[cgl_discussion] vm strict overcommit question
Zheng, Jeff
jeff.zheng at intel.com
Mon Aug 11 18:24:14 PDT 2003
To config if Heuristic or Strict overcommit, just set value of /proc/sys/vm/overcommit_memory:
0: Heuristic overcommit. Default value.
1: No overcommit handle, memory allocation always success
2: Strict overcommit. swap+part of physical memory (configed by /proc/sys/vm/overcommit_ratio).
The issue is if we want swapless strict overcommit, we have to do "swapoff -a".
Thanks
Jeff Jeff.Zheng at intel.com
BTW, I speak for myself, not for Intel Corp.
> -----Original Message-----
> From: Fleischer, Julie N
> Sent: 2003年8月12日 7:00
> To: Zheng, Jeff; cgl_discussion at osdl.org
> Subject: RE: [cgl_discussion] vm strict overcommit question
>
>
> > From: Zheng, Jeff
> > From CGL2.0 requirement, AVF6.1, vm strict overcommit, it
> > defines two methods for control of virtual memory:
> > + Heuristic overcommit
> > + swapless strict overcommit
> >
> > The POC reference web site has patchs for both linux-2.4.18
> > and linux-2.5.26. The patchs define following features:
> > + Heuristic overcommit handling.
> > + No overcommit handling.
> > + Swapless strict overcommit
> > + Strict overcommit
> > + Paranoid overcommit
> >
> > But from linux-2.5.28 to newest 2.5 kernel, only three
> > features survives:
> > + Heuristic overcommit handling.
> > + No overcommit handling.
> > + Strict overcommit. It is configurable. By applying
> > parameter to Strict overcommit, it can implement same
> > functionality as Swapless strict overcommit and Paranoid
> overcommit.
> >
> > My question is: Does AVF6.1, vm strict overcommit, means
> > "Configurable strict overcommit" + "Heuristic overcommit"?
>
> Didn't seem like this has been answered yet (although I may
> have missed it after the F2F).... My read is that since the
> requirement is toggleable yes that both swapless strict
> overcommit and heuristic over-commit just need to be turned
> on by default, but it's okay if they are configured out later.
>
> I'll add this data to the POC tracking sheet, though.
>
> - Julie
>
> **These views are not necessarily those of my employer.**
>
>
> _______________________________________________
> cgl_discussion mailing list
> cgl_discussion at lists.osdl.org
> http://lists.osdl.org/mailman/listinfo/cgl_discussion
>
More information about the cgl_discussion
mailing list