2016-04-07 13:27:26

by Martin Houry

[permalink] [raw]
Subject: Question : Compress files to send with NFS4.1

Is NFS4.1 capable to compress files before sending them on the network
and decompress them on the destination to speed up the file transfert?

Kind regards,
Martin


2016-04-07 14:06:12

by Cedric Blancher

[permalink] [raw]
Subject: Re: Question : Compress files to send with NFS4.1

On 7 April 2016 at 15:27, Martin Houry <[email protected]> wrote:
> Is NFS4.1 capable to compress files before sending them on the network
> and decompress them on the destination to speed up the file transfert?

You can get that with IP tunnel compression. Trouble is, NFS
(v2/v3/v4; and younger protocol designers are even worse by telling
you 'dunno care about high latency [ > 4 sec roundtrip ] any more'
[yeah? wanna pester? old RFS and AFS did a better job...]) are highly
sensitive to latency, and like ssh the high latency causes more
trouble than its worth.

Ced

>
> Kind regards,
> Martin
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html



--
Cedric Blancher <[email protected]>
Institute Pasteur

2016-04-07 18:16:45

by J. Bruce Fields

[permalink] [raw]
Subject: Re: Question : Compress files to send with NFS4.1

On Thu, Apr 07, 2016 at 04:06:11PM +0200, Cedric Blancher wrote:
> On 7 April 2016 at 15:27, Martin Houry <[email protected]> wrote:
> > Is NFS4.1 capable to compress files before sending them on the network
> > and decompress them on the destination to speed up the file transfert?
>
> You can get that with IP tunnel compression. Trouble is, NFS
> (v2/v3/v4; and younger protocol designers are even worse by telling
> you 'dunno care about high latency [ > 4 sec roundtrip ] any more'
> [yeah? wanna pester? old RFS and AFS did a better job...]) are highly
> sensitive to latency, and like ssh the high latency causes more
> trouble than its worth.

The only feature the protocol has that's a little bit like compression
is READ_PLUS (in 4.2), which can tell the client that a range of the
file is a hole instead of returning all the zeroes. You could think of
that as a specialized kind of compression for sparse files. Even there
we've been having some trouble ensuring there aren't performance
regressions.

--b.