VERITAS Volume Manager 3.5 User's Guide - VERITAS Enterprise Administrator (June 2002)

Load Balancing
30 VERITAS Volume Manager User’s Guide - VEA
Statistics are accumulated from when they were reset(usually they are resetwhen
the system boots/reboots, or at object creation, whichever is most recent).
- View > Compare on Refresh
Each statistic request is compared with the last one to determine where the
current high I/O activity is occurring.
Monitoring Performance
You can change the criteria that determine which disks are displayed as slow performers.
Select View > Perf. Monitoring Parameters and complete the Performance Monitoring
Parameters dialog box. This dialog box allows you to configure how many disks are
shown as slowest and next slowest performers, the sampling interval, and whether to
show the slowest performers for reading, for writing, or for both reading and writing.
Load Balancing
If disk activities are heavily concentrated on one or a small number of disks in the storage
subsystem, it maycreatebottlenecks. You canuse theMoving a Subdisk and possiblythe
Splitting a Subdisk features to spread out disk accesses more evenly across all the disks
to balance the load.
If a disk has High or Critical I/O activity (shown by a red or yellow pie symbol), you may
consider moving one or more of its subdisks to another disk that shows below average
I/O activity (shown by a blue pie symbol). The idea is to move just enough activity to
achieve balance. A careful study of the statistics for the disk with Critical activity may
identify the best subdisks to move. You should move subdisks only when a disk has High
or Critical I/O activity over a prolonged period of time and performance is affected.
Moving a subdisk to another disk has an effect on I/O as well, but it should be
compensated for by the other disk having much lower I/O activity. You would need to
look at the statistics after the subdisk move to see whether the move was effective in
balancing the load.
Disk Failure or Removal
Moving a subdisk for redundant volumes (mirrored or RAID-5) will use the redundant
data to recreate the subdisk on the healthy disk. However, for nonredundant volumes
(concatenated or striped), the data cannot be recreated and doing subdisk move will
therefore lose data, which could be recovered if the disk can be repaired. Thus, when you
attempt to move a subdisk from a failed or missing disk that has nonredundant data, a
dialog box comes up that asks you if you want to force the move. You may want to force