Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?
Silicon Graphics Software

SGI's Open Source Performance Co-Pilot 24

codesmythe writes "The Fates, through SGI nee Rackable, have granted a new beginning to Silicon Valley's once darling Silicon Graphics. Despite old mistakes and economic misfortunes, Silicon Graphics' engineering contributions are legendary: their systems (oh, the systems!), and software such as the well known OpenGL and the little known Performance Co-Pilot. PCP is an enterprise-class open source system monitoring, measurement, and visualization infrastructure — overlooked in last fall's monitoring tool discussion. Since its proprietary beginning in 1993, PCP has been re-released as open source and ported to all major operating systems. Readers of Slashdot's recent Beginning Python Visualization book review will be pleased to hear there are Python interfaces to PCP data sources. Here is an example of using Python and Blender to visualize PCP data (registration may be required). The PCP dev community is well and active, and includes several of the original team members."
This discussion has been archived. No new comments can be posted.

SGI's Open Source Performance Co-Pilot

Comments Filter:
  • by OzPeter ( 195038 ) on Friday May 29, 2009 @10:40AM (#28138775)
    PCP [wikipedia.org] will allow you to see things??
  • by Anonymous Coward

    Sure would be nice to have larger block sizes and real-time disks, etc. from the Irix version of XFS available under the GPL.

    • Re: (Score:1, Interesting)

      by Anonymous Coward
      Not to mention clustering, aka CXFS.
  • nee rackable? (Score:2, Informative)

    by grub ( 11606 )
    "nee rackable" is incorrect. nee [wikipedia.org]
    • "nee rackable" is incorrect. nee [wikipedia.org]

      According to Merriam Webster [merriam-webster.com], it is correct usage of the word. See definition two.

      • I must admit I was confused at first. I remember that Rackable bought SGI, so "Rackable nee SGI" would be somewhat logical: The SGI from back then is now called Rackable. However, since Rackable then changed its name into SGI, we also have "SGI nee Rackable".
        • >I remember that Rackable bought SGI, so "Rackable nee SGI" would be somewhat logical

          No, because Rackable was never previously known as SGI.

          You could say that a *unit* of Rackable was previously known as SGI, except that unit would not, strictly speaking be known as just "Rackable", but rather the "SGI unit of Rackable" -- and in any event, Rackable pretty much instantly changed their name to SGI (Silicon Graphics International Corporation, to use the full name) as soon as they acquired old-SGI's assets.

          • by anothy ( 83176 )
            the usage is correct; the thing currently called SGI used to be called Rackable. the fact that Rackable included, however briefly, a sub-part called SGI, which had its own prior existence, is irrelevant.

            sorry, let me clarify: the usage is correct, but pretentious. i'd be more forgiving if the accent was correct.
  • I've used it... (Score:3, Informative)

    by Anonymous Coward on Friday May 29, 2009 @01:09PM (#28140523)

    I've written systems that use Performance Co-Pilot. It is actually really sweet, and gives you so much in terms of general accounting. We tracked CPU load, network load, user/kernel time, as well as adding our own tags to track. The PCP infrastructure had automatic archive and visualization of all data, including realtime visualization, as data was being logged. You can also plot relationships (in real time).

    This stuff is certainly the non-sexy part of enterprise software development, but it is pretty cool once it is integrated (actually quite easy), and gives you so much. Not to mention, saves *huge* amounts of time because it was so complete.

    • Re: (Score:2, Interesting)

      by alex_vegas ( 891476 )
      I'm curious, how does this package handle snmp data? Looking through their commit tree, I found one commit which mentioned snmp, and nothing at all in a full text search of the docs. http://oss.sgi.com/cgi-bin/gitweb.cgi?p=pcp/pcp.git;a=commit;h=1500bd088898317c42eede0f0748f1fd09989c69 [sgi.com]
      • by wiosd ( 576937 )
        The developers are pretty helpful.
        Try posting your question on the PCP mailing list or IRC channel.
        Details on the "Developers" page at http://oss.sgi.com/projects/pcp/ [sgi.com]

        Disclaimer: I was a PCP developer in a previous life. ;)
      • by kenmcd ( 1565451 )

        PCP was developed for addressing hard performance problems in complex, often distributed, systems. The relevant performance data is hiding the in the hardware instrumentation, the operating system, the service layers (dbms, web, mail, ...), the network and the applications.

        So there is not a lot of SNMP coverage here to start with.

        Secondly, the metadata services of SNMP are rather weak and we wanted something much richer to describe the available performance data that changed from platform to platform and

    • by wiosd ( 576937 )
      ...and if it doesn't support something you want to monitor/log, the API is documented and there are examples, so you can roll your own agents (or PMDAs as they are known), create your own visualisations, etc. Disclaimer: I was a PCP developer in a previous life.

This process can check if this value is zero, and if it is, it does something child-like. -- Forbes Burkowski, CS 454, University of Washington