Compellent's one-trick storage pony is getting company in the next year that will scale up, scale out, dedupe replay blocks, add support for metro-clustering, and introduce a backup/archival storage product.
The company's current Storage Centre platform is an X86 controller, the SC030, running block-access Storage Centre 5.2 software with a stack of disk drives in enclosures underneath it. It can be given a filer personality using either Windows Storage Server or zNAS heads.
The coming SC040 will scale-up by having dual quad-core Nehalem processors, a faster and larger cache, seven PCIe slots, and a 64-bit Storage Centre 5.3 OS that will provide the horsepower for the more complex features and storage personalities that Compellent is developing. The SC040 will ship to early adopters in the third quarter of this year, and represents a huge performance boost that will enable massive scalability and deduplication - although it's perhaps somewhat late, having first been discussed a year ago.
Storage Centre 5.3 will also support the SC020 and SC030 controllers, but not the earlier SC010. Dedupe on the SC020 and SC030 controllers is not a given, as CPU resources are needed for it.
The system will have a 6Gb/sec SAS link to the backend storage enclosures, a switchless link, and host connectivity over Fibre Channel, iSCSI, or FCoE. Data encryption will be supported, and Emulex being mentioned with the suggestion of server HBA-level 256-bit encryption.
Compellent says there will be increased 2.5-inch drive support, with a 900GB, 10K product mentioned as well as a 300GB, 15K product - but neither have yet been announced by disk manufacturers.
Larry Aszman, Compellent's chief technology officer and a co-founder, said the scale-out development will make it possible to cluster Storage Centers beyond today's two-controller configuration. The plurality of nodes will be managed as a single system and some form of load balancing will occur, both using Live Volume technology with synchronous Live Volumes. The cluster interconnect has not been announced, but Ethernet is a good bet as Compellent likes open standards.
Both older and newer Compellent systems would inter-operate in such clusters. Individual nodes in the cluster could be taken down and upgraded without affecting cluster uptime.
Metro-clustering will be supported so that Storage Centre products in data centres in both side of a city could be linked and act as fail-over targets for each other using mirroring and Live Volume technology. The distance supported will depend on the network latency, and the Compellent systems on either side of the link can be dissimilar.
Backup and Archive
Compellent will also add a backup and archive personality to the Storage Centre platform. Currently, users can get data protection with snapshots called Replays.
With Data Instant Replay, once an initial snapshot of a volume is taken, only incremental changes in data need to be captured, and data volumes can be recovered to pretty much any instant in time if snapshots are taken often enough. We expect that replays can be copied by data progression software or some policy-driven mechanism, from the production or primary Storage Centre to the backup and archive one.
Aszman said "We would be envisaging some form of spindown," using drives with that capability, and that "We understand all the properties of data blocks in play. Once a block is old [enough] it could be moved to a tier that can be spun down."
Compellent's co-founding of the Active Archive Alliance (AAA), whereby archive apps present a consistent pool of archive storage including both disk (Compellent) and tape (SpecraLogic), is one aspect of its move into the backup and archive space. Data could be moved, for example, from Compellent storage to a SpectraLogic library.
Archiving software tends to be pretty rich, including ingest, indexing, search, eDiscovery, legal hold, compliance and retention features. What approach would Compellent take to delivering this?
Aszman said: "We could come up with an initial layer of functionality and then refine it in a step-wise way." He also said Compellent could/would work with partners as well. It was noteworthy that CommVault, with its Simpana archive and backup software, was a sponsor of Compellent's C-Drive event.
A CommVault spokesperson said that the company would approach potential membership in the AAA from the point of view of wanting a standard archive format so that data could be moved from one archive to another.
Aszman emphasised that this is "still at the conceptual stage," and Compellent marketing head Bruce Kornfeld talked about three different use cases for the backup and archive product. The first is for it to hold data from production Storage Centres. He said that he could see the product having some 15K drives for the most-recent or needed backup/archive data, and 7200 SATA drives for the older data.
The second case is to have the product mounted as, for example, a CommVault Backup Server or by Symantec or Microsoft's DPM. In other words, it receives data from servers over the LAN.
The third example is to have data replicated from one Compellent backup/archive product to a remote one or to the cloud. Compellent is not developing its own cloud service offering, but is devising products to sell to cloud service builders and suppliers.
Storage Centre software will support compression and dynamic block deduplication. Our understanding is that this will work on replays and run in the background, not working when data is written to disk for the first time. Instead, a background process will examine the multitude of blocks in replays, compress them, detect duplicates, and replace them with pointers.
Aszman said: "The majority of the data in a system will be in a Replay and we know how old it is and how active it is." Deduplication could come into play once a Replay was created. The deduplication will also work on the backup and archive product.
It will be Compellent-developed technology, with patent protection being pursued. Aszman said he had "an epiphany about a month ago" about how deduplication could be accomplished by Compellent.
In general we should expect a 2011 delivery for deduplication. Two wrinkles: volumes can be excluded from having their blocks deduplicated if necessary, and there will be multiple dedupe pools to enhance security. Compellent emphasizes that there will be no data-access performance impact from deduplication activities.
Concerning Compellent's addition of dedupe, John Rollason, NetApp's Product, Alliances and Solutions Marketing guy for EMEA, said: “NetApp … customers [have adopted] our deduplication across over 80,000 systems worldwide … at zero license cost. Customers can eliminate duplicate data across all tiers of storage, including primary storage.
"Deduplication technology is also available on NetApp V-Series [and] can be applied to storage systems manufactured by EMC, HP, Hitachi Data Systems, and others.” That's something Compellent is not offering to users of its 3,000 installed systems.
Compellent offers Portable Volumes - single USB-connected disk drives - as a way of preloading a remote replication target system with data. Apparently its customers want more, so the company is considering offering portable disk shelves with Fibre Channel or SAS links for the same purpose, but with much higher capacity.
Another development is to enable Portable Volumes to be read by Windows or Solaris systems. Kornfeld said: "The product is real. We will do it. It's just a case of which iteration and when."
Compellent said it may introduce the deduplication feature of ZFS into the zNAS product. It also said it had no intention to add object-storage capabilities, seeing no substantial demand among its customer base.
The net effect of these roadmap features, deliverable in the next 12 - 18 months or so, should be that Compellent will get a greater share of its customers' IT spend and have its products appeal to a wider range of customers. The scale-out, deduplication, backup, and archive features should increase its attractiveness to enterprise customers, and bring it into greater competition with all the other suppliers to those businesses.
For Compellent to become an investors' sweetheart again, though, the firm will have to avoid future failures in sales execution as well as deliver on the technologies in its roadmap. That's a multi-month - maybe multi-year - endeavour. ®