireless networks, Grid networking, and MPLS/GMPLS networks.

    Earlier versions of parts of the paper were presented at BROADNETS-2005 and LCN-2006 conferences.

    View full text