David A. Mosher

EECS Department
University of California, Berkeley
Technical Report No. UCB/CSD-85-238
June 1985

http://www2.eecs.berkeley.edu/Pubs/TechRpts/1985/CSD-85-238.pdf

This report presents a detailed timing analysis of the dynamic behavior of TCP/IP and UDP/IP as they are implemented in the released 4.2BSD version of the Berkeley UNIX operating system. The analysis is based on a series of performance measurements in the kernel, directed by a specific task. The organization of the code and the algorithms used in the implementation of each routine are discussed. The effects of this organization and these algorithms on the performance of each protocol are discussed. A model is proposed to estimate the minimum cost of a protocol implementation, and a comparison is made with the measured results. The results of this paper suggest that the implementation is responsible for a good deal of the overhead in the measured performance. In addition, the overheads of both light and medium weight protocol do not appreciably differ from each other, and are overshadowed by the performance penalties due to interface to the user and to the hardware.

BibTeX citation:

@techreport{Mosher:CSD-85-238,
    Author = {Mosher, David A.},
    Title = {A Study of An Internet Protocol Implementation},
    Institution = {EECS Department, University of California, Berkeley},
    Year = {1985},
    Month = {Jun},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/1985/5934.html},
    Number = {UCB/CSD-85-238},
    Abstract = {This report presents a detailed timing analysis of the dynamic behavior of TCP/IP and UDP/IP as they are implemented in the released 4.2BSD version of the Berkeley UNIX operating system. The analysis is based on a series of performance measurements in the kernel, directed by a specific task. The organization of the code and the algorithms used in the implementation of each routine are discussed. The effects of this organization and these algorithms on the performance of each protocol are discussed. A model is proposed to estimate the minimum cost of a protocol implementation, and a comparison is made with the measured results. The results of this paper suggest that the implementation is responsible for a good deal of the overhead in the measured performance. In addition, the overheads of both light and medium weight protocol do not appreciably differ from each other, and are overshadowed by the performance penalties due to interface to the user and to the hardware.}
}

EndNote citation:

%0 Report
%A Mosher, David A.
%T A Study of An Internet Protocol Implementation
%I EECS Department, University of California, Berkeley
%D 1985
%@ UCB/CSD-85-238
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/1985/5934.html
%F Mosher:CSD-85-238