Clustered Servers 1

[ice78991]

I am looking into setting up a network of 3 dedicated servers. One Server will run coldfusion, the other server SQL Server 2005 and the last server a CPU intensive application. I want to be able to communicate between these computers over a private network using IP addresses. Would these servers commonly be connnected via a router and is this setup known as a dedicated server cluster? How do you go about setting up such a system with a company?

 

[burtsbees]

They are commonly set up with a router and a switch. The only cluster that I know of is a cluster of the same type of server (let's say three) that all boot off of the same disk. The ones that are common that I see are Alpha ES45's (usually three) in an OpenVMS cluster. Do you need help with the entire thing, or have I explained enough? I will be glad to help.

Burt

 

[jimbopalmer]

if you use a different subnet, and leave the default gateway blank, you do not need a router at all

http://members.arstechnica.com/x/jimbopalmer/servers.jpg

In this picture, the users are all on 10.0.0.0 so substitute your actual user subnet. Then make a private subnet that is different. When the servers use the private IP addresses, they will not slow down the user network.

(this is not a cluster as I understand it, as the servers are not interchangeable, but you may speed up with this arrangement. I used this from 1993 to 1996, as ethernet would only go 10 meg, but 100 meg FDDI was too expensive to run to every user)


I tried to remain child-like, all I acheived was childish.

 

[LawnBoy]

I've got 23 servers and not a router anywhere near them. You only need a router to cross over onto a different network.

Clusters are setup in two ways (that I'm familiar with, there are probably other ways as well). All use shared disk space, one way or another...

There are high load clusters where you have two or more servers running the exact same software. All servers are handling requests at the same time, therefore you can handle more connections and requests simultaneously. Think Google.

There are also high availability clusters where (usually) 2 servers are loaded with the exact same software but only one of them is active. Should the active server fail, the secondary server will become active and perform the job.