This guide provides actual measurements of the factors effecting the speed of a web site. It provides some real world answers to frequently asked questions about Web Server Performance. We looked at the following factors:
During June and July 96 we used spare capacity on our OnTime Delivery monitoring network to monitor over 100 web sites in a dozen countries. These sites had all loaded our standard test page onto their server, and added details of their site configuration. The sites had responded to a posting in the web server newsgroups to take part in The Great Web Race. For their trouble, contestants got a free ranking of their server performance. For our trouble, we got a unique view of real world server performance.
OnTime Delivery is a speedometer for your web site. It is used by such major corporations as IBM, Hewlett-Packard and ICL to monitor the delivery performance of their web servers. Our system routinely reports the fast and slow page deliveries for each of our clients, as well as the average delivery time. From our experience, at most sites, the fast pages show little speed variation. Our assumption is this represents the server performance when the site is lightly loaded. This fast page measure was used in our analysis.
We would like to measure the factors affecting server performance under heavy load, but this would require much more cooperation from the test sites. Avoiding this hard question is not such a great loss. We believe you should aim to run lightly loaded most of the time, with a small spread between fast and slow pages. This is routinely achieved by some of our OnTime Delivery clients, and even by some quite small servers taking part in the Great Web Race.
Our test page consisted of a 2K text file, and a 30K jpeg file. This a fairly typical size for many of the pages we monitor. Our research has not shown a large difference between sending one large file or several smaller ones. This page was delivered on average in 22.6 seconds. The average for fast pages was 19.1 seconds, and for slow pages was 30.9 seconds. (See the detailed report for a definition of the terms "fast" and "slow".) The best servers delivered the page in under 14 seconds for fast pages.
The best performing servers were high end workstations (DEC, SGI) on T1 or better connections in North America. You should aim for a fast delivery time for our test page of about 14 seconds if your server meets these criteria. If not, to estimate fast delivery times:
If your server is delivering fast pages within this time, it is performing at the upper end of servers in its class. If you are content with your server meeting the average performance for its class, add 2 seconds to the above times.
We were surprised how little difference factors such as connection speed and hardware type made to actual performance. The overwhelming factor in delivery performance is server load. Moving from an ISDN line to a T1, or upgrading to a workstation may each chop one or two seconds off your delivery times. Load on your website can (and for most web sites does) add back ten or twenty seconds.
So our strong recommendation is:
A lightly loaded Mac or Pentium on an ISDN line can deliver pages several times faster than an overloaded Sun machine on a wide pipe. If the load is appropriate, a small server performs just fine. Under a large load, even a large server and Net connection will give poor results. Which leads to the obvious question:
To which we have an obvious reply. Buy a subscription to OnTime Delivery! For just $2 a week, you can download our test page to your server, and check that it is meeting the potential for its type, location and connection size. If your server is well matched to its load, it will deliver pages as well as most anything out there.
Last updated 20th Aug 96 by td@timedancer.com.