What I am looking at doing is placing a number of servers (x) in (x) number of locations (let's say 6). I know that this is widely done on very large websites but what I am looking at doing is hosting multiple domains and multiple "hosting accounts". The idea is to place all the servers on business broadband plans with all the servers being identical.

What I'm not sure of is what I need to research to do this. I plan on running RHEL or Debian and possibley run XEN, KVM or OpenVZ if a hypervisor is needed.

I'm guessing that there needs to be a DNS server or domain controller that receives all the requests and then forwards them to the appropriate server location. Is this where load balancing would take place or is that done somewhere else? Does the DNS/DC need to be a seperate machine or can it be on one of the hosts (It can be on it's own VM if needed).

Should this network always run a VPN to talk to each other or is there no need for that?

Any information on what I should research to make this work would be greatly appreciated. I'm pretty confident on the OS side (hypervisor/OS) but don't know what is the exact terminology for what I am looking at doing - pointers on where to look and or suggestions would be greatly helpful and appreciated.

Your question is a huge depends. Typically, a large scale, multi-datacenter configuration would look like the following.

The products and services you use to satisfy all of those components depends on your company culture, budget constraints, and technical prowess of staff.

Depends.

- How geographically disperse are these servers? Same datacenter, multiple datacenters with dedicated links between them, separate locations connected via internet? Based on this, the solution will be either a load-balancer for the web traffic itself, or a DNS round-robin.

- Do these servers share anything in the backend? Database, filesystem?