I would like to do a bit of benchmarking to understand better our performance and look into possible infrastructure improvements.
My current issue is that we have large databases (typically 50Gb) that takes a lot of time to load into RAM when used the first time. Usually this takes between 20min and 40min.
So a few questions:
- is that normal ? what kind of performance do you have on your own databases / servers ?
- did you implement specific infrastructure upgrades to improve your performances ? Which ones ?
- are you using a virtual infrastructure ? If yes, are you using memory extensions such as VMWare "overcommitment" function ?
Besides in terms of usability, very often we start loading a DB into memory by selecting it by mistake in the DB list. If it's a small one, no issue, but if it's a big one we're stuck for 20min... could it be possible to implement a check like "This DB is not in memory, do you want to load it ?" ?
Thanks for your input,