Parin, As every application workload is different I would suggest that you test G1 vs CMS on your specific application and installation.
In comparisons that I have done between tuned CMS and default G1 I have found the default G1 often beat out the best tuned CMS. I also found that tuning G1 often lead to poorer performance overall. When left to its defaults G1 auto tunes as the application evolves. As many IBDG/Cache applications go phases, like bulk load to runtime, tuning the GC for any one phase causes the other to perform poorly. With default G1 settings performance as your application changes phase is not immediately great but improves quickly. If your application alternates between phases then so will G1s performance but overall you will likely see better performance than a highly tuned CMS and with a lot less fussing. -Jake On Wed, Mar 29, 2017 at 8:23 PM Parin dazz <[email protected]> wrote: > Hello Team, > > Any update on this? > > Regards, > Parin > > > On Tue, Mar 28, 2017 at 9:12 AM, Parin dazz <[email protected]> > wrote: > > Hi Team, > > I want to know about geode's support for java 8? Do we have any reference > for jvm tuning with java 8? > > I found few ones related to CMS & MarkNSweep GC. Is it a good idea to use > G1 GC from Java 8 onwards? What care we must take while using G1 GC? Any > standard recommendations for G1 GC parameters? > > I understand that old GC tuning config was suggested to achieve low > latency and high through put. I would appreciate if you can share your > analysis for G1 GC as well. > > Thanks & Regards, > Parin Dazz > > >
