Performance evaluation of optical mesh restoration schemes
Introduction
To accommodate the exponential growth of the Internet, transport networks based on wavelength-division multiplexing (WDM) technology [9] are increasingly being deployed in carrier networks. A WDM network provides lightpaths, high-capacity virtual links, to higher layer client networks. Each lightpath in a WDM optical network may operate at data rates of up to several gigabits per second. Since WDM networks carry such high volumes of traffic, a single failure of any of the network components may have severe consequences.
In order to maintain a high level of service availability, carriers and/or service providers usually over-provision their network with extra capacity to protect their working components. When a network failure occurs the restoration procedure has to be invoked by a pre-assigned restoration source node according to a pre-planned restoration mechanism depending on the availability of network resources and the quality of service requirements. In this paper restoration and protection are used interchangeably.
In this performance evaluation study, we consider schemes in which backup paths for links or paths are pre-computed, but wavelength assignment is either pre-computed (static) or done dynamically upon failure (dynamic). In static restoration we compare the performance of link and path restoration in terms of capacity requirement and restoration time. For the dynamic restoration, we consider several dynamic wavelength reservation schemes such as backward and forward reservation. Their performance is evaluated in terms of restoration speed and restorability.
Many papers on optical layer restoration have been published in the literature [1], [2], [3], [4], [5], [10]. Link-based protection methods and end-to-end path protection and restoration are described in [2], [3], [4], [5]. In [1], the authors propose a distributed method to do restoration in mesh networks of Optical Cross-Connects. They assume full wavelength conversion where a lightpath can use a different wavelength on each link along its path, whereas in our model we assume an optical network with no wavelength conversion capability. Working and restoration capacity assignment problems were addressed in [2]. In [7], the authors present capacity-assignment algorithms and signaling protocols for various restoration methods and evaluate their performance in terms of capacity efficiency, restoration speed and proportion of connections that can be restored successfully. In our study, we use the same framework as in [6], [7] with more restoration schemes such as link-based restoration and more dynamic wavelength reservation methods such as single/partial/full forward wavelength reservation. Some analysis on restoration time in optical networks is considered in [3] where a model for the restoration time calculation is presented but the model does not consider any message queueing delays. In our study as well as in [1], [6], [7], however, a significantly more detailed model is used, where a failure situation is simulated with consideration of possible queueing delays in each node.
The rest of the paper is organized as follows. In Section 2, we discuss restoration strategies with different options for wavelength assignment and reservation in optical WDM networks. Section 3 describes the network architecture including our basic node-link model. In Section 4, we explain restoration signaling protocols that are based on recent proposals of Generalized Multi-protocol Label Switching (GMPLS), enhanced with our extensions specifically designed for restoration purposes. In Section 5, we present simulation results on the performance of various restoration schemes in terms of capacity requirement, restoration speed and restorability, focusing on the trade-offs between them. Conclusions are given in Section 6.
Section snippets
Protection
There are two major categories for protection schemes. One is path protection where the restoration processes are done between end-nodes of lightpaths. The other is link protection where the two end-nodes of the failed link take care of the restoration procedures. Fig. 1 shows an example of each protection scheme. In path protection, each lightpath affected by the failure is re-routed on another link disjoint path by the source and destination node of that lightpath. In link protection, the two
Network architecture
In our network architecture that is same as the one found in [6], [7], each node of the network includes the units shown in Fig. 2. The message processor, optical cross-connect (OXC) controller, and message transmitter are represented as First In First Out (FIFO) queues with deterministic service times. All of the service times for these queues are set as input parameters to our simulation model since these rates correspond to computing power of CPU, transmission rate and speed of optical
Restoration signaling: an extension to GMPLS
In order to implement various restoration schemes we propose a simple extension to one of the signaling protocols of GMPLS, namely, resource reservation protocol-traffic extension (RSVP-TE) [11], [12]. RSVP-TE is a signaling protocol for GMPLS control plane proposed by IETF. In static wavelength assignment, our extension to RSVP-TE is to add a two-way hand shake between the end-points of a backup path to make sure that the restoration path is successfully activated. For dynamic wavelength
Simulation and numerical results
We used the NSFNET topology that has 14 nodes and 21 bi-directional (42 uni-directional) links. For a given number of lightpaths N, source and destination nodes were randomly chosen. Shortest path routing was used to select the primary path for each lightpath, and the first-fit WA was used for ST-WA.
The service rates are: message processing rate (P messages/s), cross-connect configuration rate (X cross-connects/s) and message transmission rate (T message/s). We used the following values for the
Conclusions
We have investigated the trade-offs between speed, capacity and restorability for various mesh restoration methods with the goal of providing guidelines for choosing the optimal restoration scheme. Link based protection may provide a faster restoration of affected traffic at the cost of more capacity required for protection purpose. On the other hand, end-to-end path protection requires less capacity than link protection. However, since restoration processes are managed by the two end nodes of
Acknowledgements
This work was supported in part by NSF Grant ANI 9973111 and by DARPA Grant N66001-00-18949 (co-funded by NSA).
References (12)
- R. Doverspike, G. Sahin, J. Strand, R. Tkach, Fast restoration in a mesh network of optical cross-connects, in:...
- et al.
Survivable WDM mesh networks, part I – Protection
- et al.
Survivable WDM mesh networks, part II – Restoration
- et al.
Optical layer survivability: a service perspective
IEEE Communications Magazine
(2002) - et al.
Optical layer survivability: an implementation perspective
IEEE JSAC
(2002) - G. Sahin, Service and restoration routing and wavelength assignment and restoration signaling in optical networks,...
Cited by (3)
Role of computational intelligence in nanophotonics technology
2013, Computational Nanophotonics: Modeling and ApplicationsAn ear-decomposition based approach for survivable routing in WDM networks
2005, Proceedings - International Conference on Advanced Information Networking and Applications, AINA