In depth detail about the issue we have - hr system
Why we must do this testing
- do we expect high volume
- do we need to spend the money on a clustered system
- backup and fall back offered by vms
Numbers we are expecting
The approach
- Lets ask the supplier if they've done any testing themselves (we'll know a rough benchmark then of what we should expect)
- Identify use of the service/system
- Identify other ways of controlling user access (e.g. send out system updates to small batches of users instead of full amount)
- what's our normal scenario (how many users are we expecting in normal way
- what's our worst case scenario (max number of users hitting the system in X minutes)
- work our way up from normal to what we find out to be the worse case scanario, then keep going until it all explodes
- what do we need to really do the test (plenty of machines, which aren't used for anything else at the time - as our test servers)
- test network to remove network latency as an issue (in our dreams - but also, it's not quite real world testing then either)
- When do we do it?
- Who are we going to affect?
- What are we going to affect? (Understand network acrhictecture and any systems sitting in between - are we going to break something else? Tracert can be useful here)
- What do we monitor (Memory, CPU, etc.)
- How do we monitor those (Linux / Windows perfmon)
- Understand your baseline readings, so we know what will be outside of our normal scope.
- Look for patterns in monitor readings (do we have memory leaks)