I met with emerging storage vendor Compellent this week at their brand new eco-friendly headquarters in Eden Prairie, MN, outside of Minneapolis, and I was excited. I have thought of Compellent in terms of SMB to mid-market solutions that leverage industry standard server components and advanced software to create feature rich storage solutions at low cost. The fact that a third of the Compellent customer base uses single controller systems pushed my thinking towards SMB, in spite of advanced features like unlimited snapshots, thin provisioning, and broad functional convergence.
During my visit and demo though, the thing that struck me was the underlying architecture of the storage system and the flexibility that it enables. Compellent storage arrays break apart storage volumes and can place the component blocks anywhere within the system. This means that tiering can happen not at the application or volume level, but at the block level. Their Data Progression software then analyzes the usage pattern of each block and moves them to a cost appropriate tier of storage based on observed performance requirements and user defined policies for each volume. So, blocks of a volume that are used infrequently over time can move to high density, low performance SATA drives, and the blocks that are used more frequently get promoted to high performance, high cost fibre channel drives.
Users of traditional storage arrays struggle with manual application and data classification exercises and have to manually place entire files, volumes, or worse, applications on a single tier, and face significant effort and disruption to adjust that placement in the future. That’s like cutting with a chainsaw compared to a robotic scalpel offered by Compellent. For all the talk about ILM and tiering during the past several years, this is the first system that I’m aware of that can actually deliver automated and effective tiering. Combine that with the ability to apply these benefits across any application, block or file storage, fibre channel, or iSCSI, and all of it inherently thin provisioned, and this is a powerful tool to truly address the management and total cost of operations woes that plague the industry. Scalability from an I/O per second perspective is a concern with this system; it is not designed to handle high intensity enterprise workloads today. That said though, this is a disruptive architecture that could change the game in storage as it grows.
Check out Andrew’s research