A Virtual Private Server is simply a virtual machine. So, what makes one VPS different from another?
One obvious difference is the performance per dollar. When I look around at VPS providers, very few will tell you that you're getting a certain percentage of a known hardware platform. Today's platforms, such as the Intel 5xxx series processors or the AMD 1/2/8XXX processors are very, very powerful. So, even 10% of a CPU core can drive an ecommerce website successfully. But many VPS vendors put hundreds of VPSes onto a single server and provide performance that is hardly better than shared hosting. So it pays to know what you're getting.
Another difference between VPS vendors is bandwidth. What is the true bandwidth you're getting in your VPS, and even more importantly, how *good* is that bandwidth? What I mean by that is, how close (transit time wise) is your VPS to your viewers or customers? You can tell this by pinging various web sites and ISP hubs from your VPS - but how will you assess it before you buy a VPS? The best VPS vendors will have multiple high-speed connections from their data centers to major backbone networks of the internet, or even be hosted at meet-me points where major carriers' networks are terminated.
A more subtle but very important difference between VPSes is reliability. If your VPS is hosted on a standard server, if the server dies due to disk failure or other hardware problem, your VPS is offline. Perhaps the vendor has a manual method for moving your VPS from backup to another machine, but you'll be down a while and potentially lose data. Some vendors host their VPSes on highly available grid computing systems like AppLogic, which automatically restart the failed VPSes on another server within minutes of a failure, which together with redundant shared disk storage allow your application to pick up where it left off.
Vendors that I know of which have AppLogic hosted VPSes are
ENKI (
http://www.computingutility.com)
and Layered Technologies (
GridLayer)
VPS hosting providers
Linear Mode
