HP NonStop G-Series - Manual (Page 53)

Configuring and Managing SMF Processes
HP NonStop Storage Management Foundation User's Guide—523562-008
3-3
Considerations for Pooling

Longer recovery time for a physical disk failure

More overhead when updating disk statistics and in the VDP’s placement
decisions, because every VDP must register the location of every disk in the pool

Larger impact if the pool process fails
Another approach might be to use one pool per disk. But, again, this method has
disadvantages:

The benefits of placement and relocation are lost

Big disk configuration require many more pool processes

Because pools grow one or more disks at a time, pool size would have to be
doubled to increase its capacity
Alternate methods for configuring your pools will provide better solutions than a single
large pool or a disk per pool. Each pool can require a different method. It is critical to
understand the design and performance characteristics of the current or planned
application when deciding which method to use. Some methods to consider are:

One pool per CPU:

Provides performance gain from co-locating the DP2 processes, the pool
process, and the VDPs in the same CPU

Limits impact of a CPU outage to a single pool

Provides natural and easy to remember way to subdivide a large disk operation

Pooling across multiple CPUs:

Provides greater parallelism

Allows relocation to balance CPU load

Is a good strategy when the disk per CPU ratio is low

Pooling by disk attribute (for example, mirror and audited):

Simplifies placement of files with special requirements

Pooling for disk space considerations:

Provides that any VDP can use up all the free space in its pool

Limits the impact any one VDP can have on overall free space

Provides that the disk full threshold is set at the pool level

Pooling by group or application:

Provides that a group or application “owns” the pool

Limits impact of physical disk outage

Simplifies space growth planning