An empirical study of active networks

ABSTRACT

The bandwidth capabilities of computer networks has increased extensively over the past decade, driven mainly by the need expressed by end-user applications. The network uses call admission algorithms to allocate its resources before it allows a call to start transmitting packets through the network. As part of this admission process, the network should be able to provide a certain QoS (Quality of Service) guarantee for the call. The traffic generated by the different calls using the same links tends to be bursty, leading to congestion problems in the network. Congestion is primarily noticed when packets arrive at a node and occupy all the available buffer spaces. Newly arriving packets finding a buffer full have to be discarded. These dropped packets increase the probability of blocking of packets for the individual calls, violating their start up QoS guarantees. In this paper, we study how we could use an Active Network model to minimize the congestion at a network node. Processors attached to a node select certain packets in the buffer and employ a compression algorithm in order to decrease their size. This processing of packets allows for more space to be freed in the buffer thus helping alleviate the congestion problem. We present a simulation study of an active network node model, proposed by McKinnon et al. The results show that under certain fixed conditions when the processor rate and the service rate are equal, increase in the compression ratio does not lead to significant gains in network performance. An intuitive explanation of this result is also provided.

REFERENCES

• 1: S. Keshav, An engineering approach to computer networking: ATM networks, the Internet, and the telephone network, Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1997
• 2: S. Bhattacharjee, M. McKinnon, and E. Zegura. Resource Allocation in an Active Node Based on Buffered Traffic Levels. working paper.
• 3: X. Luo. A Simulation Study of Active Networks. Master's thesis, Department of Computer Science, Georgia State University, December 1999.
• 4: D. Tennenhouse, J. Smith, G. Sincoskie, and D. Wetherall. A survey of active network research. IEEE Communications Magazine, 35(1)(1), January 1997.
• 5: H. Perros and K. Elsayed. Call Admission Control: A Review. IEEE Communications Magazine, 35(11):82--91, November 1996.
• 6: J. derMerwe and S. Rooney. The Tempest -- A Practical Framework for Network Programmability. IEEE Network, May 1998.
• 7: David L. Tennenhouse , David J. Wetherall, Towards an active network architecture, ACM SIGCOMM Computer Communication Review, v.26 n.2, p.5-17, April 1996 [doi>10.1145/231699.231701]
• 8: Kenneth L. Calvert, Samrat Bhattacharjee, Ellen W. Zegura, and James Sterbenz. Directions in Active Networks. IEEE Communications Magazine, 1998.
• 9: Samrat Bhattacharjee, Kenneth L. Calvert, and Ellen W. Zegura. On Active Networking and Congestion. Technical report git-cc-96-02, College of Computing, Georgia Tech, 1996.
• 10: G. Hjálmtýsson and S. Bhattacharjee. Control-on-Demand: An Efficient Approach to Router Programmability. IEEE Journal on selected areas in Communications, 17(9):1549--1560, 1999.