By Topic

A scalable scheduling algorithm to avoid conflicts in switch-memory-switch routers

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Yang Xu ; Dept. of Comput. Sci. & Technol., Tsinghua Univ., Beijing, China ; Wu, B. ; Li, W. ; Bin Liu

Although output queued (OQ) switches are prominent for their high performance, they are not easy to implement due to the high speedup requirement. Using a special scheduling algorithm in the first stage switch, a more scalable switch-memory-switch (SMS) architecture can emulate an OQ switch, where cells must be transferred from the inputs to the shared memories per time slot without arrival and departure conflicts. Although scheduling algorithm achieves good performance, the time complexity for constructing the bipartite graph is too high to be used in practice. In this paper, we propose a new iterative random round-Robin matching (iRRM) algorithm together with its constrained version CiRRM, where no bipartite graph is required to be constructed in advance to solve the departure conflict, and thus high computation overhead is avoided. In our algorithms, both the arrival and the departure conflicts are melted in the iterations. Each iterations consist of two steps: request step and grant step, where randomness and more easily implemented round-robin principle are used respectively. Through theoretical analysis, we obtain that with M=2φ(N-1) shared memories, where N is the port number and φ is a constant larger than (2N-1)/(2N-2), iRRM/CiRRM can complete a matching within O(logM) iterations with high probability in M and the time complexity of CiRRM is only O(log2M/loglogM), which is much lower than prior algorithms.

Published in:

Computer Communications and Networks, 2005. ICCCN 2005. Proceedings. 14th International Conference on

Date of Conference:

17-19 Oct. 2005