On Wed, 2003-11-26 at 19:00, Alejandro Leyva Rabinovich wrote:

> lxrtr1 is a vm with an osa dedicated, and two virtual nics, i dont
> understand why it is ALWAYS in i/o wait, but thats not the point, i dont

VM63282 makes CP stop to consider the outstanding Read CCW for network
adapters something to wait for...

> understand why, if the dasd 1209 are less than 30% utilized, the vm user
> goes into i/o wait state, and, i dont understand exactly what means that
> the user is in that state, also, are 9291 ioreqst a lot?

Whether 9291 is a lot depends on what the I/O is. A single SSCH (the
unit that is being counted) may be 4 KB but could run upto 450KB. So the
300 SSCH's could be anywhere between 1 and 135 MB/s.

Unfortunately your RTM output does not show average connect time per I/O
(which makes me wonder about the hardware platform involved). That
number would be helpful to determine how much data is involved.

Further, since the guest has a QDIO network device it will be classified
as I/O-wait if nothing else (see VM63282).

> why does, if the volume is across three dasd's (1208, 1308 and 1138), only
> 1208 receives a high rate of ioreqst?

I believe the default for LVM is not to stripe over the volumes, so
depending on how much of the logical volume is in use you might end up
accessing data on only one of the devices. In case you *are* striping
your stripe may be large compared to the range of blocks you access in a
period of time?

Rob

Reply via email to