Using Round Robin DNS

Round robin allows you to perform a crude form of load distribution across multiple web servers (or other servers, for that matter). To use round robin, you create multiple Address records for the same IP address. When a client asks for an IP address, the DNS server will return all the resource records, like this:

www.mycompany.com           10.10.10.1 www.mycompany.com           10.10.10.2 www.mycompany.com           10.10.10.3

The next time a client asks for that name, the IP address list is rotated, like this:

www.mycompany.com           10.10.10.2 www.mycompany.com           10.10.10.3 www.mycompany.com           10.10.10.1

Clients will typically take the first address in the list, so the load is distributed among these three servers. Round robin does not take into account the load of each server, or whether these servers are online. If one of these servers in this example is offline, one out of every three web page queries will fail!

Round robin also does not perform content replication, so any data stored on a server that a user needs will not be replicated to the other two servers. If you use round robin to distribute load, you will need to plan around the need for 'sticky sessions,' since server-side data will not persist. Because a client does keep a record in its cache for much time (only about an hour), you may not experience a problem with shorter sessions, but this still could present a problem.

With a proper DNS structure in place, name resolution for your web site will go much smoother, and you can utilize technologies such as round robin to help make your site more scalable.