On Saturday, 10/20/2007 at 12:38 EDT, Patrick Spinler <[EMAIL PROTECTED]> wrote:
> For example, updating the system on a z series (e.g. upgrading z/VM) > requires an outage of *all* the linux guests hosted there. This is in > addition to bearing the cost of upgrading a linux guest (e.g. from > SuSE 9 to SuSE 10) effecting the individual guest. > > Upgrading AIX on an LPAR effects only that one AIX instance. Upgrading AIX is upgrading AIX, without regard to the "realness" of the server it is running in. Likewise with upgrading a Linux Regardless of whether Linux is running in an LPAR or a virtual machine, an upgrade of that Linux affects only that Linux. With z/VM you *can* create Linux instances that have such a high degree of sharing that an update to one is an update to all, but that's up to you. If you need to service z/VM, yes, the system has to be restarted. But that's why we have CSE clusters so that you can bring the server up on another LPAR and keep the service alive with only a brief outage. Some folks have seen Romney White (one of my colleagues) demonstrate at SHARE some z/VM technology that relocates a running guest from one z/VM partition to another without an outage. Even though it was a prototype, it demonstrated that z/VM is capable of fulfilling the "zero down time" for your guests and we are busily working to bring that to a theater near you. > Obviously, also, running a virtual machine for something that needs > any significant CPU horsepower is contra-indicated. Also, if there's > requirements for certain types of dedicated bandwidth (e.g. needing 4 > bonded gigE's for a very high throughput file or backup server) you > might as well do it on a distributed platform. The z/VM Virtual Switch provides for bonding up to 8 OSAs together in the event the aggregate throughput on that VSWITCH exceeds the capacity of one OSA. The new z6 CPU is a lot faster than the z9 (go see IBM-MAIN for discussions), so the conventional wisdom will need to be reevaluated when that CPU makes its way into a box. You're correct, though, that some CPU-bound apps will likely never be appropriate for z (or partitioning or virtualization of any flavor, for that matter). Alan Altmark z/VM Development IBM Endicott ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
