Re: -mm merge plans

From: Ryan M.
Date: Thu Mar 23 2006 - 02:56:34 EST


Hello,

Andrew Morton wrote:
> A look at the -mm lineup for 2.6.17:
>

>
> mm-implement-swap-prefetching.patch
> mm-implement-swap-prefetching-fix.patch
> mm-implement-swap-prefetching-tweaks.patch
>
> Still don't have a compelling argument for this, IMO.

I hate to make a comparison based on the little information there is, but Windows Vista will have something like prefetch, albeit, exponentially more intrusive (read MS' explanation on their website). However, when I see that such technology is being embraced by the competitor to help improve the desktop (and it follows, the server space) and I see this better, non-invasive solution nearly rejected, I can't help but feel rather disappointed.

To prefetch applications from swap to physical memory when there is little activity seems so obvious that I can't believe it hasn't been implemented before.

I play World of Warcraft which saps away 1gb of physical memory in a heart-beat. During that period of time I run gobs of other networking applications as any desktop users might. They sometimes swap. I leave World of Warcraft and I can see them prefetched back to physical memory while not being removed from swap. To do something useful, particularly when it helps interactivity, when idle is a very smart use of resources. It is even smarter to save future swapping of those same applications because they're likely not to be removed from swap unless swap space is becoming limited. It is practically free and for long running desktop systems that seems a necessity.

I don't have quantitative evidence, but I think the objective of swap prefetch speaks volumes itself. I can imagine it being useful in server-space, because having had anything swapped out in the past and it not having to be swapped again after prefetched could seriously help reduce disk accesses -- extremely important during heavy i/o loads.

I hope others join me in explaining the very usefulness of swap prefetch -- it'd be great for someone with the time and abilities to provide quantitative reasons for this existing in mainline.

best,
Ryan M.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/