Listed requirements vs real-world use

Alex Nelson3

I realize that has already been asked adnauseum, but wanted to bring it to light again incase anything has changed.

We're looking at setting up Bamboo, Crucilble, FishEye, Greenhopper, and Jira. System requirements are a good start, but not overly helpful in that they don't say anything about how many users are being accounted for with the requirements.

The fact is that we and many other companies need to submit requirements to those in organization responbile for provisioning servers and databases. It makes us look silly to ask for servers/DBs based on the listed system requirements and then find out that we need to have 16GB RAM instead of 4GB or 4 CPU cores instead of 1 after reaching X (mystery number) of concurrent users.

I've read the suggestion that we find this out on our as part of the trial process. This still isn't very helpful as it wastes our time and still makes us look silly if we need to go back and ask for beefier servers.

It would be extremely helpful if included with the requirements were guidelines as to how things will likely need to scale. e.g. 1 more CPU core per 10 users, etc.

Nic Brough -Adaptavist-

I'm afraid there simply isn't an answer.

You can't estimate the resource usage of any complex system like these because the answer depends on how you use them, how you configure them, your surrounding infrastructure and so-on

I completely understand why you are asking and have every sympathy, because I've been there many times myself. It's really hard and quite frustrating to have to sit there and say "I really don't know" in any talk about resources.

Unless you can say "we will have the same usage pattern as existing system that belongs to X", there's no useful answer. 1 CPU core per 10 users is too simple a formula, because it has only one factor and that's not enough. Even if it were a valid starting point, before it can tell you anything, you have to combine it with information on the nmber of issues those 10 users are going to filter on, across how many projects, with how many custom fields to search and/or display, and... and... and...

Random example - one place I worked had 20,000 users and were running Jira fine on Server X. My current main client has 6000 users and virtually identical hardware and are moving to more CPUs because they're struggling. It's just not a simple question.