NFS slow, how many nfsd should run ?

NFS slow, how many nfsd should run ?

Post by WONG SAI-KE » Tue, 19 Jun 2001 16:44:00



I had just configured a NFS server on eDesktop 2.4 and tried a NFS client
on Redhat 6.1 to mount it.  But I found that the speed was slow.  Slow
means for 2 cases:

1.  The first time to access the directory, there were several seconds
    of delay.
2.  During vi of a file, the response was slow

I checked with ps -ef on the NFS server and found:

root       632     1  0 17:09 ?        00:00:00 nfsd 4
root      1013   632  0 17:14 ?        00:00:00 nfsd 4
root      1014   632  0 17:14 ?        00:00:00 nfsd 4
root      1015   632  0 17:14 ?        00:00:00 nfsd 4

Is it the only reason is there're too few daemons are running ?
How can I increase it ?  How to calculate how much is the best
(opt for speed) ?

Thanks, rgds

SK

 
 
 

NFS slow, how many nfsd should run ?

Post by God_Of_Pai » Wed, 20 Jun 2001 06:59:07


This may not help you but... NFS connection speed will be greatly hampered
if the NFS serving computer does not know the name of the client making a
request. If you are not using a DNS on your LAN, if your IP address are
assigned and static, make sure that your NFS server has an entry in it's
/etc/hosts file for each client that wants to make a connection. My
/etc/hosts file is this.
192.168.0.2             Jason.Ept               Jason
192.168.0.5             sasquatch.EPT           sasquatch
192.168.0.80            StoVoKor.Ept            StoVoKor
192.168.0.86            OlympusMons.Ept         OlympusMons
192.168.0.89            UtopiaPlanitia.Ept      UtopiaPlanitia
192.168.0.200           Brother.lpd             Brother
192.168.0.250           KVort.Ept               KVort
192.168.0.254           Gateway.gwr             Gateway

If connection speed is not the problem but transfer speed it I would look
at hardware as the problem.


> I had just configured a NFS server on eDesktop 2.4 and tried a NFS client
> on Redhat 6.1 to mount it.  But I found that the speed was slow.  Slow
> means for 2 cases:

> 1.  The first time to access the directory, there were several seconds
>     of delay.
> 2.  During vi of a file, the response was slow

> I checked with ps -ef on the NFS server and found:

> root       632     1  0 17:09 ?        00:00:00 nfsd 4
> root      1013   632  0 17:14 ?        00:00:00 nfsd 4
> root      1014   632  0 17:14 ?        00:00:00 nfsd 4
> root      1015   632  0 17:14 ?        00:00:00 nfsd 4

> Is it the only reason is there're too few daemons are running ?
> How can I increase it ?  How to calculate how much is the best
> (opt for speed) ?

> Thanks, rgds

> SK


 
 
 

NFS slow, how many nfsd should run ?

Post by WONG SAI-KE » Wed, 20 Jun 2001 11:29:41


: This may not help you but... NFS connection speed will be greatly hampered
: if the NFS serving computer does not know the name of the client making a
: request. If you are not using a DNS on your LAN, if your IP address are
: assigned and static, make sure that your NFS server has an entry in it's
: /etc/hosts file for each client that wants to make a connection. My
: /etc/hosts file is this.
: 192.168.0.2             Jason.Ept               Jason

I have something similar.  But just the name may not be exactly the
one used in the client machine.  The reason was historical.  I think
that shouldn't be a problem, right ?

: If connection speed is not the problem but transfer speed it I would look
: at hardware as the problem.

I think the hardware is OK, its a P3 733MHz/128MB/18GB running the
OpenLinux 2.4 .  I found the newer distribution of Linux suffer from
similar problem of Mac OS and Windoze - big and slow.

SK

 
 
 

NFS slow, how many nfsd should run ?

Post by cameron.k.. » Fri, 22 Jun 2001 18:35:56


| I had just configured a NFS server on eDesktop 2.4 and tried a NFS client
| on Redhat 6.1 to mount it.  But I found that the speed was slow.  Slow
| means for 2 cases:
|
| 1.  The first time to access the directory, there were several seconds
|     of delay.
| 2.  During vi of a file, the response was slow

You probably need to be specifying the rsize=8192 wsize=8192 options in
/etc/fstab for all of your NFS mounted filesystems.

--

Praise Slackware, our baud and saviour!
--

 
 
 

1. Slow LINUX NFS preformance (nfsd)

I have recently added a LINUX machine to or network of SUN and HP
workstations. The new LINUX hardware is a Pentium P6 200Mhz, 2940UW
SCSI, 3c905 10/100, running Slackware 3.0 kernel Ver 2.0.0. I am
running the amd automounter to mount the remote file systems (about 30
disks on 12 workstations). The problem is on the LINUX machine I get
very
good performance copying files FROM a remote disk TO a local disk.
(about
1.2 seconds to copy a 7Mb file). But I get extremely poor performance
copying
FROM the local disk TO a remote disk, or when copying BETWEEN two remote
disks
from the Linux machine (about 30 seconds to copy a 7Mb file). I have
tried using normal mounts without amd, but the result is the same. I
have
also observed the same performace on a separate network from a P5
running
slackware ver 2.00.
Has anyone got any ideas as to what could be causing this problem? Is it
a problem with the nfsd setup?

Thanks for any advice on this problem.

--Simon


2. error in serial device

3. non-functional NFS (can't get nfsd to run)

4. Where is liby and libl (for lex and yacc) ?

5. HELP: Slow NFS client to server causing Very slow NFS install

6. boot question

7. Kernel NFS: equivalent to userland nfsd -r in kernel NFS

8. Welcome to comp.unix.shell [Frequent posting]

9. NFS really slow (too slow!)

10. PC-NFSD slow response mounting drives

11. really slow NFS (was: Faster NFS wanted: advice solicited)

12. nfs very slow, locking files across nfs ?

13. One NFS connection fine, one NFS connection slow