Starting with the release of Strategy ONE (March 2024), dossiers are also known as dashboards.
In Strategy 2021 Update 4.1, execute a one-million-rows report and notice workstation shows 1 GB memory consumption while Linux OS only shows around 0.2 GB memory increment.
The initial memory of the Linux OS is about 9.7G as shown below:
At the time of completion of one-million-rows report, Linux OS memory increased to 9.9 GB. An increment of 0.2 GB demonstrated by the Linux ‘top’ utility as shown below:
The memory usage from workstation at the completion of one- million-rows reports is 1 GB as shown below:
This is working as designed with Strategy 2021.
It's not necessary that the increased memory at the server level (read from the top command, in the RES field) is more than all the memory consumed by the running jobs (from the job monitoring window in Workstation, which refer to the current memory usage for a job).
The reason is some heap memory may not be released immediately after the usage, it will be cached and reused later to increase the efficiency/performance.
The performance counter "Total in-use memory (MB) in server process heap" and "Total cached memory (MB) in the server process heap" may reveal the heaped memory part.
Refer to following Knowledge Base Articles for how to enable the counters:
KB19155: Recommended performance logging counters to enable when troubleshooting memory or performance issues in MicroStrategy Intelligence Server Universal in UNIX environments
KB19154: Recommended performance logging counters to enable when troubleshooting memory or performance issues in MicroStrategy Intelligence Server in Windows environments
MSTRSvr is increased from 3.2 G to 3.9 G while workstation shows 0.8 G when executing a dossier.
From Performance counter, it shows ‘Total in-use memory (MB) in server process heap’ increases from 3113M to 3336M, and ‘Total cached memory (MB) in the server process heap’ increases from 205M to 508M.
There is 205MB cached heap before the job execution, this part is already counted in the server used memory and can be reused by the job. So, the math is 0.7GB (increased memory at the server level) + 0.2GB (cached heap) > 0.8GB (used memory for the job), which is reasonable.
It is an expected behavior and there is no action needed in Strategy side.
