Compellent micro-manages data to gain efficiency
Sixteen times increase in tiering granularity
CeBIT By micro-managing data with its coming 64-bit software, Compellent is aiming to increase its automatic data placement tiering granularity sixteenfold and greatly increase SSD use efficiency.
Currently its Storage Centre O/S is 32-bit software and it automatically moves 512KB pages of data between faster and slower tiers of storage based on its activity level. Hot data moves up the tiers to solid state drives (SSD) and inactive data moves down the tiers to bulk SATA drive storage.
The number of pages it can track is limited by its 32-bit addressing scheme. In the next few weeks the O/S will change to a 64-bit version with consequent much larger addressing possible. Compellent also has a relative excess of processor cores in its Storage Center controller.
Talking at CeBIT Steven Dahlin, Compellent's EMEA sales director, said the 3U Storage Center controller box has dual quad-core processors and: "We're only using one core per controller right now."
The core usage count will go up when the synchronous version of LIve Volume is introduced, software that will replicate changes to volumes in one Compellent array to another. It will increase again with an sixteenfold increase in the granularity of which automatic data placement is carried out.
Dahlin said: "We have the most granularity in the market – 500 times – at 512KB. With a 64-bit OS our page size can become more granular; 32K, why not? We just need the resource to handle that many pages ... Having 64-bit is like turning on a turbo charger."
This would mean a huge increase in the number of data blocks tracked and, hopefully, less data moved up and down the storage tiers as not all data in the current 512KB pages is active. There should be a great increase in the efficient usage of the fastest and most expensive storage as a result.
Dahlin also suggested that Compellent's coming in-house deduplication could also use 32KB blocks. He confirmed that Dell is productising its Exanet scale-out NAS technology. This would then be a natural fit with Compellent storage. Currently there is a Nexenta ZFS-based NAS head available, providing file access to the Storage Center's block-based storage.
We can envisage an Exanet-based scale-out NAS head or the integration of Exagrid technology directly into the Storage Center O/S, although such retrofit integration would be more difficult than a straightforward layering exercise. Our impression is that Compellent is heading full tilt deeper into the data centre with the base hardware becoming clustered pairs of Storage Center arrays federated into greater storage units with Live Volume, and providing unified file and block storage with the Exanet technology. ®