p5 Performance & Usage Statistics - Aix

This is a discussion on p5 Performance & Usage Statistics - Aix ; I'm pretty new to AIX and LPARs, so I need some assistance with monitoring the performance and usage. I have 4 p5s, each of them have varying number of LPARs on them. What is the best way to gather CPU, ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: p5 Performance & Usage Statistics

  1. p5 Performance & Usage Statistics

    I'm pretty new to AIX and LPARs, so I need some assistance with
    monitoring the performance and usage.

    I have 4 p5s, each of them have varying number of LPARs on them. What
    is the best way to gather CPU, memory, and other performance statistics
    not for the LPARs individually, but for the boxes themselves as a
    whole? I'm taking over the administration of these servers and want to
    establish more defined criteria for determining where to place new
    LPARs.


  2. Re: p5 Performance & Usage Statistics


    RickWaukCo wrote:
    > I'm pretty new to AIX and LPARs, so I need some assistance with
    > monitoring the performance and usage.
    >
    > I have 4 p5s, each of them have varying number of LPARs on them. What
    > is the best way to gather CPU, memory, and other performance statistics
    > not for the LPARs individually, but for the boxes themselves as a
    > whole? I'm taking over the administration of these servers and want to
    > establish more defined criteria for determining where to place new
    > LPARs.


    You might want to look at NMON V11 and the new NMON Analyser
    V3.2. I think NMON Analyser V4 is alpha next month and Stephen
    Atkins is adding what you want. Here is a copy of a recent email
    on the subject. Notice the last item. Looks like what you want.

    NMON V3.2 is now available for download from the Wiki

    http://www-941.haw.ibm.com/collabora...e/nmonanalyser

    this release is in support of NMON11 and includes various improvements
    suggested by users (e.g. a WLMPCPU sheet showing physical CPU for WLM
    classes and new "User and Arg" columns on the TOP sheet). It also has
    a few bug fixes that I didn't retro-fit to V3.1 when they were
    identified (mainly to do with printing, WLM, Linux and input files with
    >65K lines as I recall).


    Now that this is out of the way I'm starting work on V4 and, if you're
    interested, check out the Wiki for the Alpha release in November. V4
    involves a complete restructuring of the code and will do the
    following:
    Allow users to define their own graphs rather than rely on my ideas of
    what graphs should be produced
    Will support output from any program that generates output in NMON
    format
    Will have no dependencies between sections (i.e. will let you delete
    sections from the input file before processing)
    Will (hopefully) be faster and have a smaller memory footprint - but
    then again...
    Will merge multiple input files to provide either a consolidated view
    of a partitioned system or a trend analysis for a single system/LPAR

    Stephen Atkins, Certified IT Specialist, MBCS CITP
    System p Product Manager, UKISA
    IBM UK Ltd, NHBR-1PH, Portsmouth, PO6 3AU
    Email: steve_atkins@uk.ibm.com


  3. Re: p5 Performance & Usage Statistics

    Do iostat, vmstat, topas, sar, etc. (nmon?) measure stats on the lpar
    itself, the machine as a whole, or both?

    On Oct 12, 11:33 am, "RickWaukCo" wrote:
    > I'm pretty new to AIX and LPARs, so I need some assistance with
    > monitoring the performance and usage.
    >
    > I have 4 p5s, each of them have varying number of LPARs on them. What
    > is the best way to gather CPU, memory, and other performance statistics
    > not for the LPARs individually, but for the boxes themselves as a
    > whole? I'm taking over the administration of these servers and want to
    > establish more defined criteria for determining where to place new
    > LPARs.



  4. Re: p5 Performance & Usage Statistics


    RickWaukCo wrote:
    > Do iostat, vmstat, topas, sar, etc. (nmon?) measure stats on the lpar
    > itself, the machine as a whole, or both?
    >
    > On Oct 12, 11:33 am, "RickWaukCo" wrote:
    > > I'm pretty new to AIX and LPARs, so I need some assistance with
    > > monitoring the performance and usage.
    > >
    > > I have 4 p5s, each of them have varying number of LPARs on them. What
    > > is the best way to gather CPU, memory, and other performance statistics
    > > not for the LPARs individually, but for the boxes themselves as a
    > > whole? I'm taking over the administration of these servers and want to
    > > establish more defined criteria for determining where to place new
    > > LPARs.


    The utilities mentioned above do only show data for the LPAR in which
    they are run.

    You might check out the IBM AlphaWorks "Graphical LPAR Monitor for
    System p5 Servers" for monitoring the server as a whole.
    http://www.alphaworks.ibm.com/tech/lparmon


  5. Re: p5 Performance & Usage Statistics

    dhoffman@kc.rr.com schrieb:

    > > Do iostat, vmstat, topas, sar, etc. (nmon?) measure stats on the lpar
    > > itself, the machine as a whole, or both?


    > The utilities mentioned above do only show data for the LPAR in which
    > they are run.


    ^^^^^^^^^^^^^^^^^^^^

    Not quite correct....

    If a partition has "Allow shared processor pool utilization authority"
    set (you can set this on the HMC), it can also monitor the available
    (unused) processors of the whole machine.

    E.g. in lparstat look at the app column:

    root@aix-test: / # lparstat 1 5

    System configuration: type=Shared mode=Uncapped smt=On lcpu=8 mem=12288
    psize=8 ent=0.70

    %user %sys %wait %idle physc %entc lbusy app vcsw phint
    ----- ---- ----- ----- ----- ----- ------ --- ---- -----
    17.4 7.7 2.6 72.3 0.23 32.3 2.5 6.57 2684 34
    15.3 7.9 0.7 76.1 0.21 30.5 2.0 6.54 2782 41
    19.6 16.6 2.0 61.8 0.34 48.9 4.1 6.26 2564 57
    24.1 14.6 5.9 55.4 0.35 50.6 4.9 6.23 4197 79
    17.9 13.5 5.8 62.8 0.30 42.6 3.1 6.58 3829 62

    See the lparstat man page....

    HTH

    Pete


+ Reply to Thread