Abhishek,
The fact that you are running on AWS is enormously significant. It's quite possible that your problem has more to do with your provisioning, or your OS configuration, than anything else. If "mechanical sympathy" implies that your application and the underlying hardware have to work together in harmony to see optimal performance, then *that * *harmony is a holistic thing that has to encompass application, JVM, OS kernel, virtualization layer,* *host OS, hardware BIOS, firmware etc. * When we explore issues in this forum we do ourselves a disservice if we only focus on, for example, the characteristics of the garbage collector and pretend that it exists in a vacuum. There are a few default settings in popular Linux distributions that are unhelpful to performance sensitive Java applications. As a starting point I would suggest that you might use *pidstat* to try and identify what is occurring a the OS level when these minor GCs happen. A command line similar to pidstat -d -h -r -u -v -w -p <PID> run every second both with and without the -t option would be useful. On Friday, December 22, 2017 at 1:29:01 AM UTC-5, Abhishek Agarwal wrote: > > Thank you for the pointers. We have gc logs but didn't help much. > > Few corrections: > Old gen size is ~ 20 Gigs > Young gen size is ~ 1.5 Gigs > > The skew between young gen and old gen size has been fixed. Here is the > snippet of gc logs when time increased to ~ 200 seconds. By the way the > service is running on AWS. > > > > -- > Regards, > Abhishek Agarwal > > -- You received this message because you are subscribed to the Google Groups "mechanical-sympathy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
