disk bottleneck in XP128

[XJIM]

I have a XP128 attached to my HP-UX with 4 optic fibers and Autoptah. The Xp128 use disk arrays for the LUNS and have a cache of 12 GB
In the GlancePlus Disk Graph some LUNS (DISK) have 100% when i copy large files (1 or 2 GB) to filesystem in this disk.
Why the LUNS of XP128 have 100% only a copy files?
Is there a disk bottleneck or Filesystem bottleneck?

[Sridhar Bhaskarla · Accepted solution]

Hi,

Utilization at 100% doesn't necessarily mean there is a bottleneck. But if you see Qlen (in the same window) on that disk, then there may be. This means the system is pushing faster than the disk subsystem can handle.

-Sri

[A. Clay Stephenson]

This is a common "problem" when using Glance (or any other host-based) tools to measure IO on high performance devices like disk arrays. The only thing Glance knows (or can know) is that a tremendous amount of IO is going through what it sees as one physical device. It has no way of knowing that what it sees as a disk (or LUN) is actually comprised of multiple physical disks within the array. Almost certainly, you have no real problem --- just the appearance of one.

I have seen cases where admins would divide a VG into several LUN's just so things would appear better to Glance although the actual I/O rate was no better.