home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sun.admin
- Path: sparky!uunet!mcsun!sunic!ugle.unit.no!trd.sdata.no!halvard
- From: halvard@trd.sdata.no (Halvard Halvorsen)
- Subject: Re: Kernel Config -- Can maxusers be too big?
- Message-ID: <1992Jul22.143012.28974@ugle.unit.no>
- Sender: news@ugle.unit.no (NetNews Administrator)
- Organization: Skrivervik Data A/S
- References: <1992Jul19.160905.11322@kodak.kodak.com> <14emaeINNd8b@early-bird.think.com>
- Date: Wed, 22 Jul 92 14:30:12 GMT
- Lines: 74
-
- from the 4.1.2 release manuals - chapter 8.3 :
-
- maximum MAXUSERS values :
-
- arch mem<64mb 128mb 640mb 2.5gb
-
- sun4 225 225 NA NA
- sun4c 225 225 NA NA
- sun4m 185 180 155 45
-
- well it clearly says MAXUSERS <= 225 for sun4 - so you are clearly running
- an 'unsupported configuration' ...
-
-
-
-
- In article <14emaeINNd8b@early-bird.think.com>, barmar@think.com (Barry Margolin) writes:
- |> In article <1992Jul19.160905.11322@kodak.kodak.com> dennett@sunshine.Kodak.COM (Charlie Dennett) writes:
- |> >Can the maxusers parameter in the kernel config file be so big that
- |> >performance is affected? Now, I have seen it where it was too big
- |> >and the kernel would not boot because there was not enough memory.
- |> >However, assuming memory is sufficient, can a large maxusers actually
- |> >hurt performnce?
- |>
- |> Most (all?) of the tables whose sizes are derived from maxusers are wired
- |> in memory. Therefore, this memory is not available for use as paging
- |> space. If you increase maxusers enough that the working set often doesn't
- |> fit in memory, you'll page more. This will slow the system down (how much
- |> depends on the difference between the working set and available memory).
- |>
- |> > (the target of my attack is a 4/490 running 4.1.1 with
- |> >128 meg of memory. It is a major server -- 18 spindles and 16+ gig --
- |> >for both disks and cpu cycles and also supports a half dozen or so
- |> >Xterminals. Maxusers was at 128. I tried raising it to 256. It booted
- |> >fine. Pstat -T told be the max number of files was perposterous at
- |> >over 14000. What does pstat consider non perposterous? It's running
- |> >at 256 now. We'll see what happens when the users start wailing on it
- |> >Monday.)
- |>
- |> Pstat considers nproc or nfiles more than 10,000 to be preposterous. I
- |> don't think it's intended to be a judgement about your configuration
- |> decisions, but a sanity check; unexpectedly large values could be the
- |> result of /vmunix not being the kernel you booted, so it's looking in the
- |> wrong place for the value.
- |>
- |> >Side issue: As you may be able to tell by this posting and a couple
- |> >of my previous ones, in case you noticed, I am trying to teach myself
- |> >and to understand performance issues. Is this group a reasonable
- |> >place to discuss performance issues and tuning? If not, what would
- |> >be an appropriate place?
- |>
- |> This is probably a good place, as would be comp.unix.large. Also, there
- |> are some good books, such as "System Performance Tuning" by Mike Loukides,
- |> published by O'Reilly & Associates (they have lots of other good books for
- |> Unix users and administrators).
- |> --
- |> Barry Margolin
- |> System Manager, Thinking Machines Corp.
- |>
- |> barmar@think.com {uunet,harvard}!think!barmar
-
- --
-
- 0 halvard halvorsen - systems consultant - Skrivervik Data A/S
- \/|/\ email : halvard@trd.sdata.no
- | telephone : + 47 7 59 23 74
- / \/\ fax : + 47 7 59 23 30
- \/ ___________________________________________________________________
-
-
-
-
-
-
-