Date: Mon, 10 Jun 1996 16:58:19 -0700 (PDT) From: Steve Khoo <steve@gordian.com> To: Rob Snow <rsnow@lgc.com> Cc: freebsd-hackers@freebsd.org Subject: Re: NFS problem: Irix 5.2 server, 2.2-960501-SNAP client Message-ID: <199606102358.QAA14378@delphi.gordian.com> In-Reply-To: Rob Snow's message of Fri, 7 Jun 1996 17:33:42 -0500 (CDT) References: <199606071932.MAA08925@delphi.gordian.com> <Pine.SGI.3.93.960607172725.4450A-100000@dympna>
next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> "Rob" == Rob Snow <rsnow@lgc.com> writes:
Rob> On Fri, 7 Jun 1996, Steve Khoo wrote:
Rob> [SNIP]
>>
>> Thanks for the all responses and suggestions... It looks like the
>> problem is related to diskstriping on the Irix 5.2. We've been
>> playing around with diskstriping on the sgi to improve performance. I
>> truned off diskstriping and rebooted the sgi lastnight and all seems
>> to be fine now. I supposed it is possible that the sgi was somehow
>> hosed and rebooting fixed it. However, this is unlikely because nfs
>> mounts to other systems were still working; I did umount and mount on
>> sunos and hp machines to the sgi and worked without a problem. I'll
>> have to test it again with striping truned ON on the sgi later, just
>> to double check; Since this sgi is a production unit, I can't be
>> rebooting it too many times.
Rob> Just checking it on my Indigo2 - Irix 6.2 and mounted an xfs partition
Rob> with no problem. I've seen problems with Irix and mounting other
Rob> filesystems, usually solved by -o vers=2 to force v2 instead of v3.
>> Anyway, we are pretty impressed with this P6 unit so far. We did some
>> benchmarking with a realworld example(we use a gcc cross-compiler for
>> embeded mips system and built one of our source tree), and here are
>> the resuls:
>>
>> SGI 21 mins
>> P6(NFS) 16 mins
>> P6(local) 7 mins
Rob> Good looking numbers!
>>
>> The source tree is on the SGI server. P6(NFS) is the compile time for
>> the P6 with the source tree mounted via NFS from the SGI. P6(local)
>> is with the source tree copied over to a local disk on P6.
>>
>> System specs:
>>
>> SGI Challenge S server (Irix 5.2)
>> 150 Mhz R4600 MIPS Processor
>> 160MB RAM
>> Fast Wide Differential SCSI Controller
>> Seagate Barracuda FWD SCSI drives: 2's and 4's
>> 10BaseT NIC
Rob> Is this a PC (Primary cache only) or SC (L2 cache added)?
It has 256K L2 Cache.
>>
>> P6 (FreeBSD 2.2-960501-SNAP, kernel rebuilt without FAILSAFE option
>> and all the other unnecessary drivers)
>> Intel Pentium Pro 200 Mhz CPU
>> 64M EDO RAM
>> Asus P/I-P6RP4, Orion chipset stepping B0 motherboard
>> NCR Fast Wide Differential SCSI Controller
>> 4G Seagate Barracuda FWD SCSI drive
>> SMC 9332DST 10/100BaseT DC2114 based NIC
>> ----------------------------------------------------------------------
>> Steve E. Khoo Gordian
>> Systems Manager 20361 Irvine Ave
>> Internet: steve@gordian.com Santa Ana Heights, CA 92707
>> Phone: (714)850-0205 FAX: (714)850-0533
>> ----------------------------------------------------------------------
Rob> --
Rob> Rob Snow
Rob> Sr. Systems Administrator
Rob> Landmark Graphics Corporation
BTW, what kind of nfs performance should I be getting from the above P6
system as an NFS server? I'm getting pretty disappointing results:
On an isolated 100BaseT network I'm only getting:
IOZONE performance measurements:
247132 bytes/second for writing the file
4488887 bytes/second for reading the file
This is from a p5-100Mhz 512K PB Cache / 32M RAM nfs client.
SEK
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606102358.QAA14378>