Drupal has gotten some big money to spend on a server infrastructure. Now all they need is a good plan for that infrastructure. The one they have now and are planning to use is one that well, sucks! They are planning on using a dns round robin. Dns round robins are the very last thing a high volume site should consider. My experiences with SQL Server and round robin systems have lead me to investigate further into better ways. So here is an idea I came up with about 2 years ago for a classroom project. This was be fore I knew anything about Linux virtual servers or the like. We were simply trying to get something like wingate to respond to SQL Server requests.

The problem with dns round robin is that you don't get any real indication of when something is broken. By broken I mean not working to its everyday optimum. Let say Drupal.org is getting hit heavily and one of the webservers gets stuck. This would pull resources and slow down that server but the dns would still point to it. In this case the whole site would be slowed down by 50% or a percentage of that. If the server actually goes down then it gets even worse. Since the dns does not know that the seerver is down it would still point to it. Some Visitors would still get a busy site or nothing at all. This is pretty much what Drupal is experiencing with just one server. So while dns round robin wil split the load it will not ensure response levels or reliability. DNS redundacy also suffers from difficulties in interpreting traffic statistics, caching and other things depending on how the DNS is setup. You have to do things like open the network up to pinging so that pinging can be used to test for availablilty of a server on the network.

There is no database replication in the plan. Since Drupal.org uses Drupal CMS then this should be the first thing to be considered. The Drupal CMS relies exclusively on database connectivity. Much of that connectivity is has no error checking behind it.

There is no redundancy in their plan at all. Using something like Octagate switch, a software Reverse Proxy would lead to a more scalable and redundant infrastructure. Not to mention being more secure and easier to administrate.

This article brought to you by the Hiveminds Magazine - Staff. Contact us if you want to post an article or announcement anonymously