Glossary

SAN consolidation

SAN consolidation merges several block storage arrays, and the fabrics that connect servers to them, into fewer and larger systems. Within a broader consolidation it is the host-centric project: each server must be repointed, so the work is measured in hosts and maintenance windows rather than terabytes.

How it differs from other consolidation

Storage consolidation sets the estate-wide goal. NAS consolidation is paced by file counts and permissions. A SAN project is paced by servers: every host that sees a volume has its zoning, paths, boot settings and backup jobs tied to a specific array, and each of those changes when the array goes away.

The sequence of a SAN project

  1. Inventory each array. List the volumes, the hosts that use them, their peak load and the replication and recovery runbooks that name the array.
  2. Remove what never needed block storage. Old SANs accumulate file shares, backup targets, media and archives. These move to file storage or object storage first, so the expensive target holds only databases and virtual machine disks.
  3. Size for combined peaks. Each old array had headroom for its own busy hours. A merged system faces every peak together, such as month-end batch runs on several databases, so the target is sized on measured combined demand.
  4. Migrate host by host. Servers move in waves, with new zoning and volumes, and testing between waves. Thin provisioning on the target avoids carrying old reservations across.
  5. Rewrite replication and recovery. Array-to-array replication pairs and disaster recovery runbooks are rebuilt for the new layout, then a recovery test is run against it.
  6. Retire the old arrays. Only after the last host and the recovery test have moved are the contracts ended and the hardware removed.

What the merged system makes worse

A controller fault or a bad firmware update that once reached one array now reaches every workload on the merged system. Maintenance windows become harder to schedule because more applications share them. The gain is a single pool of spare capacity in place of a reserve on every array, fewer switches and firmware levels to track, and fewer arrays reaching end of support in the same year. Licensing for array-based features such as snapshots and replication often shrinks too, since it is usually sold per system.

A worked view of the reserve

Five arrays each kept about 20 percent free as a buffer against running out. Merged into one storage area network system, a single shared buffer covers all five, and much of that idle capacity can be used or never bought again at refresh.