Multi-Criteria Path Finding Using Multi-Queues Based Bidirectional Search for Multiple Target Nodes in Networks

Abstract

The convergence of networking and cloud computing leads to an increasing number of cloud based services and applications, which require to be processed by network nodes with enough resources. There are multiple available nodes in the network and network operators need to find paths to forward the traffic to these nodes for processing. In addition, services and application may have diversified requirements on the paths in terms of bandwidth, delay, jitter, etc. Therefore, network operators need to determine the optimal paths considering multiple criteria. However, single-criterion based shortest path algorithms are often used to compute paths from a source node to one target node, which may result in uneven traffic distribution and low resource utilization. Furthermore, the paths obtained are only optimal to this criterion and may not satisfy the Service Level Agreement (SLA) requirements. Therefore, in this paper, we propose a multi-criteria path finding method to obtain all the Pareto-optimal paths from a source node to multiple target nodes (destinations). We extend an existing effective multi-criteria routing algorithm (ParetoBFS), and integrate a multi-queues based bidirectional search mechanism. Additionally, we use a topology sampling technique to accelerate path computation. We evaluate the performance of these path finding methods using various topologies with multiple criteria including bottleneck and additive types. Experimental results demonstrate that our approach can reduce the running time by 54% -89% in different topologies compared to the ParetoBFS algorithm. By employing topology sampling, we further improve our algorithm's speed while still finding most of the Pareto-optimal paths.INDEX TERMS bidirectional search, multi-criteria, path finding, path selection, routing.

Multiobjective Path Problems and Algorithms in Telecommunication Network Design—Overview and Trends

“…In paper [171], the problem of finding the set of efficient solutions of a MOPP seeking the calculation of paths from one source node to different destination nodes is tackled, for application to cloud computing. The resolution process is based on the extension of a Pareto BFS method proposed in [172], with which it is possible to enumerate all paths from one source node to a destination node and uses a pruning strategy to readily identify the paths which are not efficient.…”

“…In multiple cases, these computational studies also analyze the effects of the application of the proposed routing methods in terms of the resulting network performance measures, using some type of simulation of end-to-end service demands, depending on the network technological and architectural context and on relevant features of the network environment. Specific examples of such computational studies are in references [18,157,170,171]. Also, in a few of the methodological papers in Section 2, real-world applications are presented, see for instance [33,118].…”

Heuristic Greedy-Gradient Route Search Method for Finding an Optimal Traffic Distribution in Telecommunication Networks

“…inefficient use of network resources [19][20][21]; -lack of prompt response to overloads when generating traffic and changing the network topology in existing satellite routing protocols [1,[22][23][24][25][26][27][28][29][30]; -inability to work with different types of traffic [19,31]; -focusing on a single criteria when solving optimization problems with multiple objectives [19,[21][22][23][24][25][26][27][28][29][30][31].…”

“…In addition to various problem settings for finding optimal routes, there are various methods to find them: heuristic algorithms [19,28,29], greedy algorithms [21], gradient methods [31], reinforcement learning [20], and neural networks [23]. In [3], the authors discuss the importance of global adaptive routing in large-scale systems with a large coverage area.…”