Thesis
Exploiting network processors for low latency, high throughput, rate-based sensor update delivery
Washington State University
Master of Science (MS), Washington State University
2009
Handle:
https://hdl.handle.net/2376/101782
Abstract
GridStat is a novel publish-subscribe middleware framework which delivers rate-based streams of sensor updates for the power grid. Its novelty includes providing different subscribers to the same sensor update stream with different guarantees in terms of end-to-end delay, rate, and number of disjoint delivery paths. GridStat implementations on a generalpurpose, dual-core CPU have very good per-hop forwarding latency on the order of 0.1 milliseconds. However, they can only support approximately 20 thousand forward/sec. In this thesis, we consider the problem of how to exploit high-speed network processors to increase the forwarding throughput for a GridStat status router (publish-subscribe forwarding engine). This thesis presents an analysis of the problem, and a design for GridStats routing algorithm on top of the Intel IXP 2400 series network processor. Experimental analysis shows how it is possible to support up to approximately 1 million forwards/second with a forwarding latency of approximately 0.01 milliseconds.
Metrics
Details
- Title
- Exploiting network processors for low latency, high throughput, rate-based sensor update delivery
- Creators
- Kim Christian Swenson
- Contributors
- David E. Bakken (Degree Supervisor)
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University; Pullman, Wash. :
- Identifiers
- 99900525011201842
- Language
- English
- Resource Type
- Thesis