Compellent Storage Center Baldwin Park CA

Compellent's Storage Center aims to take the strain out of ILM (information lifecycle management) by delivering a storage solution that automates data classification. One of the biggest problems with an ILM policy is managing the migration of data throughout an organisation and creating an automation phase to determine where it should physically reside and when it should be archived.

Compellent's Storage Center aims to take over this phase. The units data progression feature automatically migrates data from high-cost, high-performance fibre channel (FC) to low-cost, high capacity SATA storage as the popularity of the former declines.

Up to three storage tiers are employed, with support for both FC and IP SANs and it offers an excellent range of virtualisation features. The really smart bit about the data progression is that not only does it work at the block level and not the file level, but it's a two-way process so older data that is suddenly called up for active duty will be promoted up through the tiers.

On review we have Compellent's ILM bundle, which comprises a controller and two sixteen bay disk arrays. The new Series 30 controller is based on a Supermicro 3U rack chassis equipped with dual port and quad port 4Gbps QLogic FC cards, which provide connectivity to the disk arrays and host systems. The disk array providing the higher tier storage came with eight 10K FC drives, whilst the other was kitted out with eight SATA drives. For IP SAN testing the review system also came with an optional dual-port QLogic iSCSI HBA.

Installation is straightforward and we found the redesigned web interface a big improvement over its predecessor. It's much easier to use thanks to a new quick start wizard to get you up and running.

Compellent's DBA (dynamic block architecture) technology is a key feature, which uses metadata to track all data blocks. It comes into play during this phase as you need to create Disk Folders, which can use all available drives or selected drives and you can mix FC and SATA as well. Each Disk Folder contains virtual volumes, which you then assign to your FC or IP SANs.

Each Disk Folder has up to three storage tiers, which Compellent's data progression feature uses to migrate data from FC to SATA storage as it becomes less active. The first tier contains 15K FC drives, the second has 10K FC drives, whilst the third contains all the SATA drives and the system automatically places disks in specific tiers based on their spindle speed and interface.

We created a single Disk Folder with two active tiers and within this we set up two volumes for our iSCSI hosts and another for our FC test host. Usefully, only two settings are used to determine data migration as you set the number of days a block must be unused for before it is moved to a lower tier and a ratio that decides how quickly data is moved up and down the tiers.
[pb/]
A new feature is Compellent's Fast Track, which recognises that the outside of a disk platter is faster than the inside and organises data accordingly. Thin provisioning is also on the menu as virtual volumes can be much larger than the amount of available physical storage. You can choose from MB, TB and PB for each virtual volume and DBA allocates blocks dynamically as the volume grows in size.

Compellent's replays are smarter than snapshots as they use a lot less space. DBA is used to determine where blocks were located when the replay was taken so they only increase in size when new data is written to a volume. Under normal circumstances replays will be placed on the third tier where storage costs are the lowest but you can have them on higher tiers if you wish and use data progression to migrate them down when they are unused.

As you'd expect with an enterprise storage solution, replication is a standard feature. A number of options are available as you can replicate between remote sites using either synchronous or asynchronous operations. Furthermore, it's up to you what you want to replicate as you can pick and choose from volumes and replays.

For testing we introduced the Storage Center to the lab's Gigabit network and 4Gbps FC SAN. For iSCSI testing we used quad-core Xeon 5400 Dell PowerEdge 1950 and 2900 servers running Windows Server 2003 R2 and Microsoft's iSCSI initiator. With one server logged onto a dedicated volume we saw the Iometer utility report a raw read throughput of 113MB/sec - near maximum for iSCSI over Gigabit.

With a second server logged onto the other iSCSI port we saw a cumulative throughput of 198MB/sec, showing some contention for disk resources was occurring. We tested FC performance using a Boston Supermicro dual 3GHz 5160 server equipped with an LSI Logic 4Gbps dual-port HBA. Using the same parameters as for the iSCSI test Iometer reported a good raw throughput of 350MB/sec for a single direct-attach connection.

Overall performance is good but Compellent will need to make the switch from FC to SAS drives for Tier 1 storage as they deliver equally good performance for a lot less cash and can be mixed with SATA drives in the same disk shelf. We also raised the subject of support for 2.5in SFF drives as they are more environmentally friendly and were advised this is currently on the cards as a future release.

The concept of hierarchical storage management has been around for a long time but Compellent's Storage Manager is one of few solutions currently available that is capable of managing and automating the data classification and migration processes effectively. We found it particularly easy to deploy with extremely good expansion potential and the DBA technology makes it quite unique.

Author: Dave Mitchell

Compellent Storage Center