Apple Xsan 1.x - Manual (Page 33)

Chapter 3 Deployment Examples 33
Planning Decisions
The following paragraphs show how the planning questions in Chapter 2 are answered
for this small computational cluster example.
How much storage is needed?
Three Xserve RAID systems, one for metadata and the 1.8 TB of input data and two
others for up to 8 TB of results, are adequate.
What organization of storage makes the most sense for user workflow?
The “user” in this example is an application processing a large data set, so no special
organization is needed to help humans navigate or manage the storage hierarchy.
Root-level access to the large data sets is adequate.
What levels of performance do users require?
Performance is an important goal in this example. So, all of the computational nodes
are attached directly to the Fibre Channel network.
How important is availability?
Ensured availability of the file system during long, unattended computational sessions
is important, so we’ll include a standby controller.
What are the requirements for security?
Security is satisfied by the fact that the SAN is physically restricted to the cluster itself.
There is no access from outside.
What RAID schemes should be used for the RAID arrays?
Xserve RAID systems are set up to provide RAID 5 arrays for a balance of performance
and recoverability.
How many volumes are needed?
This example uses a single volume with two affinities, one for input data and the other
for results.
Which storage pools make up each volume?
For this example, the single volume consists of three storage pools:
 1 pool for metadata
 1 pool with affinity for source data sets
 1 pool with affinity for results
How are individual volumes organized?
Because the volumes are used by an application and not by human users, we’ll store
the data sets at the root level of the volumes.