We are having problems on servers that host 1500+ domains. ASP will fail as well as Frontpage's ability to log in. If we isolate all sites to "high" process, then the machine runs out of RAM - it has 4GB of physical RAM.
If we have all sites as "medium" or "low", the machine does not have excessive CPU/RAM usage, but eventually after around 2-6 hours ASP will fail. We get a "New Application Failed" error message when ASP dies.
By doing a NetStop W3SVC or using IIS Reset, we can fix the problem
temporarily. How can we fix it for good?
While Microsoft does have tools to help tune Web servers for massive amounts of users, it sounds like you're already familiar with them. You're doing the right thing by isolating Web sites from each other. I would also assume that if you have 4GB of RAM configured, you're using Windows 2000 Advanced Server and have it configured to take advantage of the maximum 3GB of RAM. However, IIS is simply not built for this type of environment--it's tuned for smaller amounts of RAM.
I would strongly recommend changing your systems architecture. Generally, it is much more cost-effective to deploy multiple, smaller, cheaper Web servers. Start configuring hardware with two-processor Web servers with 1GB of RAM. Purchase as many of these as necessary to accommodate your user base.
Also, look forward to .NET Server and IIS 6.0. IIS 6.0 includes
features that greatly improve the reliability of Web servers operated in
your type of environment. Good luck.
This was first published in October 2001