All about server farms, in just the one post?
Two ways basically;
Clustering (using software) and Hardware, or a combination depending on how resilient the network has to be.
Failover Clustering is mainly about availability, and is where the machines operate in a redundant capacity, so should something fail the next in line takes over.
Load Balancing (This is the more common server farm mode) is where a device handles the incoming requests and passes the tasks on to parallel computers (or clustered groups)
There are also High Performance Clusters which is the kind of thing that Google runs on, where the software is specially developed to run tasks in parallel on individual processing nodes.
Using a HPC does not mean that everything will run quicker BTW (in much the same way that a dual core CPU doesn't mean Windows will run any quicker  ) (Anybody remember the INMOS Transputer? )
|