Balaji Prabhakar

We have designed a congestion control algorithm, QCN - Quantized Congestion Notification, for standarization in Data Center Ethernet (IEEE 802.1Qau). Several presentations were made during the course of the development of the QCN algorithm, showing how it evolved and the tests that were performed by our group and others. This work is a part of the efforts of the Data Center Bridging Task Group, and a current draft of the standard can be found here. A preliminary paper describing the QCN algorithm and a control-theoretic analysis of its stability is here. It turns out that the BIC, CUBIC and QCN algorithm operate according to a method, which we have called the Averaging Principle (AP), and this makes them more stable in the face of increasing round trip times, hence well-suited to the short-buffered environment typical in data centers. The AP is a general idea, applicable to any control system, not just congestion control systems. Slides from a recent talk provide an overview of QCN and AP.

Other topics of interest in Data Centers: designing a large, high-speed and low-latency interconnection fabric, and ensuring reliable transmission, despite errors on the wire, by using new types of coding.

List of Presentations for the QCN Algorithm at IEEE 802.1 QAU

Publications

1. M. Alizadeh, S. Yang, M. Sharif, S. Katti, N. McKeown, B. Prabhakar, S. Shenker, "pFabric: Minimal Near-Optimal Datacenter Transport," Proceedings of the ACM SIGCOMM 2013 , pp 435-446, June 2013

2. V. Jeyakumar, M. Alizadeh, D. Mazieres, B. Prabhakar, C. Kim, A. Greenberg, "EyeQ: Practical Network Performance Isolation for the Multi-tenant Cloud," REM , Vol. 1005, Issue A1, Pages A2, April 2013.

3. V. Jeyakumar, M. Alizadeh, D. Mazieres, B. Prabhakar, C. Kim, W. Azure, "EyeQ: Practical Network Performance Isolation at the Edge,” USENIX Association, Proceedings of the 4th USENIX conference on Hot Topics in Cloud Computing, p 8, December 2012.

4. M. Alizadeh, S. Yang, S. Katti, N. McKeown, B. Prabhakar, B., S. Shenker, "Deconstructing Datacenter Packet Transport," Proceedings of the ACM SIGCOMM 2013 , pp 435-446, June 2013

5. M. Alizadeh, A. Javanmard,B. Prabhakar, "Analysis of DCTCP:Stability, Convergence, and Fairness," Proceedings of the ACMSpecial Interest Group on Computer Systems Performance, SIGMETRICS 2011, June 2011.

6. M. Alizadeh, A. Kabbani, B. Atikoglu, B. Prabhakar, "Stability Analysis of QCN: The Averaging Principle," Proceedings of the ACM Special Interest Group on Computer Systems Performance, SIGMETRICS 2011, June 2011.
7. M. Alizadeh, A. Greenberg, D.A. Maltz, J. Padhye, P. Patel, B. Prabhakar, S. Sengupta, M. Sridharan, "Data center TCP (DCTCP)," Proceedings of ACM SIGCOMM 2010, August 2010.
8. A. Kabbani, M. Alizadeh, M. Yasuda, R. Pan, B. Prabhakar, "AD-QCN: Approximate fairness with quantized congestion notification for multi-tenanted data centers," Proceedings of the IEEE 2010 Symposium on High-Performance Interconnects, pp.58-65, August 2010.
9. M. Alizadeh, B. Atikoglu, A. Kabbani, A. Laksmikantha, R. Pan, B. Prabhakar, M. Seaman, “Data center transport mechanisms: congestion control theory and IEEE standardization,” Proceedings of the 46th Annual Allerton Conference on Communications, Control and Computing, September 2008.
10. Y. Lu, R. Pan, B. Prabhakar, D. Bergamasco, V. Alaria, A. Baldini, “Congestion control in networks with no congestion drops,” Proceedings of the 44th Allerton Conference on Communication, Control and Computing, September 2006.