As port of an attempt to reduce the amount of manual work need to administer our databases, I am considerring building some statistical routines to help predict database behaviour.

I have aprocess which, on a weekly basis, stores the size of various databse elements (for those who know ADABAS: number of records, number of extentes (DS/UI/NI), DS size etc.). I also have data about buffer usage etc.

What I want: * Predict the growth of databases (let's say I'd like to predcit for 3 months in advance) * Discover erratic behaviour (based on past experience) * Discover relationshpis between differnet databases (or files).

Anyone doing this kind of thing? If so, I'd be happy to hear your ideas. If someone can show me actual result/code I'd be verey glad.

Any practical solutions, pointers to methods (statistical or otherwise), is also welcome.

If/when the the project succeeds - I'll try to publish (possibly on the net). Hopefully the results will be good enough to be reused by others.

-- Ehud Lamm mslamm@pluto.mscc.huji.ac.il