The Past and Future of Shared Servers.
In pursuing server consolidation we need to carefully balance what has worked in the past what has failed in the past. It is easy to point to IBM's OS/390 (and its predecessors) as a model for shared servers. This OS did a great job of making computing resources available to a wide array of applications, providing a high degree of process isolation and tweaks to ensure an equitable distribution of resources. On the other hand the product release cycle on the OS/390 is extraordinarily slow (or it is at our shop anyway).
Important technologies like DB2 and CICS, even MVS itself, often lag several release cycles behind the current release. This happens because of the enormous amount of coordination and testing required to ensure stable applications do not break under the new release. I believe it takes us close to 15 months to put in a new release of CICS for just this reason. These elongated release cycles are part of the reason for our move to the Windows computing environment; we expect to be more nimble (as application developers) when we have more direct control over the hardware and the applications running on it (i.e. just our own).
As we move back in to a shared computing resource environment we need find a balance that enables us to employ the benefits from both environments (the mainframe's resource sharing structure, the PC's expense and agility) while minimizing the negatives (mainframe's sluggish release cycle, PC's floor space and underutilization).
8:20:55 AM
|
|