Re: [PATCH 0/3] readahead drop behind and size adjustment

[Fengguang Wu]

On Sat, Jul 21, 2007 at 10:44:28PM -0400, Dave Jones wrote:
> On Sun, Jul 22, 2007 at 10:39:23AM +0800, Fengguang Wu wrote:
>  
>  > It makes sense to raise it beyond 128K.  1M default readahead
>  > absolutely makes sense for sequential workloads.  For the desktop,
>  > this increases boot speed and readahead misses, both due to more
>  > aggressive mmap read-around. Most users will be glad to feel the
>  > speedup, and happily ignore the readahead misses, which may be
>  > "invisible" in case of large memory.
>  > 
>  > In theory, the distributions can do the same tuning.  So we have an
>  > interesting question for Dave:
>  >         Does fedora desktop raise the default readahead size? Why or
>  >         why not?  It goes so far to do userland readahead ;)
>  
> Fedora takes whatever defaults for readahead the kernel.org kernel has.
> The only reasoning being if anyone reported VM bugs, we'd be able
> to say to interested upstream developers "we're running the stock VM".
> without having to get the user to try and reproduce on unpatched
> kernels.

Thank you.  Now I'm more confident that the kernel should have a
reasonable default readahead size. The current one is 5+ years old and
should be updated now.

>  > - drop behind
>  > 
>  > Sorry, I still doubt it will benefit all/most workloads. Leave it off
>  > by default, and leave the enabling decision to Dave? I do hope that
>  > it help general desktops.
> 
> It's not a subject that I'm intimatly familiar with, and when it
> comes to decisions like this, I tend to just take whatever the
> upstream defaults are.

- It will avoid large-file-reads-thrashing-my-desktop problem,
  so most desktop users should like it. But sure there will be counter
  cases when a user want to keep the data cached.
- File servers may hurt from it. Imagine a mp3/png file server. The
  files are large enough to trigger drop-behind, but small (and hot)
  enough to be cached. Also when a new fedora DVD iso is released, it
  may be cached for some days. These are only the obvious cases.

So I opt for it being made tunable, safe, and turned off by default.

-
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/