On Tue, Sep 1, 2009 at 8:21 AM, Florian<florian.bi...@gmail.com> wrote:
> This is what I expected. I fear also that the chance is doubled to crash the > system when the PAGING volumes are spread on both DS-8000. So I will correct > this soon. I used to joke that it would be best to put all our critical resources on a single volume to reduce the chances that any of them would be affected by a disk failure. The problem obviously is the amount of trouble you have when that volume would be affected. With modern disk subsystems I understand it is unlikely for a single logical volume to fail, but rather the full subsystem when something happens with it. But that risk is not the only aspect, you also need to look at performance. Many installations run two z/VM systems and two disk subsystems (with mostly PPRC). In such a setup it does make sense that when one disk subsystem would fail you only impact one z/VM system. When the subsystem dies, z/VM will not isolate you from the errors and the mirrors must be broken by hand. This will impact all virtual machines using the logical volumes that were affected and possibly basic z/VM functions. A common scenario is that in such a case you get the development systems off the surviving z/VM to pick up the (restarted) production load of the impacted system. This means that without further increasing the impact of a disk subsystem failure, you can spread I/O by separating production and development systems over the two disk subsystems. Rob