What Storage Is Best For Server Virtualization, Part II
In my <a href="http://www.informationweek.com/blog/main/archives/2010/08/what_storage_is.html">last entry</a> and the first part of this series we discussed some of the key capabilities to look for when selecting a server virtualization strategy, but as a friend of mine pointed out I never really declared one storage type the best. In this entry we will start to give you some steps to follow in making that selection.
In my last entry and the first part of this series we discussed some of the key capabilities to look for when selecting a server virtualization strategy, but as a friend of mine pointed out I never really declared one storage type the best. In this entry we will start to give you some steps to follow in making that selection.The reality is that there are so many variables in the selection process, many particular to your environment, that I'm not sure that anyone can tell you conclusively without knowing those details which solution is best. The truth is that almost any shared storage platform can be made to work for your server virtualization project. How much effort you put into making it work, how much time you have to spend keeping it working and how adaptable it is to change are really the key issues.
The first is to get as good a grasp as possible on the scope of your server virtualization project's storage needs. This is not just the amount of capacity required but also the number of physical servers that will be connected to that storage as well as how many virtual servers will those physical servers be supporting. From a virtual aspect I suggest doubling or even tripling what you think will be the most virtual machines that will be needed. Almost every virtualization project that I have been involved with ends up supporting far more virtual machines than it originally intended.
The number of hosts and the ensuing number of virtual machines has a direct impact on the next step; developing and understanding how much storage I/O you are going to need. The challenge with calculating storage I/O needs in a virtualized server environment is that the I/O demand is completely random. There are also two points of storage I/O contention. The storage bandwidth itself to connect the hosts to the storage and the I/O capability of the storage. The simplest solution, assuming you can afford it, is buying more network and storage I/O performance than you calculate you will need. For network I/O that may not be as far fetched as it sounds. For smaller organizations, often the performance of 1Gb iSCSI provides more than enough bandwidth. Up from 1GbE you don't have to be a big enterprise to justify as 10Gb Ethernet becomes increasingly affordable.
Getting I/O at the storage controller itself gets a little more challenging. The key to keeping storage simple in any environment but a server virtualization one in particular, is to limit the amount of discrete storage systems you have to manage. One system that can do it all is ideal, two is manageable, three or more can turn into a nightmare. I/O, or lack there of, is one of the issues that will drive you to an additional storage system quicker than others. To prevent this look for systems that affordably can give you more I/O performance than your environment will ever need or systems that can have I/O performance added to them without replacing the storage system.
In our next entry in this series we will look at more steps to take in narrowing down what storage is best for server virtualization in your environment.
Track us on Twitter: http://twitter.com/storageswiss
Subscribe to our RSS feed.
George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Find Storage Switzerland's disclosure statement here.
About the Author
You May Also Like