|
HP OpenVMS Systemsask the wizard |
|
The Question is: I am migrating from a VAX 6620 VMS 5.5-2 with HSD controller to a 4100 Alpha OpenVMS 7.1. The application is identical (RMS files managed by COBOL programs). While I can see that the CPU performance is much better, the batch programs do not seem any faster. They obviously are waiting for DIO to complete. At the moment the Alpha has very fast disks RZ1CB attached to a regular controller. My test program runs around 140 Disk I/O Operation Rate in both cases. Any idea where I should look to improve the Alpha performance ? Would an HSZ70 subsystem improve anything ? Is there any RMS parameters to watch especially since I go from 5.5 to 7.1 ? Help would be appreciated. Tks The Answer is : If your application is performance-limited by the storage interconnect, then moving to a faster storage interconnect will help, while moving to a faster CPU will generally provide lower (if any) improvement. The original note does not clearly indicate if the existing HSD storage has been migrated to the AlphaServer 4100 series system, or if a newer storage subsystem -- other than an HSZ70 -- is in use. If the existing DSSI HSD storage subsystem was migrated for use on the new AlphaServer system, then the reported I/O levels would appear consistent with an I/O-limited application. And thus a faster I/O subsystem would help. The Wizard would recommend basic performance tuning of the RMS files including passes of CONVERT and CONVERT/RECLAIM as appropriate, proper file extension size settings, checking for (and avoiding) large numbers of duplicate keys on indexed files, use of EDIT/FDL to design an appropriate indexed file structure, appropriate use of local and global buffers, reasonable sizes for the process multibuffer and multiblock counts, and other similar RMS tuning operations. Memory and CPU can sometimes be used to offset I/O limitations, through the use of more aggressive data caching operations.
|