HP OpenVMS Systems

ask the wizard
Content starts here

Determining Disk I/O Performance?

» close window

The Question is:

How can I get the maximum I/O Operation Rate supported by a hard disk? Somehow
 MONITOR knows this because the command "MONITOR DISK /PER" displays I/O
 Operation Rate by percent. Where does MONITOR get the maximum from? Is it
 hardcoded? in a file somewhere
? or is it probed on startup? f$getdvi() can  provide the current operation
 count since boot (OPCNT) but not the maximum supported I/O Operation Rate as
 far as I can see. From the percent displayed by MONITOR on a running system I
 calculated that a Compaq
 10K rpm 18 Gb hard drive supports max. 330 I/O Operations per second. But I
 assume that his figure varies with rpm, SCSI version, storage capcity, etc. I
 need this figure in order to provide what the MONITOR command does but in a
 DCL that I'm making to b
e used as an alternative to having to use multiple utilities such as MONITOR,
 SHOW, INSTALL and ANALYZE to get the resource usage by percent.
Thank you very much.

The Answer is :

  MONITOR simply totals all I/O performed on the entire system for the
  interval, and has no specific knowledge of the disk capabilities.
  The information displayed by MONITOR is a percentage of the aggregate
  I/O on the OpenVMS system, and not a percentage of the maximum possible
  I/O rate to a particular disk device.  For example, if in an interval
  only one disk does all of the I/O, then it will have done 100% of the
  I/O displayed by MONITOR -- even if the device performed a single I/O.
  Disk I/O statistics are available from the disk drive vendor, though
  the values seen in any particular environment are exceedingly dependent
  on the I/O transfer sizes, the I/O access patterns, the effectiveness
  of the various I/O caches that often exist, host and bus and controller
  contention, and a wide variety of other factors.
  A (very simplistic) rule-of-thumb for aggregate I/O follows:
      Disk   I/O Per
      RPM    Second
      ----   -------
      3600     60
      5400     80
      7200     100
      10K      120
      15K      140
  The effective rates are obviously highly dependent on the details of the
  actual I/O activities.  Spirialing data onto a modern disk can provide
  rates of thousands of I/Os per second, for instance -- this is what is
  often refered to as the "spiral transfer rate".  If your data is not
  sequential, then performance tends to be affected by the access time.
  RAIDsets of magnetic disks, of course, can obviously also improve the
  read I/O rates, as RAID controllers often perform read I/O transfers
  using the set of data that first passes an available disk drive head,
  regardless of which constituent volume is involved.
  At present, the HSV-series storage controllers typically provide the
  highest available disk I/O throughput.

answer written or last revised on ( 6-MAR-2002 )

» close window