2.4.18-5 server and client with < 2.4.18-5 read slowness

[seth vidal <skvidal@phy.duke.edu>]

[-- Attachment #1: Type: text/plain, Size: 1060 bytes --]

Hi all,
 I've been trying to sort out the problems with rhl 7.3 2.4.18-5.
I've got the mounts using rsize and wsize of 8192 and in async mount
(and async export on the server (for testing nfs speed, not disk speed))

What I've found is this:

if the server is 2.4.18-5 the read and write performance is fine with
either older kernels or 2.4.18-5 (on 100tx cards 9-10MB/s on read and
write w/multiple threads)

If the server is < 2.4.18-5 the performance for clients 2.4.18-5+ is
good for writes, lousy for reads, but the performance for < 2.4.18-5
clients is good for both reads and writes.

(2MB/s read performance and 9-10MB/s write performance)

Is this occurring b/c of the rpc congestion fixes that went into the
client for 2.4.18-5+ or is it related to the rpc_ping function (which
appears to be showing up A LOT in the tcpdumps of traffic b/t the two
machines)

Is it likely that the newer clients will just not be happy with older
servers b/c of the changes to the client code? Is there any way to work
around this?

-sv



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 232 bytes --]