What settings should we use to optimise the use of the available memory?
We're currently using 10.22.
If you look at the sizing guide the requirement for an Enterprise class UCMDB Server is 64GB.
While there are recommended JRE settings for the probe there aren't any for the UCMDB Server.
The Support Matrix document suggests that only 32GB is needed.
In one document from Effectual the following advice is given:
For large capacity planning, it is recommended to change the Heap size to 8GB for 40 million CIs and relationships or more. If working with an Oracle database, 4GB is supported, but 8 GB is recommended.
In other documents I've seen 24GB mentioned.
What would HPE recomment for a 10.22 implementation on a server with 64GB installed?
What other settings should we be using?
Toward the end of the deployment guide there is a section called "Large Capacity Planning Overview". This section describes the number of CIs and Relationships expected and memory required to handle each number. With 64GB you should be able to handle 125 million CI's and relationships if you configure your system as follows. (taken from deployment guide 10.31)
Be sure that your database server can handle this load.
After setting these parameters and restarting UCMDB you should see it consuming much more of the system memory in the task manager of the system.
Hope this helps. Its my first post ever. I'm new to the community.
The 10.20 Deployment Guide gives the following.
So a maximum Java memory of 16GB.
So you see my confusion you install 64GB of memory (or should that 32 GB) but the maximum that UCMDB will use is 16GB.
Should we be using the 10.30 recommendations for 10.20?
Parameter ≤ 40 million > 40 million; ≤ 60 million Location wrapper.java.initmemory 2048 8192
wrapper.java.maxmemory 8192 16384 dal.object.condition.max.
result.size 50000000 50000000
table.high.threshold.oracle 6000000 6000000 dal.joinf.max.result.size 4000000 4000000
I would use the 10.20 recommendations because your running 10.22. Leveraging the additional memory and increasing the number managed CI's are benefits of upgrading to 10.3x. Following the portion of the deployment guide you posted will increase the memory available to UCMDB 10.2x, but you won't be able to use all of the memory on your server and be covered under the documented configuration until after you upgrade. I see your confusion with the recommended assignment of 64GB of memory to the UCMDB server and setting it to use a max of 16GB. I would just set it as the documentation recommends and stress test it to see how if works under those parameters. This may not be an issue at all if you have under 60 million CI's.