Should I “Enable Memory Analysis” on a Web Intelligence Processing Server?
It was always best practice to leave the feature ‘Enable Memory Analysis‘ turned on, as this enabled a number of mechanisms to kick in and reduce memory consumption. I described some of them in a reply to an
However, with Web Intelligence now sitting on a 64 bit architecture the product no longer has a 2 GB limit of memory to run within and so these mechanisms are likely to be unnecessary. This assumes you have enough RAM on your server!
Our development team have been busy testing the use of the product with this feature disabled. Rigorous testing identified a defect in a third party ‘dll’ on the Windows Platform. Non-Windows was ok. With that issue now fixed in BI 4.1 Support Pack 7 it’s now time to switch our Best Practice recommendation.
From SAP BusinessObjects Business Intelligence 4.1 Support Pack 7 onward and BI 4.2 onward, we are now saying de-select ‘Enable Memory Analysis’
If you are using an earlier version, on the Windows Platform, then you should leave this feature enabled to prevent a possible crash of the Web Intelligence Processing Server.
What’s the benefit? Quite simply, performance. You should experience an improved performance as the product will not need to garbage collect as often, this sometimes ‘freezes’ the product.
Enjoy to the new power of Web Intelligence, but please just make sure you have enough RAM and monitor the memory usage of your server. If your server runs of RAM it will start to use swap and that will slow things down.
If you have limited RAM and you find your server is using swap, you may need to keep this enabled. Feel free to increase the Lower/Upper and Maximum Thresholds, but my advice would be to keep the ‘ratios’ between these values the same. Don’t set them all to the same value in other words. You’ll want to allow the memory recovery mechanisms to kick in one at a time.
Feel free to comment to this blog with your experiences of disabling this feature and provide SAP and other customers feedback.
Regards, Matthew (Twitter: @MattShaw_on_BI)