home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.protocols.nfs:1920 comp.sys.sun.hardware:3475
- Path: sparky!uunet!gatech!prism!gt1111a
- From: gt1111a@prism.gatech.EDU (Vincent Fox)
- Newsgroups: comp.protocols.nfs,comp.sys.sun.hardware
- Subject: Re: NFS I/O Ops/seconds
- Message-ID: <64081@hydra.gatech.EDU>
- Date: 22 Jul 92 14:45:26 GMT
- References: <1992Jul22.061146.15641@u.washington.edu>
- Followup-To: comp.protocols.nfs
- Distribution: usa
- Organization: Georgia Institute of Technology
- Lines: 47
-
- In <1992Jul22.061146.15641@u.washington.edu> kint@rio.engr.washington.edu (Rick Kint) writes:
- > We're buying a Sun to serve as an NFS server in one of our departments.
- >The Sun rep commented that a typical Ethernet allows 300 NFS ops per second,
- >so if you're on a single wire any server bandwidth beyond that is wasted.
-
- > I have no feel whatsoever for the numbers that go into this figure,
- >can anyone in netland (a) explain this in words of two syllables or less, and
- >(b) does anyone know the figures for current systems?
-
- Hmmm. I'm a bit skeptical. All NFS operations are not created equal.
- A write obviously takes a lot longer to complete than a getattr. I've seen
- claims up to 1,000 or better. Auspex claims this I feel certain.
-
- > I'm sure that this includes assumptions about what represents a
- >typical mix of operations; since a getattr is presumably quicker than a write
- >(for example), how useful are these numbers anyway?
-
- Not much. Sun reps usually know less than you do.
-
- > ObCurmudgeonlyRemark: I was reading about the SS10/41, with 1 MB of
- >cache. Good Lord, a meg of cache on a workstation, why when I was your age
- >we thought a meg of *main memory* on a *mainframe* was a lot...
-
- I remember when we had teletypes with 300 baud acoustic modems, and
- we *liked it that way*!
-
- There's not much you can do to accelerate Sun's NFS on the read-side,
- but you can speed up writes quite a bit. We use the PrestoServe
- board, which buffers up NFS writes and reduces the response-time
- between concsecutive write operations. InterStream's eNFS software
- also replaces the nfsd's with enfsd's and does better scheduling.
- If you are doing more than 7-10% writes (do an nfsstat -s), you can
- really speed things up. You also will want to increase your clients
- to more than just the default 4 biod's to take advantage of the
- faster server. InterStream was sending out evaluation copies for
- free last I head. The try-out license number lasts about a month
- I think if you want to evaluate it on-site before purchase. Comes
- with benchmarking software to determine optimum configuration.
-
- Prior to adding these two things, I might get 100-200K/sec write
- performance on moderate-size (512K-1meg) files, now it's more like
- 400-500K/sec. This with a Sparc II and the built-in SCSI controller.
- I'm sure with a better boxen and maybe an NC4000 ethernet board, you
- could push things a bit faster.
-
- --
- The War on Drugs is just a part of the War on the US Constitution.
-