On Server Fault, I asked this question, trying to get some ideas about how to figure out what size of a server I need.

Clearly the question wasn't the right one.

However, I'm pretty sure people on this site buy (and size) servers frequently. How do I ask the question to get at that sort of checklist?

Pointers appreciated!


2 Answers 2


Server sizing is a deeply personal and technical thing, and generally requires more data than can be put into a ServerFault post. This is why they nearly always get closed. There are some things people know we need to know, and they're pretty good about including them:

  • Concurrent connection count
  • Total number of users
  • Some idea as to how much data will be used
  • Technology frameworks in use

Good, but not enough to size anything on. The one variable that is very, very hard to evaluate with no access to the system (which is what all of us answerers will be doing) is code performance. In the grand algebra of how much server do you need, how efficient your code is in handling each transaction has deep impact. Benchmarking is the only real way to get this data.

And for sizing questions involving off-the-shelf software where code-efficiency could reasonably be viewed as common knowledge to the right subset of people, the other variable, how your users use the system has significant impact as well.

These two items:

  • Code efficiency
  • User usage patterns

Are things we can't judge for you. Unless you have that, we can't help you size something beyond generic guidance. That may be what you're looking for, but if we keep going down that path we'll develop a stable of 'canonical answers' that we'll close-as-dupe questions to.

A better question to ask than, "how much server do I need," is, "How can I figure out how much server I need." This second question allows us to answer with techniques for figuring it out, rather than hard values we're wild-ass-guessing on. Specific technology frameworks have their own ways of determining performance, and you may learn quite a bit about benchmarking in the process.

  • Great answer! Thank-you! I'll see if I can formulate something more confined and answerable based on this... or perhaps several different such questions.
    – Peter K.
    Commented Jul 23, 2011 at 16:22
  • 1
    hmm, this seems like a classic "canonical answer" per blog.stackoverflow.com/2011/01/… and perhaps it would work as such on the main site? Commented Jul 24, 2011 at 9:22

As sysadmin1138 said, server sizing questions are large and difficult to answer, involving a great many variables which are hard to describe in a server fault post. I also find that there's a certain amount of "you can't even understand the answer" (as per joel's answer to this MSF question) in most questions of this type -- you start describing why the question can't be answered, and the questioner goes off the rails... and it just ends up not being worth it.

In your specific case, there was yet another factor at work -- you weren't just asking a question with insufficient information, you were asking for something that doesn't exist -- "industry standard server sizes". It's a nonsensical request, and I feel that anyone with IT experience would both (a) know such a thing doesn't exist, and (b) know that such a thing couldn't exist, because computer hardware and software is so variable and changes so fast that any "standard" would be obsolete before it could get a chance to take hold.

  • Thanks for taking the time to answer! The reason I was mis-guided about "standard server sizes" is that the company I used to work for (UTX), and when we used CSC as the IT infrastructure supplier. Whenever we needed a server, they basically had a standard list of servers to choose from. So I was (erroneously) thinking that this might be standard in the industry.
    – Peter K.
    Commented Jul 23, 2011 at 23:20
  • It might be instructive to meditate on the way that CSC described their various offerings on that list. I'll bet a large quantity of money that it wasn't along the lines of "capable of serving 1500 simultaneous users", it would have been more "this is the CSC Whizzo Enterprise+ grade 4 server" -- a term meaningless for anything more than saying to your salesweasel "please send me five of $THATONE". To properly select a server for your task, I'm guessing you still would have had to examine the specifications of the various configurations.
    – womble Mod
    Commented Jul 23, 2011 at 23:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .