|
HP OpenVMS Systemsask the wizard |
|
The Question is: How to replace disk in VAX DSSI cluster and improve performance? I've a VAX 4505 cluster with 2 HSD10 & 2 R400X. The current cnfig is that each HSD10 has 7 disks in it and 2 of the them is mirrored by Volume Shadow as DSA1($4$DUA11，$4$DUA12)， and DSA2($4$DUA21, $4$DUA22), separately. Now it suffers a per formance problem(DIO is about 70, IO queue is 3). 1) I want to replace the 4 disks with larger disk to get more capacity. What's the detail commands and procedure? 2) ay I get performance improvement by this way? 2.1) create a stripeset with the 2 new disks in each HSD10, and create 2 partitions in the stripeset. 2.2) Mirror one partition in a HSD10 with the other one in the ther HSD10 by Volume Shadow? 3) How to keep the disk device names unchanged after the replacement so that all appliactions can work as before without any code or configuration modification? The Answer is : The OpenVMS Wizard would tend to replace the entire configuration, as the central performance limit is likely that of the DSSI (and, for that I/O traversing it, the Q-bus), and as the performance of the VAX 4000 model 505 is itself rather glacial by current system performance standards. DSSI has the performance characteristics of SCSI-1, and that limit will very likely throttle aggregate I/O performance here regardless of the particular disks installed. An I/O queue depth of 3 indicates that your disk I/O is overloaded -- the OpenVMS Wizard generally considers a depth average above 0.5 to be bad, as that indicates half of all I/O is waiting. Your queue depth of 3 indicates I/O saturation, with all I/O waiting. Depending on locality of reference and on the I/O load, you might want to look at what is generating the I/O load and at improving the host-based caching -- adding physical memory. A performance evaluation based on the performance manual might find a glaring error in the configuration -- overly small working sets forcing excessive paging I/O, for instance. But it is also as likely that the current I/O requirements and system hardware are simply overloaded. Please contact your hardware reseller or sales representative for assistance in configuring the storage upgrade, or (likely better) in replacing the current system hardware configuration with a newer one.
|