How to translate cache to RAM utilization

[#1]

I am trying to understand high RAM usage on our system. Base application utilization plus an adjustment for concurrent users puts me at 20GB, but we consistently see RAM hover near 80GB (low working set). The difference has got to be with the caching that is going on, but I would like to understand how cache translates to RAM utilization and if that can be quantified.

On an average month there are 100 users making 3,000 unique selections, and the applications contain 1,500 unique expressions. What impact do these have on RAM? Are there any formulas to determine ultimate utilization based on selections and cached expressions? I've read 'The QlikView Cache' by HIC but it doesn't call out quantifying RAM per cache.

Any insight would be helpful. Thanks!

[Anonymous]

Kris - I don't know of an explicit formula that will account for selections and expressions and yield a RAM utilization...however my understanding is that the server will continue to cache all new expressions and selection states up to the low working limit you set (in this case 80GB).  This will often be much higher than the base app size plus concurrent usage load.  Lots of selections and lots of distinct expressions adds up quick = the situation you describe.  Once the working set limit is reached, the server cycles older items out of the cache and new selections and/or calculations are loaded into it.  I think the short answer is your high RAM usage is not abnormal or a problem. The server is efficiently using as much RAM as you are allowing it to.

Kyle