Rob, are you sure CP selects the channel to use?  That was the case in the
370 era.  AFAIK, in XA, CP will create a mask telling which channels *can*
be used, but it is the IO subsystem that will select the channel and maybe
even queue the IO if all channels would be busy.  In XA mode CP only queues
an IO if it knows the device is busy (ie it launched an IO itself), device
busy due to IO's from other LPARs or CECs will be queued by the IO
subsystem.

2010/5/21 Rob van der Heij <rvdh...@gmail.com>

> On Thu, May 20, 2010 at 3:39 PM, Billy Bingham
> <billy.bingham...@suddenlink.net> wrote:
>
> > Hello all,
> > Risking the old saying that it's better to have people think you're
> stupid
> > than to open your mouth and remove all doubt :)
> > If VM is doing I/O to one DASD volume only, say a DDR DUMP or RESTORE,
> and
> > everything else is equal, will VM use multiple channels to get to the one
> > volume, or will it do all it's I/O on whatever channel it gets first and
> > only that one channel?
>
> DDR will wait for an I/O to complete before it starts the next one.
> And with all I/O to the same real volume, it would also be serialized
> by z/VM. So if nothing else is going on z/VM can do with just one
> channel. I don't think CP will try to balance over the channels, but
> picks another one when the first choice is busy. Some DASD
> configurations encourage CP to prefer one channel over the other.
>
> In your scenario, you would see it all go via one channel. Something
> like this (with apologies for those who read mail in proportional
> font):
>
> ESACHAN  Velocity Software-Test VSIVM4  ESAMON 3.7
> Channel Performance Analysis            CHANNEL 00
>
>                   Pct Channel <--------------Data
>         <Channel> Utilization <---Reads/Second-->
>  CHP Shr Class Typ  LPAR Total LPAR TOTAL Pct  Max
>  --- --- ----- --- ----- ----- ---- ----- --- ----
>  00 Yes FICON FC    0.1   0.1    4     6   0 195K
>  01 Yes FICON FC    5.8   5.8 6405  6416   3 195K
>
> I can't really think of real life situations where that would be a concern.
>
> Rob
> --
> Rob van der Heij
> Velocity Software
> http://www.velocitysoftware.com/
>



-- 
Kris Buelens,
IBM Belgium, VM customer support

Reply via email to