by Liem Nguyen, Corporate Communications Manager — February 01, 2010
This is in followup to my post on performance considerations for automated tiered storage. In a recent blog, Martin Glassborow @storagebod posed some good questions about how admins can stay ahead of data loss in a tiered storage environment. He asked for vendor feedback, so this is Compellent’s.
In a Compellent storage system, customers can specify the volumes that get automatically tiered, so you know what applications are tiered and which ones aren’t. A customer may even choose to “lock” an application/volume to a single tier for a specific business reason.
In a tiered environment, or any for that matter, we also recommend clustering controllers for load balancing and failover (the Compellent controllers don’t have to be the same model so you can cluster a 1st generation controller from 2004 with a currently shipping controller).
In case of physical disk failure, our PhoneHome proactive monitoring will alert both the customer and Compellent Copilot team. Often, a new drive is dispatched before the customer is even aware there’s a problem. In the meantime, hot spares can be used for rebuild. Because with thin provisioning we only use up space when data is written, generally rebuild times are faster than for those drives with a bunch of zeros written to them (ie. In SANs using RAID ranks and groups).
In case of data loss, admins can also roll back to locally stored replays (pointer based snapshots) to quickly recover a volume without having to declare a DR scenario. If it’s a physical server issue – say a power supply conks out - customers can use replays in conjunction with boot from SAN to quickly boot and configure a bare-metal standby server with an image centrally stored on the SAN, and mount the volume to the new server. The administration for the entire process can be handled in several ways – via click-through menus from the Storage Center GUI, scripting using PowerShell cmdlets for Windows servers, or the command utility.
If you do have to declare a DR scenario, you can use Enterprise Manager software to initiate failover to replicated volumes stored at the DR SAN site. Enterprise Manager can be accessed remotely and can used to manage replication between multiple sites. Within Compellent’s environment, a production SAN with automated tiered storage can replicate to a secondary site that’s configured as a single or multiple tiers of completely different storage type, RAID level and speed. This lowers the overall cost of replication and DR because you don’t need a copy-exact config between the two sites to do replication. Some customers want the DR SAN to be a full production site for business continuity purposes so may deploy automated tiered storage at the second site too. Then you can revert back to the primary SAN after you’ve resolved whatever issues caused you to declare a disaster. Enterprise Manager also provides dashboards and reports so admins can monitor capacity utilization by volume, RAID level, disk type and business unit, and the status of the tiered or untiered storage at any site.
A lot of buzz is being generated about automated tiered storage but what’s most important is to think about tiered storage as an integrated part of a Fluid Data environment. What other aspects of automated tiered storage are you thinking about?