Re: NFS and SAMBA on RELENG_5 vs RELENG_4

From: Eric Anderson (anderson_at_centtech.com)
Date: 01/20/05

  • Next message: stheg olloydson: "Re: FreeBSD 5.3 I/O Performance / Linux 2.6.10 | More Info" 
    Date: Thu, 20 Jan 2005 11:35:58 -0600
To: Mike Tancsa <mike@sentex.net>

    Mike Tancsa wrote:
    > At 11:11 AM 12/01/2005, Eric Anderson wrote:
    >
    >
    >> Mike - let me know if you want any tweaks that I've done for anything
    >> specific.
    >
    >
    > Hi,
    > Yes, by all means please post any suggestions you have. I am
    > going to take the time and try and test a bit with a half dozen client
    > machines on both RELENG_4 and RELENG_5 to see what differences there
    > are. If you can suggest some benchmark tools / methods that would be
    > great.

    I can tell you this - you must increase the number of nfsd threads to a high number, if you plan on really hammering the machine with nfs and lots of clients. I recompiled the nfsd binary with it tweaked to allow 256 threads, and that still isn't quite enough. You need something on the order of: 1 per active machine using nfs * 1.10. The hard part is finding out how many active machines you have. I usually start with about 20% of my total machines mounted to the server, and then watch the nfsd threads cpu time. If the lowest thread is using more than about 3-4% of the time of the 10-15th top nfsd process, then you need to bump up the number. That may be confusing..

    The other thing you can do, as an easy general beef up, is bump the maxusers sysctl to something higher, like 512. As you beat on the server, you should also watch nmbufs via netstat -m and all it has to offer. You may want to tweak some other tcp sysctl's (like send and recv space) by bumping them up a notch. I have found that lots of memory is a key to a good fast NFS server, and of course decent NICs that don't hog cpu time and a good disk subsystem (hardware RAID50, 10, or 5, depending on your needs).

    There's more, but I can't think of them right now.. :)

    Let me know if you have specific questions..

    Eric 

    -- 
------------------------------------------------------------------------
Eric Anderson        Sr. Systems Administrator        Centaur Technology
I have seen the future and it is just like the present, only longer.
------------------------------------------------------------------------
_______________________________________________
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscribe@freebsd.org"
  • Next message: stheg olloydson: "Re: FreeBSD 5.3 I/O Performance / Linux 2.6.10 | More Info"

    Relevant Pages

    • Re: SATA: RAID 5 controller recommendations
      ... Mike Tancsa wrote: ... >> I have seen some network based SMP related performance problems ... >> To unsubscribe, send any mail to ...
      (freebsd-performance)
    • Re: Testers needed: Joes MFC of USB code
      ... On Wed, 25 Feb 2004, Mike Tancsa wrote: ... I just applied it to a couple of boxes. ... a previous MFC candidate? ... To unsubscribe, ...
      (freebsd-hackers)
    • Re: make buildworld problems on RELENG_5 (linprocfs)
      ... Mike Tancsa wrote: ... >> re@ approved an instant MFC of this, since this is a build fix. ... To unsubscribe, ...
      (freebsd-current)
    • Re: http://www.openssl.org/news/secadv_20060905.txt
      ... At 10:53 AM 9/5/2006, Mike Tancsa wrote: ... I manually added in the diffs and everything seems to compile and function with some limited testing. ... make install ... To unsubscribe, ...
      (FreeBSD-Security)
    • Re: RELENG_4 panic
      ... On Wed, 21 Apr 2004, Mike Tancsa wrote: ... > The box is a somewhat heavily used outbound smtp smarthost. ... trace. ... To unsubscribe, ...
      (freebsd-stable)