For performing processor or memory intensive work on an integrated server, the performance characteristics are equivalent to a stand-alone server using dedicated disk drives. Since the integrated server disk drives are allocated out of iSeries™ storage, the disk performance is dependent on the iSeries.
Greater disk performance capacity with iSeries shared disks
On most stand-alone servers a few disks are dedicated to each server. For applications with a small average disk load, the performance is adequate. However, there can be periods of time where the server performance is limited by the capacity of those few dedicated disks.
When the same group of servers is integrated with the iSeries, the virtual disks are spread across more iSeries hard disks. The total average disk load does not need to be any greater than for a group of servers with dedicated disks. But, when an individual server temporarily needs more disk performance capacity, it is available through the larger set of iSeries disks.
On servers with dedicated disks, the disk response times tend to be relatively steady. For example, you might take advantage of the predictable response time and configure the Windows Performance Monitor to produce alerts when disk response times exceed typical thresholds and indicate exceptional conditions which may need your attention.
On an integrated server, the iSeries storage, CPU and memory are shared between the integrated server and iSeries applications. It is normal for Windows disk response to swing through a larger range. Short periods might occur where I/O operations from multiple integrated servers, or other iSeries operations contend for the same disk. Some disk intensive iSeries applications (like SAV and RST), can reduce the disk performance seen on the Windows server for a period of time. This can make it more difficult to choose a threshold value for short time periods.
Consider the entire group of disks when you evaluate storage bottlenecks
The iSeries server storage space appears as one disk drive within Windows. When the Physical Disk average queue length (in Windows Performance Monitor) exceeds two, the server performance is not necessarily disk constrained. Assuming that memory paging issues have been ruled out, a queue length of two or a Windows disk utilization of 100% only points to a storage bottleneck if there is only one physical disk drive to perform the operations. There are usually multiple disks on the iSeries server in the storage space ASP operating in parallel. Typically, two times the number of disks in the ASP might point toward a disk bottleneck. You might also need to account for the average queue lengths of all the servers using the storage ASP.
