TM1 Forum

Hi All,

Has anyone experianced the Drill to Transactions (to an MS SQL DB) to cause an increase in the Memory in Garbage?

We seem to have a spike in the RAM usage about 1 week out from Month End which is when users start hammering the server with queries for their Financials & Transactions.
From the Message Log we can see that there are a lot of these queries getting actioned over the last 2 days and have subsequently seen our RAM (from Task Manager) grow from 55GB to 92GB (i.e 100% usgage).

I still have the feeling this most likely relates to a "toxic" templates with poorly built stargate views that some users have to build within their Flash reports...

Just wanted to eliminate the not so obvious!

It doesn't surprise me that lots of transactional drill-through seems to increase garbage memory. It does surprise me that the garbage memory would continue to grow over time.

When TM1 performs a transactional drill-through the cross-tab data set it fetches back is external to the TM1 data model so if extra memory is needed then this would go straight into garbage, this part makes sense. For a really large query this could even be quite a lot of memory, say a few MB. But the whole point of garbage memory is that TM1 should internally recycle it, if this isn't happening then there's an issue with how TM1 is handling the reallocation of memory. By the sounds of it this is something you should really be investigating further.

Any pointers on how to go about investigating how TM1 is handling the reallocation of memory?

From the Perfromance Monitor last night we were seeing 112GB of RAM used & 62GB of this was in Garbage (on a server with 92GB of RAM).

You can use the windows performance monitor tool to collect memory usage metrics from TM1 over time. You can set it up to grab both the tm1 process virtual bytes (Windows measurement) and the tm1 specific pooled memory/garbage memory counters. Not sure if you can both collect the data and view it in real time, but the tool makes it easy to view a graph. You should see garbage memory shrink to zero (or close to it) before more memory is allocated from the OS. Also, by having the graph you should be able to see what time during the day memory spiked and you can try to correlate with tm1top/message log files to see what activity was going on around the same time.