home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.hp
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!destroyer!fmsrl7!ef2007!bkelley
- From: bkelley@ef2007.efhd.ford.com (Brian Kelley)
- Subject: Re: Disk performance question on hp9000/730
- Message-ID: <BtLz9n.AMq@ef2007.efhd.ford.com>
- Keywords: disk performance hp 9000 730 PA RISC SCSI
- Organization: Ford Motor Company
- References: <duff.714850491@starbase>
- Date: Wed, 26 Aug 1992 20:51:22 GMT
- Lines: 54
-
- In article <duff.714850491@starbase> duff@starbase.mitre.org (David A. Duff) writes:
-
- >i've already spent some time reporting this problem to hp, and i've even
- >had a service person come out.
-
- You should have had the service person check your shift keys :-)
-
-
- >before i pursue this further, i'd like to be absolutely sure that what
- >i'm experiencing is a problem that is specific to my machine. i'm
- >wondering if anyone out there has tried running iozone on a 730 or 750
- >(should be comparable performance). what numbers are you getting?
- >
- >i'm consistently seeing the following:
- >
- ><800K/second writing
- ><600K/second reading
-
- Something is wrong, but you didn't give much info. What OS version?
- How much RAM do you have?
-
- On a 720 with diff SCSI and a 1.2g disk, 32 MB ram and 8.07, I get:
-
- pms709>/tmp/iozone 23
- ...
- Writing the 23 Megabyte file, 'iozone.tmp'...11.560000 seconds
- Reading the file...14.870000 seconds
-
- IOZONE performance measurements:
- 2086267 bytes/second for writing the file
- 1621872 bytes/second for reading the file
-
- I have observed very little performance difference between our 720/730/750
- machines when conducting this simple test. I never get numbers that low
- (on an otherwise unloaded machine).
-
- Note that disk buffers are statically allocated in 8.05 and 8.07. You
- may want to tweak the setting of the NBUFS kernel variable. Check TFLROM
- under 'bufpages' for more info. Note that I have not had to touch any
- kernel variables to obtain the above performance - you shouldn't have to
- either.
-
- >i suspect a problem with the i/o board on my machine. anyone seen or
- >heard of similar problems? also, if anyone has performed these or
- >similar tests, please post or send me your results.
-
- Something is definitely not right.
-
- Brian
-
- ---
- bkelley@pms001.pms.ford.com
-
-
-