Characterizing the resource demands of TCP/IP
Communication costs can significantly affect the performance behavior of distributed applications. We consider the problem of estimating the resource demands of a commonly used network communication subsystem, namely TCP/IP. We show that the resource demands associated with TCP/IP vary significantly for different message sizes and behave in a non-intuitive manner. For example, larger messages can require less CPU time than shorter messages. We find that different implementations of TCP have implementation-specific signatures that describe the relationship between message size and CPU demand. These signatures can be used to support performance-oriented design and management of distributed applications. A characterization procedure for finding an implementation's signature is presented. The procedure is non-intrusive and uses measurement tools that are or can be made available in most Unix environments. In this way the procedure can be repeated by analysts on many platforms.
Unable to display preview. Download preview PDF.
- 1.M.A. Bauer, P.J. Finnigan, J.W. Hong, J.A. Rolia, T.J. Teorey, and G.A. Winters. Reference Architecture for Distributed Systems Management. IBM Systems Journal, 33(3):426–444, 1994.Google Scholar
- 2.Chran-Ham Chang, Richard Flower, John Forecast, Heather Gray, William R. Hawe, K. K. Ramakrishnan, Ashok P. Nadkarni, Uttam N. Shikarpur, and Kathleen M Wilde. High-performance TCP/IP and UDP/IP Networking in DEC OSF/1 for Alpha AXP. Digital Technical Journal, 5(1), Winter 1993.Google Scholar
- 3.David D. Clark, Van Jacobsen, John Romkey, and Howard Salwen. An Analysis of TCP Processing Overhead. IEEE Communications Magazine, pages 23–29, June 1989.Google Scholar
- 4.J. B. Postel (editor). Transmission Control Protocol. RFC 793, September 1981.Google Scholar
- 6.Connie Umland Smith. Performance Engineering of Software Systems. Addison-Wesley, 1990.Google Scholar
- 7.W. Richard Stevens. Unix Network Programming. Prentice Hall, 1990. ISBN 0-13-949876-1.Google Scholar