Rollout: Compellent's Storage Center 3.5

Storage vendors have been calling their entry-level and midrange SAN disk arrays modular for several years. But Compellent Technology's Storage Center 3.5 fits the bill better than any of the competition with its winning feature set and flexible storage options.

Rather than spend R&D dollars on developing hardware disk controllers and enclosures, Compellent uses industry-standard servers as disk controllers and OEM supplier Xyratex's SBOD (switched bunch of disks) enclosures that incorporate Fibre Channel (FC) switching components and JBOD (just a bunch of disks) for SATA drives.

Standard, But Outstanding

Although using standard components can result in a plain-vanilla array system, Compellent's storage system is anything but. Unlike those on more conventional systems, such as the significantly more expensive EMC Clariion CX3 or Hitachi Data Systems (HDS) Thunder 9500 series, the Storage Center's ports are not dedicated. Each of the four available slots on the server-based controllers can hold a four-port FC or a one-port Gigabit Ethernet HBA (host bus adapter) for iSCSI. All the ports can be used for server connections or replication, and the FC ports can be used for back-end connections to disk enclosures. Storage Centers can support up to 784 drives--significantly more than the competitors--and the controllers can be configured into a two-node active-active cluster.The real difference between Storage Center and conventional arrays is software. It starts with how logical volumes are built. With most arrays, a RAID set would be created from a group of similar drives and slices of that RAID set would be allocated into volumes servers can access. Rather than build a RAID set from a group of drives, Storage Center places data on drives in individual stripes. In RAID 10 arrays, data is written to mirrored stripes on two drives. For RAID 5, data is written to stripes on four or eight drives and parity info goes to an additional drive. The next stripe is written similarly, but not necessarily to the same drives. This striping RAID model boosts performance and enables two of Storage Center's advanced features: thin provisioning and data progression.

Thin provisioning helps avoid wasting space, by using storage space only as an app needs it. This feature has been available only on high-end arrays from 3Par and on NAS-SAN hybrids. With thin provisioning, managers allocate space to each app, but the logical volumes don't use physical disk space until the host server actually writes data to a given block. If you do over-allocate, Storage Center sends e-mail alerts when it starts to run short of space.

Data progression is Storage Center's method of keeping the most frequently accessed data stripes front and center, and migrating older data to lower-performance drives. By keeping track of such metadata as the frequency and last access dates for each stripe, Storage Center can migrate less-used data automatically from RAID 10 drives to RAID 5 on SATA or nearline drives, SATA-like drives with an FC interface that can be mixed and matched with high performance drives in SBOD enclosures. We set up a test volume to store primary data on high performance FC drives and snapshots on nearline drives.

Compellent's implementation of replication and snapshots, which Compellent calls replays, are also standouts, thanks to Storage Center's well-detailed templates. With almost no effort, we set up and applied a replay template that would take and retain snapshots at four different frequencies over separate time periods.Storage Center's volume replication is rich. Besides supporting the standard synchronous, asynchronous and snapshot replication, Storage Center has unique twists. Most arrays can replicate to their doppelgangers, then take snapshots on the target. Storage Center takes the process one step further by replicating data and the snapshot's metadata, to guarantee a snapshot on the target represents the same point in time as that on the source.

Sophisticated Wan Replication

QoS templates make Storage Center's WAN replication the most sophisticated we've seen. We set up a QoS template that throttled network utilization during the day while we replicated less critical volumes. If snapshot replication falls behind due to limited bandwidth, Storage Center can eliminate multiple changes to the same disk blocks in subsequent replays, called thin replication. Bandwidth utilization is monitored through the Enterprise Manager. You can even set up replication to a virtual remote array and monitor bandwidth utilization before you replicate.

All in all Storage Center is a winner: 4-Gbps FC performance, iSCSI connections, thin provisioning, flexible replication and a competitive price.

Howard Marks is founder and chief scientist at Networks Are Our Lives, a network design and consulting firm in Hoboken, N.J. Write to him at [email protected]. 0