Rate-distortion based link state update☆
Introduction
The tremendous growth of the global Internet has given rise to a variety of applications that require quality-of-service (QoS) beyond what is provided by the current best-effort IP packet delivery service. One of the challenging issues is to select feasible paths that satisfy different quality-of-service (QoS) requirements. This problem is known as QoS routing. In general, two issues are critical to QoS routing: the state distribution and the routing strategy [1]. The routing strategy is used to find a feasible path which meets the QoS requirements; this has been extensively studied in the literature [2], [3], [4], [5], [6], [7]. The state distribution addresses the issue of exchanging the state information throughout the network and can be further decomposed into two sub-problems: when to update and how to disseminate the link state information. This paper focuses on the state distribution, especially on the former sub-problem. A number of research works have also been reported on how to disseminate the link state information [8], [9], [10], [11], which is, however, beyond the scope of this paper.
Many proposed QoS routing solutions assume that accurate link state information is available to each node. However, this is impossible in real networks. Moreover, to facilitate accurate enough link state information would impose a significant bandwidth and processing overhead on the network resource, i.e., the network resource will be greatly consumed if an update of the link state information is triggered whenever a minor change to the QoS parameters occurs. Current link-state routing protocols such as OSPF [12] recommend that the link state is updated periodically with large intervals. For instance, a link disseminates its state information every 30 min in OSPF. Consequently, because of the highly dynamic nature of link state parameters, the link state information known to a node is often outdated. As a result, the effectiveness of the QoS routing algorithms may be degraded significantly. To overcome this problem, several link state update policies (threshold, equal class, and exponential class based update policies) have been proposed in [13]. Given a predefined threshold value (τ), an update is triggered in the threshold based update policy if ∣bc − bo∣/bo > τ, where bo is the last advertised value of the available bandwidth, and bc is the current available bandwidth. However, in equal class and exponential class based update policies, the bandwidth is partitioned into classes and an update is triggered whenever the available bandwidth crosses a class boundary. The only difference between them is that the bandwidth is partitioned into classes of equal size ((0, B), (B, 2B), …) in equal class based update policy, while it is partitioned into unequal classes, whose sizes ((0, B), (B, (f + 1)B), ((f + 1)B, (f2 + f + 1)B), …) grow geometrically by a factor f, in the exponential class based update policy, where B is a predefined constant. Recently, many works considering the effects of the stale or coarse-grained information on the performance of QoS routing algorithms were reported in the literature. In [14], extensive simulations were made to uncover the effects of the stale link state information and random fluctuations in the traffic load on the routing and setup overheads. In [15], [16], the effects of the stale link state information on QoS routing algorithms were demonstrated through simulations by varying the link state update interval. A combination of the periodic and triggered link state update is considered in [17]. Instead of using the link capacities or instantaneous available bandwidth values, Li et al. [18] used a stochastic metric, Available Bandwidth Index (ABI), and extended BGP to perform the bandwidth advertising.
As reviewed above, although many link state update policies have been proposed, there is still a lack of a rigorous theoretical foundation. As a result, they may not be efficient enough and may waste the network resource. In this paper, we will provide a theoretical framework for link state update, from which we will propose a high performance link state update mechanism. We theoretically demonstrate that from the perspective of QoS routing, our proposed link state policy outperforms its contenders in terms of the update rate and false blocking probability of incoming connections. The rest of the paper is organized as follows. The problem is formulated in Section 2. In Section 3, we provide an insight on link state update based on Information Theory. We propose an efficient link state update policy in Section 4, and present the simulation results in Section 5. Finally, concluding remarks are given in Section 6.
Section snippets
Problem formulation
A key challenge for a network simulation is the selection of the network topology and the traffic patterns. Owing to the constantly changing and decentralized nature of current networks, it is rather difficult to define a typical network topology applicable for exploring any protocols [19]. Moreover, the results of a simulation over different network topologies with different traffic patterns may vary dramatically. In this paper, the problem of designing an efficient link state update policy in
An insight from information theory
Many studies have been done to characterize the Internet traffic, revealing interesting facts such as Long-Range Dependence or multi-fractal behaviors [20], [21], [22]. Therefore, we cannot assume that the source signal (available bandwidth) is memoryless. Each QoS parameter of a link is viewed as a random process with memory that is independent of time in this paper. By applying the sampling theorem, each continuous random process can be converted into an equivalent discrete-time sequence of
The proposed link state update policy
For simplicity, we only consider the case that c1 = c2 = c in this paper, and thusThe key difference between our proposed link state policy and those in [13] is that instead of partitioning the bandwidth into classes of equal sizes or exponentially growing sizes, we divide the bandwidth into classes by taking into account of the requested bandwidth of the connections and the available bandwidth, for the purpose of minimizing the source distortion under a
Simulations
Note that given a blocking probability p of a connection on every single link and without considering the relation among links, its overall blocking probability to traverse through an h-hop path is: 1 − (1 − p)h. Moreover, the number of link state updates of a network is simply the sum of the updates of all the links in the network because each link updates its own link state information independently. Therefore, the performance of a link state update policy in a network can be reflected by its
Conclusions
In this paper, we have proposed an efficient link state update policy. Through theoretical analysis and extensive simulations, we have shown that it greatly outperforms its contenders, i.e., it achieves a much lower false blocking probability with a very low update rate. As a result, we can increase the performance of QoS routing using the proposed link state update policy.
Gang Cheng received both B.E. and M.E. degrees in Information Engineering from the Beijing University of Posts and Telecommunications, Beijing, PR China, in 1997 and 2000, respectively, and Ph.D. in Electrical Engineering from the New Jersey Institute of Technology in 2005. His research focuses on QoS guarantee issues in high speed networks, Internet routing protocols and service architectures, and modeling and performance evaluation of computer and communication systems, especially on QoS
References (22)
- et al.
Algorithm for efficient generation of link-state updates in ATM networks
Computer Networks and ISDN System
(1997) - et al.
An overview of quality of service routing for next-generation high-speed network: problems and solutions
IEEE Network
(1998) - et al.
On multiple additively constrained path selection
IEE Proceedings on Communications
(2002) - R. Guerin, A. Orda, QoS based routing in networks with inaccurate information: theory and algorithms, in: Proceedings...
- T. Korkmaz., M. Krunz, S. Tragoudas, An efficient algorithm for finding a path subject to two additive constraints, in:...
- L. Gang, K.G. Ramakrishnan, A prune: an algorithm for finding k shortest paths subject to multiple constraints, in:...
- S. Chen, K. Nahrstedt, Distributed QoS routing with imprecise state information, in: Proceedings of 7th International...
- et al.
Quality of Service routing for supporting multimedia applications
IEEE Journal on Selected Areas on Communications
(1996) - et al.
Topology broadcast algorithms
Computer Networks and ISDN Systems
(1988–1989) - B. Bellur, R.G. Ogier, A reliable, efficient topology broadcast protocol for dynamic networks, in: Proceedings of IEEE...
Cited by (10)
A scalable method for DCLC problem using hierarchical MDP model
2013, Computer CommunicationsA tree-based particle swarm optimization for multicast routing
2010, Computer NetworksCitation Excerpt :These solutions are found to be highly efficient. Based on rate-distortion analysis, Cheng and Ansari [19] investigate the issue of how to update the link state information efficiently and effectively, proposing a high performance link state policy that is capable of minimizing the false blocking probability of connections under a given update rate constraint. Shaikh et al. [17] explore the effects of stale link state information and random fluctuations in traffic load on the routing and setup overheads and investigate how the inaccuracy of link state information interacts with the size and connectivity of the underlying topology.
Minimum required information to achieve a performance target in a network with memoryless states
2015, IEEE Transactions on CommunicationsOn the interplay between resource-management overhead and performance in sensor networks: An information theoretic approach
2013, Proceedings - IEEE Global Communications Conference, GLOBECOMOn the cost of knowledge of mobility in dynamic networks: An information-theoretic approach
2012, IEEE Transactions on Mobile ComputingThe information theoretic perspective on context aware network design
2012, Advanced Materials Research
Gang Cheng received both B.E. and M.E. degrees in Information Engineering from the Beijing University of Posts and Telecommunications, Beijing, PR China, in 1997 and 2000, respectively, and Ph.D. in Electrical Engineering from the New Jersey Institute of Technology in 2005. His research focuses on QoS guarantee issues in high speed networks, Internet routing protocols and service architectures, and modeling and performance evaluation of computer and communication systems, especially on QoS routing, traffic scheduling, queue management, and congestion control in high speed networks. He is currently employed with VPISystems Inc.
Nirwan Ansari received B.S.E.E. (summa cum laude), M.S.E.E., and Ph.D. degrees from NJIT, University of Michigan, and Purdue University in 1982, 1983, and 1988, respectively.
Since 1997, he has been working as a full-time professor in the Department of Electrical and Computer Engineering at NJIT. He authored with E.S.H. Hou Computational Intelligence for Optimization (Kluwer, 1997, translated into Chinese in 2000), and edited with B. Yuhas Neural Networks in Telecommunications (Kluwer, 1994). He is a technical editor of the IEEE Communications Magazine, Computer Communications, the ETRI Journal, as well as the Journal of Computing and Information Technology. His current research focuses on various aspects of broadband networks and multimedia communications. His research has been supported by various federal and state agencies, and private industries. He has also contributed over 80 refereed journal articles, and numerous conference papers and book chapters.
He organized (as General Chair) the First IEEE International Conference on Information Technology: Research and Education (ITRE2003), and was instrumental, while serving as its Chapter Chair, in rejuvenating the North Jersey Chapter of the IEEE Communications Society which received the 1996 Chapter of the Year Award and a 2003 Chapter Achievement Award, served as Chair of the IEEE North Jersey Section and in the IEEE Region 1 Board of Governors during 2001–2002, and currently serves in various IEEE committees including as TPC Co-chair/Vice-chair of several conferences. He was the 1998 recipient of the NJIT Excellence Teaching Award in Graduate Instruction, and a 1999 IEEE Region 1 Award. He is frequently invited to deliver keynote addresses, tutorials, and talks.
- ☆
This work has been supported in part by the National Science Foundation under Grant 0435250, and the New Jersey Commission on Science and Technology via NJWINS.