Activity Number:
|
398
|
Type:
|
Topic Contributed
|
Date/Time:
|
Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Physical & Engineering Sciences*
|
Abstract - #301535 |
Title:
|
Traffic Modeling and Performance Analysis of Commercial Web Sites
|
Author(s):
|
Li Zhang*+ and Cathy Xia+ and Zhen Liu and Mark Squillante and Naceur Malouch
|
Affiliation(s):
|
IBM T. J. Watson Research Center and IBM T. J. Watson Research Center and IBM T. J. Watson Research Center and IBM T. J. Watson Research Center and INRIA
|
Address:
|
P.O.Box 704, Yorktown Heights, New York, 10598, USA P.O.Box 704, Yorktown Heights, New York, 10598, USA
|
Keywords:
|
traffic analysis ; self-similarity ; sub-exponential ; performance ; waiting time
|
Abstract:
|
We study the performance of Commercial Web sites based on the access logs from the servers. A problem with existing access logs is the coarse granularity of times, e.g., arrival times. We demonstrate and quantify the significant differences in performance obtained under different assumptions about the arrival process of user requests derived from the access logs, where user response times can differ by more than an order of magnitude.
We observed self-similar behavior of the arrival process. The request size distribution follows a sub-exponential, but not heavy-tail distribution. We exploit the properties of the self-similar processes as a theoretical foundation for constructing the arrival process at finer time scales. Our approach maintains consistency between the properties of the arrival processes at both coarser and finer time scales. We further investigate various issues concerning the waiting times of user requests under an FGN arrival process together with light and sub-exponential tailed distributions. We provide conditions on the heaviness of the tail distributions that affect the dominant factor for the waiting time asymptotic.
|