CSE581 – Internet Technology

Paper Group 4

Phil Cayton

03/04/02

 

·        Bibliographic Information:

1. Hari Balakrishnan, V. Padmanabhan, S. Seshan, M. Stemm, R. Katz, TCP Behavior of a Busy Internet Server: Analysis and Improvements. Technical Report UCB/CSD-97-966, University of California, Berkeley, CA, August 1997.

2. H. Balakrishnan, H. S. Rahul, and S. Seshan. “An Integrated Congestion Management Architecture for Internet Hosts”. In Proceedings of ACM SIGCOMM 1999, pages 175--187, Sep 1999.

3. L. Breslau, E. Knightly, S. Shenker, I. Stoica, and H. Zhang, "Endpoint Admission Control: Architectural Issues and Performance," In Proceedings of ACM SIGCOMM 2000, Stockholm, Sweden, August 2000.

4. Stefan Savage, Neal Cardwell, David Wetherall, and Tom Anderson, “TCP Congestion Control with a Misbehaving Receiver”, In ACM Computer Communications Review, October 1999.

·       In-Class Presentation: <slides.ppt>

 

·        Summary of Paper Group:

The whole paper group centered on TCP performance, and discusses ways in which TCP can be changed, tweaked, or otherwise improved to increase throughput, better control congestion and recover from loss faster.  All these papers were very well written and contained interesting (if somewhat dated) material.  The papers presented problems in TCP, discussed their impacts and suggested improvements that would make TCP perform better in the Internet environment. 

The “TCP Behavior of a Busy Internet Server: Analysis and Improvements” paper analyzes the performance of a very busy internet server used during the 1996 Atlanta Olympic games.  The authors discuss how performance was impacted by such a heavy amount of dense traffic. They suggest improvements such as: enhanced loss recovery, integrated congestion control and loss recovery, applications level multiplexing, persistent connections in HTTP and TCP-INT implementation. 

The second paper – “An Integrated Congestion Management Architecture for Internet Hosts” – presents a method of managing network congestion from an end-to-end perspective using a congestion manager.  Their architecture consists of an (optional) feedback mechanism and a mechanism which maintains network statistics and coordinates data streams between particular network nodes.  The goal of this congestion management scheme is to ensure stable congestion control, efficient flow scheduling, and better overall performance than “best-effort” networks. 

The third paper – “Endpoint Admission Control: Architectural Issues and Performance” – attempts to address the problem of efficient and timely admission control on a busy server.  It attempts to combine the benefits of Integrated Services and Differentiated Services architectures while diminishing the negative aspects of these two approaches.  The paper also provides a discussion of several scheduling mechanism topics (i.e., FIFO vs. Fair Queuing, coexisting with Best Effort Traffic, Multiple Levels of Service), and some endpoint probing algorithms (e.g., Acceptance Thresholds, Accuracy, and Thrashing).  It also provides opinions and studies to show what works and what could be improved.   

The fourth paper – “TCP Congestion Control with a Misbehaving Receiver” – is a very good discussion on some of the more accessible TCP vulnerabilities and suggests relatively simple methods to plug these holes.  This paper was the most interesting to me personally.  It contains details on three relatively simple “hacks” or “spoofs” an unprincipled Internet citizen can use in order to gain better bandwidth (at the expense of other users).  The hacks it discusses are: Ack-Division, Optimistic-Acking and Multiple-acks.  The authors explain that each of these attacks is relatively easy to implement and carry out.  The paper shows in great detail the effects and dangers of these attacks and suggests some very lucid changes to TCP which could alleviate these problems.