This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
1. - Modern data center design 101: Capacity planning: Read more in this section
- Capacity planning for virtual environments
- Let there be chaos in strategic planning
- Predicting the unpredictable in capacity planning
- Thinking redundantly for better capacity planning
- Strategic capacity planning for private clouds
Explore other sections in this guide:
- 2. - Ideas for data center facilities planning
- 3. - Choosing potential hardware systems: Modular and converged options
- 4. - Meet the Experts
Is there a rule of thumb for planning for server capacity? For example, if I know that I need five "beefy" servers to handle my in-house needs, should I buy an extra one or two -- or 20% more CPU/memory than needed -- for server redundancy? Assume that we don't expect to grow in terms of employees or clients, or that we have already handled that. Also, assume that we have all the spare parts we need -- this would just be to have on hand for additional capacity.
This depends entirely on your situation, as there is no all-encompassing rule of thumb when it comes to capacity planning. Also, you never mentioned how much if any of your server infrastructure will be virtualized. That's a huge variable when attempting to plan for server capacity.
Given the assumptions you've outlined, I would focus the entire capacity planning effort on the concept of server redundancy. This means a determination must be made with regard to what portions of your organization's infrastructure are vital and what portions are simply nice to have. All things being equal, I would allocate 25% more CPU resources than what your organization currently needs toward your most vital infrastructure. Then I would examine what has been deemed a "nice to have," and I would allocate approximately 15% of extra CPU resources toward this portion of your organization.
Another item you have to think about with regard to redundancy is storage. Again, all things being equal, I would calculate the total amount of hard drive space within your network and allocate some sort of network resource(s) with an amount of storage space equal to or greater than the total you previously calculated.