Poor PE4600 NFS performance

Rick Gaudette Richard.Gaudette at Colorado.EDU
Tue Aug 6 00:53:01 CDT 2002


Hi Aly,


> > 
> > It is the onboard broadcom chip.
> > 
> 
> 	I am going to venture a personal opinion, I think that Broadcom cards aren't the 
> best, Intel is better in performance, I wonder if that has anything to do with 
> it. I guess its worth an expensive experiment to try another card maybe a Intel 
> or 3Com gig card ??? :) :) :)

I think the 3C996BT is a broadcom 5701 chip.  That seems to work fine
with the 2.4.18 kernel on another machine.

> 
> > 
> > 
> > Would this still apply if there are typically only one or two users
> > reading at any given time?  Our typical use model is that one or two
> > users might be trying to read a 1-2 GB data file at any given time. 
> > That is the type of use model we wish to get the best performance for.
> 
> 	I don't think this would do much in terms of 1-2 users at a go, unless you are 
> planning on servicing a Computer Science network with 120 Unix workstations and 
> whole bunch of students. I would consider maybe putting in another gig nic if 
> you are reading this amount at a go. Later maybe ??
> 
> > 
> > Anyone have any recommendations here?  We have tried a bunch but the
> > default seems to work best.  The man pages say the default is 8K but
> > when we specify 8K we get degraded performance.  The defaults seem to
> > perform the same as 32K.
> > 
> 
> 	I seem to stick to nfs settings in the man pages, the default is 1024bytes. They 
> seem to claim that thru' put is greatly improved with rsize=wsize=8192bytes. So 
> I am surprised that you are having a bad performance.
> 
> 
> > 
> > UDP.  I would like to try to use TCP but as far as I can tell the RH
> > kernels don't support it.  Anyone know how to enable TCP based NFS on RH
> > kernels.
> 
> 	Man nfs says that if you add tcp to the end of the options then it will start up 
> as a tcp nfs system as opposed to udp, RH ???. Again subject to experimenting 
> adding that at the end of the /etc/fstab line that I provided maybe adding tcp 
> may do the trick. Most of the Linux NFS is v2 maybe toggling it to version 3 may 
> provide more improvement ??? (Pure experimentation !)

My understanding is that nfs daemon and the kernel need to support it. 
Trond has a patch to allow TCP NFS support but I haven't applied it to
anything yet.  I was hoping that the stock RH kernel would support TCP,
they seem to have applied some of Trond's patches
http://www.fys.uio.no/~trondmy/src/

If I add the tcp option to the mount options I am not able to mount at 
all.

> 
> > 
> > A bit more info that I have noticed.  nfsstat reports that the number of
> > retries is pretty significant compared the number of calls.  The clients
> > all say that there is 1 retrans for every 3 calls to the PE 4600.  This
> > doesn't sound good to me but I am not sure how to fix it.
> > 
> 
> 	My system seems to provide more information for V3 NFS as opposed to that of V2. 
> I guess V3 maybe the defaults ?
> 
I agree V3 appears to be the default for RH linux implemementations.

Thanks for the input

Rick

> 	HTH.
> 
> 	Cheers,
> 
> 	Aly.
> 
> 
> -- 
> Aly Dharshi
> aly.dharshi at uleth.ca
> System Administrator ORS Servers
> University of Lethbridge
> 
> 	"A good speech is like a good dress
> 	that's short enough to be interesting
> 	and long enough to cover the subject"





More information about the Linux-PowerEdge mailing list