Hello,

I have a machine setup as a fileserver running slackware 10 and kernel 2.4.31. Several of the directories on that server, such as my music collection and a directory of kennel club information, are mounted to points in my home directory on my slackware 11.0 desktop machine (kernel 2.6.18) using nfs.

Over the last several months while ripping CDs using grip, I have noticed that although the actual ripping part has gone pretty quickly, the encoding, which requires reading the ripped WAV file off the server and writing back an MP3 file, takes a long time. (Ripping and encoding a CD recently took over eight hours. Subjectively, it also seems that the more CDs I rip, the longer each subsequent CD takes, but I can't prove that. Could just be a sign of my growing frustration.) This last Friday, I sent off an email with a 3.9MB attachment that was stored in one of these nfs-mounted directories. It took over 25 minutes for the file to transfer and the email to be sent.

So, I tested the transfer times, and here is the command I used and the results:

bash-3.2$ time dd if=/dev/zero of=HSV/transfertest bs=16k count=8k && time dd if=HSV/transfertest of=HSV_transfertest bs=16k
8192+0 records in
8192+0 records out
134217728 bytes (134 MB) copied, 1908.7 seconds, 70.3 kB/s

real 31m49.268s
user 0m0.004s
sys 0m0.392s
8192+0 records in
8192+0 records out
134217728 bytes (134 MB) copied, 23659.7 seconds, 5.7 kB/s

real 394m20.464s
user 0m0.016s
sys 0m1.264s

(HSV is my home directory mount point for depot:/srvr/hsv. I was in my home directory at the time, so HSV_transfertest was written to a local, non-nfs mounted directory.)

The network connection is running at 100Mb/s, as shown by `ethtool eth0` run on both machines:

On the desktop machine:
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Advertised auto-negotiation: Yes
Speed: 100Mb/s
Duplex: Full
Port: MII
PHYAD: 1
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: pumbg
Wake-on: d
Current message level: 0x00000001 (1)
Link detected: yes

On the server:
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Advertised auto-negotiation: No
Speed: 100Mb/s
Duplex: Full
Port: MII
PHYAD: 32
Transceiver: internal
Auto-negotiation: off
Supports Wake-on: pumbg
Wake-on: p
Current message level: 0x00000007 (7)
Link detected: yes

I ran tcpdump during the most recent test, and the output looks like this during the transfer from the fileserver (depot) to the desktop (trooper):

19:09:24.989946 IP depot.nuthole.de.nfs > trooper.nuthole.de.413269765: reply ok 1472 read
19:09:24.990003 IP trooper.nuthole.de.430046981 > depot.nuthole.de.nfs: 124 read [|nfs]
19:09:24.990044 IP trooper.nuthole.de.295829253 > depot.nuthole.de.nfs: 124 read [|nfs]
19:09:24.990430 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.990600 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.990738 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.990866 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.990995 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.991124 IP depot.nuthole.de.nfs > trooper.nuthole.de.295829253: reply ok 1472 read
19:09:24.991165 IP trooper.nuthole.de.362938117 > depot.nuthole.de.nfs: 124 read [|nfs]
19:09:24.991200 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.991208 IP trooper.nuthole.de.446824197 > depot.nuthole.de.nfs: 124 read [|nfs]
19:09:24.991329 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.991992 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992146 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992276 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992405 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992534 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992661 IP depot.nuthole.de.nfs > trooper.nuthole.de.362938117: reply ok 1472 read
19:09:24.992721 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992773 IP trooper.nuthole.de.463601413 > depot.nuthole.de.nfs: 124 read [|nfs]
19:09:24.992866 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.992994 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.993829 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.993994 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.994123 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.994252 IP depot.nuthole.de > trooper.nuthole.de: udp
19:09:24.994380 IP depot.nuthole.de > trooper.nuthole.de: udp

There is also a lot of ARP requesting going on, or at least it seems so to me:

19:08:40.251161 arp who-has trooper.nuthole.de tell depot.nuthole.de
19:08:40.251180 arp reply trooper.nuthole.de is-at 00:04:61:42:94:13 (oui Unknown)
19:09:27.188887 arp who-has depot.nuthole.de tell trooper.nuthole.de
19:09:27.189002 arp reply depot.nuthole.de is-at depot
19:09:41.279370 arp who-has trooper.nuthole.de tell depot.nuthole.de
19:09:41.279390 arp reply trooper.nuthole.de is-at 00:04:61:42:94:13 (oui Unknown)
19:10:51.628302 arp who-has depot.nuthole.de tell trooper.nuthole.de
19:10:51.628410 arp reply depot.nuthole.de is-at depot
19:13:31.619194 arp who-has depot.nuthole.de tell trooper.nuthole.de
19:13:31.619302 arp reply depot.nuthole.de is-at depot

So, can anyone offer me a clue where I can start looking for the problem here?

What are your mount settings on the client? What are your export settings on the server?

The mount settings on the client as show by mount are:

depot:/srvr/music on /home/tschulze/music type nfs (rw,sync,rsize=16384,wsize=16384,intr,addr=192.168.1.13,addr=192.168.1.13)
depot:/srvr/hsv on /home/tschulze/HSV type nfs (rw,addr=192.168.1.13,addr=192.168.1.13)

The export settings on the server are:

/srvr/music 192.168.1.0/255.255.255.0(rw,no_root_squash,sync,subtree_check)
/srvr/hsv 192.168.1.0/255.255.255.0(rw,no_root_squash,sync,subtree_check)

To see if the rsize and wsize settings have any effect, I am running the same command again, except I am using the directory mounted to depot:/srvr/music. I'll let you know the result.

rsize and wsize seem to help transfers to the server:

8192+0 records in
8192+0 records out
134217728 bytes (134 MB) copied, 549.925 seconds, 244 kB/s

real 9m10.259s
user 0m0.060s
sys 0m2.324s
8192+0 records in
8192+0 records out
134217728 bytes (134 MB) copied, 23155.5 seconds, 5.8 kB/s

real 385m55.944s
user 0m0.012s
sys 0m1.728s

But transfers from the server to the desktop are about the same. So, could it be a problem with a setting on the desktop machine?