Perhaps you could achieve a lot with VM parameters according to your load?
> On Feb 4, 2016, at 1:35 PM, jim_b_o <ja...@inaseq.com> wrote: > > The GC pause times are a very important metric for our app and I'm sure for > many others. Our system is latency critical and delays cost money. GC > pauses are totally concrete. Events are being processed all the time (2 > main types: one type handled in 60-70 microseconds, other type in 2-3 > milliseconds) and throwing a GC pause into the middle of that processing > used to add another 10-12 milliseconds but now adds 16-20 milliseconds. > > Fortunately we are not burning memory too quickly so the GCs aren't > frequent. But load is sufficient that most times one occurs it will be > impacting one or more events. The impact on average latency is negligible > but the financial cost of additional latency on select events is not. > > > > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/Artemis-use-of-PhantomReference-impacting-GC-performance-tp4706961p4706978.html > Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.
