I am using V8 with Node, and profiling is showing 30% time in GC running 
with these flags:

--nouse_idle_notification --expose_gc --max_old_space_size=8192

My process is calling gc manually every 2 minutes, and each call takes 2.5 
seconds. This should be about 2%.

If I run trace-gc there's just too much info to make sense. Anyway to 
aggregate this in the profile output? Or to breakdown GC into components in 
the profile output rather?

The server has 32GB of RAM and the process itself is only using about 2GB 
of RAM and it's no problem for it to use much more.

V8's GC seems to be too aggressive if it's spending 30% of CPU on GC?

Would max-new-space-size help? I have tried to look for more info on what 
this does but have not found anything.

Ideally, I'd like GC to just do nothing at all, and then every hour or so 
take a 5 second hit.

RAM can grow as much as necessary.

Would anyone be able to help?

-- 
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users

Reply via email to