Data center services with multiple compute paths?

[anandtechuser07]

Anyone here work in a data center or in network operations? I'm looking to set up a system that provides fault tolerance and scalability through replicated services.

For example, suppose some "service" can be run through a web server, an SQL server, and a compute node in serial fashion. I would then like to replicate that whole "service" through a different set of servers but with different characteristics, like though a web server and an SQL server only, where the SQL server runs some kind of stored procedure in lieu of a compute node.

Basically I want to provide redundancy and parallel execution through different data and compute paths. Anyone know of best practices, or have some URLs to related information?

 

[spidey07]

What you're looking for is content switching/load balancing. It's a device that accepts incoming connections and directs them to the appropriate services/servers. These servers could be anywhere in the world for redundancy sake. If this is a single data center you can just run a pair of content switches in a redundant scenario.

The decision to send connections to one set of services or another is very customizable as well. Cisco and F5 own this arena. The only real difficulty is truely understanding the application and how it works to make load balancing most efficient.

Go to Cisco for a look at ACE
F5 networks and BigIP

Also - if you want more information on load balancing/redundancy there's plenty of people that do this stuff everyday for a living in the networking forum.

 

[drew99]

what spidey said. Some data centers offer Load Balancing/CSS services to customer in lieu of you forking out the money for the equipment and configuration. That way you have little capital cost and bosses like that

 

[anandtechuser07]

Thanks for your help. I posted a similar question on the Networking forum.