News for hackers. Stuff that matters.
Re:Java Turns 25 – Whats Next? [pdf] by deepakarora3 on Sunday September 20, 2020 @07:44PM (#24538401)
All OK except when you need the same level of performance for those 0.001% of requests when the GC kicks in and takes the response time outside acceptable limits. Due to this reason alone, my company is planning to move off a popular Java based API gateway and to a C++ envoy side car implemented service mesh. And I am wondering if this is really worth it.
Re:Java Turns 25 – Whats Next? [pdf] by throwaway189262 on Monday September 21, 2020 @12:38AM (#24539858)
Consider trying ZGC first. The main selling point is low worst case pause time. I've seen some tests where P99.9 pause time was less than 5ms, vs several hundred for older collectors.

In my limited tests I never saw a GC pause over 5ms. I was basically hammering a Spring Boot application with HTTP load tester.

Re:Java Turns 25 – Whats Next? [pdf] by pkolaczk on Monday September 21, 2020 @02:48AM (#24540540)
GC doesn't need to pause to have a negative effect on application latency.

While not-pausing is generally a huge improvement (after several decades of GC development), now what about thrashing of CPU caches?

Re:Java Turns 25 – Whats Next? [pdf] by sideeffffect on Thursday September 24, 2020 @08:28AM (#24577921)
Shenandoah might help you. See my other comment for more details:

https://news.ycombinator.com/item?id=24571054

Re:Java Turns 25 – Whats Next? [pdf] by zmmmmm on Sunday September 20, 2020 @10:07PM (#24539087)
They should definitely investigate the new low-latency gc options in the JVM.