[erlang-questions] hybrid heap question

Ulf Wiger ulf@REDACTED
Wed Feb 28 21:52:26 CET 2007


Den 2007-02-28 17:25:00 skrev Kenneth Lundin <kenneth.lundin@REDACTED>:

> Because the hybrid heap involves a heap which is sharedbetween all  
> processes it is harder to do a GC without
> impact on all processes during that time. Incremental GC
> is therefore more of a necessity for Hybrid heap than for
> the separate heap solution at least in the SMP emulator.
> Implementing the incremental GC for the SMP emulator will
> also be harder than for the non SMP emulator I think.

Given the obvious advantages of SMP Erlang, it's difficult to see
any kind of future for hybrid heap if it doesn't work together
with SMP. This is why I expressed my hope that they will eventually
work together.

We have run some internal benchmarks on "real"(*) applications with
SMP Erlang, and I think I can say that they have been extremely
engouraging. We've run several million calls without problems
(naturally, without recompiling anything either), and the
speedups have exceeded our expectations.

A big Thank You to the OTP team. It's nice to see that you can
surprise us even after 10 years of close cooperation.

(*) Real, as in lots of I/O, linked-in drivers, etc.
A few thousand processes in the system at any given time -
not terribly much, but enough to keep the scheduler threads
on a dual-core busy.

Does anyone else have some experience with SMP in OTP R11B-3?

BR,
Ulf W
-- 
Ulf Wiger



More information about the erlang-questions mailing list