logging in or signing up vaidya Teresa1 Download Post to : URL : Related Presentations : Share Add to Flag Embed Email Send to Blogs and Networks Add to Channel Uploaded from authorPOINTLite Insert YouTube videos in PowerPont slides with aS Desktop Copy embed code: (To copy code, click on the text box) Embed: URL: Thumbnail: WordPress Embed Customize Embed The presentation is successfully added In Your Favorites. Views: 220 Category: Education License: All Rights Reserved Like it (0) Dislike it (0) Added: April 08, 2008 This Presentation is Public Favorites: 0 Presentation Description No description available. Comments Posting comment... Premium member Presentation Transcript Network Border Patrol Celio Albuquerque, Brett J. Vickers and Tatsuya Suda: Network Border Patrol Celio Albuquerque, Brett J. Vickers and Tatsuya Suda Jaideep Vaidya CS590F Fall 2000Need: Need End to End Congestion Control / Avoidance Mechanisms not enough. Unresponsive flows Rogue TCP stacks Network level mechanisms - Necessary Evil Try to push as much to the border of the network as possible Essential Idea: Essential Idea Core Stateless Congestion Avoidance Mechanism Exchange of feedback between Edge Routers Per-flow rate monitoring at Egress Routers Per-flow rate control at Ingress RoutersGoals: Goals Eliminate congestion collapse resulting from undelivered packets When combined with fair queueing, achieve approximately max-min fair bandwidth allocations for competing network flowsNBP Egress Router: NBP Egress Router NBP Ingress Router: NBP Ingress Router Feedback Control Algorithm: Feedback Control Algorithm Decides how and when feedback packets are exchanged between edge routers. Necessary for discovering source, communicate per-flow bit rates & detect network congestion by estimating RTT BFFs can be generated asynchronously. (RTT cannot be calculated in this case)FeedBack Packets: FeedBack Packets Rate Control Algorithm: Rate Control Algorithm Regulates rate at which each flow enters the network. Converge on set of per-flow transmission rates, preventing congestion collapse. Maximize link utilization. Similar to TCP congestion control (Slow start and Congestion avoidance phases) Handles synchronous and asynchronous packets differentlyRate Control Algorithm contd.: Rate Control Algorithm contd. Activated on receipt of feedback packet. Synchronous feedback Update baseRTT Calculate mrc (minimum rate change) Change rate based on phase Aysnchronous feedback Use old mrc and modify rate based on current phaseResults from Simulation Experiments: Results from Simulation Experiments Preventing congestion collapse Results from Simulation Experiments: Results from Simulation Experiments Max-min fairnessFairness Results: Fairness Results NBP by itself is not able to provide fairness. With WFQ or CSFQ, NBP provides approximate fairness, and avoids congestion collapse Results with WFQ are better than results with CSFQ. CSFQ’s fairness mechanism engages only when congestion is detected CSFQ is an approximation of WFQImplementation Issues: Implementation Issues Scalable Flow Classification Scalable inter-domain deployment Scalable fairness Incremental Deployment Multicast Multi-path routing Integrated or Differentiated serviceConclusion: Conclusion Pros Good paper. Limitations noted. Stop gap solution. Adequate for goals noted. Cons Incremental deployment not easy. Deployment based on whether we expect QoS technology to be available soon. Overload is directly proportional to number of flows. Would not work well with HTTP 1.0 (more number of flows)Questions?: Questions? You do not have the permission to view this presentation. In order to view it, please contact the author of the presentation.
vaidya Teresa1 Download Post to : URL : Related Presentations : Share Add to Flag Embed Email Send to Blogs and Networks Add to Channel Uploaded from authorPOINTLite Insert YouTube videos in PowerPont slides with aS Desktop Copy embed code: (To copy code, click on the text box) Embed: URL: Thumbnail: WordPress Embed Customize Embed The presentation is successfully added In Your Favorites. Views: 220 Category: Education License: All Rights Reserved Like it (0) Dislike it (0) Added: April 08, 2008 This Presentation is Public Favorites: 0 Presentation Description No description available. Comments Posting comment... Premium member Presentation Transcript Network Border Patrol Celio Albuquerque, Brett J. Vickers and Tatsuya Suda: Network Border Patrol Celio Albuquerque, Brett J. Vickers and Tatsuya Suda Jaideep Vaidya CS590F Fall 2000Need: Need End to End Congestion Control / Avoidance Mechanisms not enough. Unresponsive flows Rogue TCP stacks Network level mechanisms - Necessary Evil Try to push as much to the border of the network as possible Essential Idea: Essential Idea Core Stateless Congestion Avoidance Mechanism Exchange of feedback between Edge Routers Per-flow rate monitoring at Egress Routers Per-flow rate control at Ingress RoutersGoals: Goals Eliminate congestion collapse resulting from undelivered packets When combined with fair queueing, achieve approximately max-min fair bandwidth allocations for competing network flowsNBP Egress Router: NBP Egress Router NBP Ingress Router: NBP Ingress Router Feedback Control Algorithm: Feedback Control Algorithm Decides how and when feedback packets are exchanged between edge routers. Necessary for discovering source, communicate per-flow bit rates & detect network congestion by estimating RTT BFFs can be generated asynchronously. (RTT cannot be calculated in this case)FeedBack Packets: FeedBack Packets Rate Control Algorithm: Rate Control Algorithm Regulates rate at which each flow enters the network. Converge on set of per-flow transmission rates, preventing congestion collapse. Maximize link utilization. Similar to TCP congestion control (Slow start and Congestion avoidance phases) Handles synchronous and asynchronous packets differentlyRate Control Algorithm contd.: Rate Control Algorithm contd. Activated on receipt of feedback packet. Synchronous feedback Update baseRTT Calculate mrc (minimum rate change) Change rate based on phase Aysnchronous feedback Use old mrc and modify rate based on current phaseResults from Simulation Experiments: Results from Simulation Experiments Preventing congestion collapse Results from Simulation Experiments: Results from Simulation Experiments Max-min fairnessFairness Results: Fairness Results NBP by itself is not able to provide fairness. With WFQ or CSFQ, NBP provides approximate fairness, and avoids congestion collapse Results with WFQ are better than results with CSFQ. CSFQ’s fairness mechanism engages only when congestion is detected CSFQ is an approximation of WFQImplementation Issues: Implementation Issues Scalable Flow Classification Scalable inter-domain deployment Scalable fairness Incremental Deployment Multicast Multi-path routing Integrated or Differentiated serviceConclusion: Conclusion Pros Good paper. Limitations noted. Stop gap solution. Adequate for goals noted. Cons Incremental deployment not easy. Deployment based on whether we expect QoS technology to be available soon. Overload is directly proportional to number of flows. Would not work well with HTTP 1.0 (more number of flows)Questions?: Questions?