HP OpenVMS Systems

Increasing I/O queue depth?

» close window

The Question is:

 
I have become alarmed by the statistics show by MONITOR DISK /ITEM=ALL on my
 Alpha 4100.
 
We see the IO Queue Length climbing steadily, even when the IO rate behaves
 normally (varying between 0 and a few 100's)
 
The IO Queue length does fall sometime too, but generally it just climbs up to
 several hundred!!!
 
Yesterday I saw it peak over 1000 and average over 700! In these circumstances
 it never reduces below many hundred either. Sometime by morning (not much
 overnight activity) it has returned to zero.
 
Surely this cannot be correct.
 
Otherwise the system seems to behave and perform normally, however recently
 there have been 3 system hangs and I'm concerned that these may be related.
 
Thanks,   James.
 
 

The Answer is :

 
  A queue-depth above one-half would indicate that more than half of all
  I/O operations to the disk are waiting.  Disks with a queue depth of
  700 would indicate a severely overloaded spindle.
 
  Assuming you have typical access throughput to the target disk spindle,
  this would appear to be a problem with the I/O queue length mechanism
  or with the MONITOR utility.
 
  Please check for and apply the mandatory ECO kits and any other relevent
  ECO kits, and (if the problem persists) please contact the Compaq Customer
  Support Center directly -- more information on the disk device and on the
  OpenVMS system will be required.
 

  

     

     
answer written or last revised on ( 20-DEC-2000 )
     
» close window