Techniques for Website Sizing and Scaling
Techniques for Website Sizing and Scaling
Abstract
The performance of successful Web-based services has all the allure of a roller
coaster ride: accelerated fiscal growth
1
combined with the ever-present danger of
running out of server capacity. This talk presents a case study [1] based on the
author's own engagement at one of the hottest e-commerce Web sites in the world
2;
so hot,they had more than their fair share of ''
CNN
Moments'' (the curious can play Lotto with any of:
Amazon,
eBay,
Yahoo,
America Online,
EA-Games,
other...).
Several spreadsheet techniques will be
presented for forecasting both short-term and long-term trends in the
consumption of back-end server capacity. Two new performance metrics [2]
are introduced for site planning and procurement:
the Effective Demand,
|
|
^
Ucpu
|
= |
N
Â
m = 0
|
am Xm , |
|
where the Xm¢s are a set of selected performance metrics gathered from
monitored web servers, and the Doubling Period:
where a is the curvature parameter from a nonlinear regression model.
Finally, some remarks will be made about the impact of the growing market for multi-media
services over the web.
References
- [1]
-
Gunther, N. J. 2001.
''Performance and Scalability Models for a Hyper-growth e-Commerce Website,'' in
Performance Engineering: State of the Art and Current Trends,
(Eds.) Dumke, R., Rautenstrauch, C., Schmietendorf, A., Scholz, A.,
Springer Lecture Notes in Computer Science, # 2047.
Heidelberg: Springer-Verlag.
- [2]
-
Gunther, N. J. 2000.
The Practical Performance Analyst,
Lincoln, Nebraska: iUniverse.com Inc.
(Read it
online)
Footnotes:
1
This is still true for the more successful web sites,
despite what you may have heard or read about the newer ''dot-bom'' economy in the Silicon Valley.
2
As a consultant, I remain bound by legal Non-disclosure Agreements.
File translated from TEX by TTH, version 2.25.
On 23 Sep 2003, 10:31.