Fragments

Distributed Data Centers

August 28th, 2008 | by Mitja Robas |

Are you building a backup data center (BDC) and experiencing headaches due to the related expenses? Or perhaps you already have a BDC and want to reduce the operational expenses?

One way to cut expenses is to look at the second data center as the active data center, and utilize its resources by deploying a distributed data center. The benefits are clear:

• reduced expenses per resource utilization
• load distribution (and thus scalable performance) among data center resources
• high availability of applications and data access for any user
• improved client response time due to client proximity, etc.

What issues have to be addressed to deploy a distributed data center?
For application delivery, the servers of the data center are the cornerstone. If you want to use the servers in both data centers, you must deploy a proper selection method. Typically, this is achievable by using global server load balancing (GSLB) to distribute client traffic to servers.

Apart from strictly network-oriented tasks (routing, switching, etc.), the distribution of client traffic is accomplished with proper site selection by engaging the following:

• HTTP redirect by leveraging the HTTP redirect function
• DNS-based site selection with »site selector«
• route health injection (RHI) and L3 routing

The data center architect also should not forget about server/application health monitoring, typically a task of application and content engines such as Cisco ACE or CSS.

Distributed data centers do have some disadvantages, such as the need to mirror data not only from primary to secondary DC but also vice versa. And don’t forget session persistence, which has to be addressed properly. However, distributed data centers offer enough advantages to outweigh these minor costs.

See also ACEDES and ACESM courses for more information on Data Centers.