Posted: Sun Mar 05, 2006 3:35 pm
Some will, some wont. Many of them will just have many servers running separately all with unique RIPE addresses (external IP addresses). Then when they fill one server too much they'll start adding accounts on a new server with a new IP.nickman013 wrote:Oh Thats cool. Thanks for the info. I am guessing all webhosts have clusters.
An example of somewhere clustering is used would be google. I dread to think how many servers google has but I can guarantee they'll be running a huge cluster. The project I'm employed on uses a cluster of MySQl/Linux servers too since we're serving data to a lot of schools and get literally hammered with requests during the day. Setting the cluster up was one of the first things I worked on when I started that job before I ever started writing code
You can also set something up called LVS that sits at the gateway to your cluster. The LVS is a server which acts as a load balancer. If a server is already busy dealing with a request the LVS will look for another one and pass the request over to that. The implication being, if you only have one LVS and it goes down nobody will be able to connect
A common practise with DNS is also something called "Round Robin" which basic cycles connections around multiple servers rather than just the one... it's just another way of bringing the load down on a network
As you can tell by my rambling... there's a lot of options and a lot of useful things you can do with clusters