With more and more companies facing the ramifications of data center failures, the need to create a solid redundancy plan is getting more attention. Creating redundancy of a component to ensure that backup is available is considered essential.
For a data center, redundant components can include servers, as well as fans, operating systems, hard disk drives, and other components of a network system.
As you examine your options for redundancy in data center capacity, you may be questioning how much is enough — especially since duplication can be considered excessive or unneeded in some cases.
Here are some factors to consider as you create redundancy in your data center operations:
Weigh the risks of failure. Before developing a redundancy plan, it’s important to examine the risks of failure and the costs associated. Only then can you get buy-in on investing the resources needed to develop redundancy. Examine everything from a “small” failure to a major, lengthy failure and its impact on operations.
Determine how much is too much. You can go overboard with creating a redundancy plan, as Brian Kirsch of Milwaukee Area Technical College points out. For example, it’s not necessary to “redundant everything” for a tier-one application, as one company initially set out to do. Once they discovered the expenses related to that plan, they scaled back to determine what was actually needed. As Kirsch explained, traditional hardware redundancies with additional network and power connections can be considered the baseline for most virtualized infrastructures. Otherwise, companies will run into under utilized resources.
Kirsch said to keep in mind that with a virtualized environment, the system you’re trying to protect is not a single system. Keep in mind that it may contain dozens of systems and applications. The more complex, the higher level of redundancy you should explore than you would for a traditional server.
Establish redundancy on different levels. Redundancy plans can be established for numerous levels — from components all the way up to regional, geographic and world levels. All companies should focus on redundancy plans that take into consideration component failure, assembly failure, room failure and building failure. On the most basic level, an N+1 approach is essential; it allows for having one more component than would normally be needed. On the next level — assembly failure — redundancy would be created for the failure of a complete server or a storage system. Storage mirroring and multiple network interface cards can provide resistance to failure. Room failure can be addressed by building two data centers within the same facility. The same concept could apply to building failures.