We have been asked to consider changing the GC method used for AppMon to use Gencon. It appears it is using G1GC. It is also being recommended that if we do change to Gencon we have the minimum heap set to at least half of the maximum heap size as part of the transition to Gencon. Are there any recommendations or reasons for or against changing the GC method used? Is there any documentation to assist with making a change to the GC method?
Yes, AppMon Server is where the change was recommended.
Our middleware engineering team has made the recommendation based on the success they have had with moving to gencon for other products in the environment.
Are there any particular details as to why it would be advised against that i could provide?
I don't work on the Memory optimization team, so I can't answer the question regarding GC selection. However I would challenge anyone who claims that a given GC algorithm that works good on App1 is therefore a good algorithm for App2. Every app is different. I know that there's a fair bit of effort put into optimizing our memory management as what we're doing is memory intensive. I would turn the question around and ask what is proposed to be improved by a different GC algorithm. That's the question that the Eng team will ask.
GenCon is an algorithm that is optimized for environments which have lots of short lived (Eden) objects, but that's not the memory pattern of the AppMon Server.