Re: NFS client performance 1.5 orders of magnitude too slow? (fwd)

Godmar Back (gback@cs.utah.edu)
Thu, 11 Mar 1999 13:02:32 -0700 (MST)


>
> > allow me to override the NIS settings. He's said I should lobby for a change
> > of the Linux kernel's default from 4 to 8k instead.
>
> The default of 4K is right especially for small machines. NFS lacks a
> size negotiation phase.

Of course, a negotiation phase would be best.

Small as in small client or small as in small server?

Okay, can I lobby for leaving the default at 4k, but making it a run-time
configurable option (which I can change via /proc, like
/proc/sys/sunrpc/nfs_debug) or via sysconfig or whatever mechanism Linux
provides to change such settings? By doing so, you could leave the
default you like in place and I could set a better default that reflects
my local circumstances where 8k appears to be always better.
By doing so, I wouldn't have to worry about amd and NIS at all.
Erez could trust that a local sysadmin would set a reasonable wsize
default.

If you guys like this idea, I would volunteer to do it myself (this
change should only be a matter of changing a few defs and adding a
few create_proc_entry calls, right?)

- Godmar

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