Benchmarking Server Performance, Part 5
by Heidi Brumbaugh
Balancing Benchmarks, Throwing Knives
No ringmaster who had to measure knife-throwing performance results ever had to deal
with the dizzying array of variables involved in a Web server setup. Although benchmark
developers can debate philosophically between the merits of standard vs. customized
measurements, what's best for you depends on what questions you want answered. If
you're shopping for a server and anticipate heavy traffic, you'll probably want to compare
standard benchmarks across a range of hardware/software configurations as part of your
fact-finding process. If you want to make performance upgrades on an existing system-
especially if your files are particularly large or involve a lot of CGIs or other processor-
intensive operations-customized benchmarks are likely to give you more meaningful
numbers.
Now for the fun part: Just as there's debate over which types of measurements are best,
there's disagreement among experts over whether they're useful at all.
The problem here isn't bad benchmarks-it's good servers (which, come to think of it, isn't
really a problem at all). If you look at the benchmark numbers for low to moderate traffic
on average-sized sites, there's little variation between the servers' performance. You're
better off focusing your energy comparing other variables, such as cost, vendor reliability
and service.
David Strom, a technical journalist and industry consultant who has used and written
(webcompare.internet.com/bench.shtml)
on benchmarks extensively, doesn't think they should be weighed very heavily.
"Performance issues aren't as important-the differences in servers aren't as noticeable as
other access bottlenecks such as dial-up lines." He'd recommend you put more energy as
Webmaster into design, making your site easy to navigate. "If you're concerned about
performance, buy the fastest CPU and the most RAM you can afford."
Benchmarks would be most meaningful if you're running a corporate network, where your
connections are faster, or for heavy traffic sites. "Once you're getting 10,000 hits a day,
it's time to get serious about measuring performance."
[Move on to the last part of the article]
Web Developer Site Feedback
Web Developer
Copyright © 1999-2001 internet.com Corp.
All rights reserved.
