Since Compellent was founded in 2002, it had the advantage of being able to design its Dynamic Block Architecture from the ground up. This architecture manages all data at the level of 2MB granularity, with options for 512KB and 4MB, which requires effective management of the metadata that is associated with the block. The metadata includes such information as what physical disk the data resides on, whether the data is the original customer production copy or a snapshot copy, time-stamping the frequency of access of the block, and what tier of storage the data is on (where the fastest tier is solid state disk (SSD), the next-fastest tier is Fibre Channel or SAS hard disk drives, and the slowest but most cost effective tier is composed of SATA drives). Given this information, policies can be put in place to use a data movement engine to automatically move a 2MB block to the most appropriate tier of storage. Of course, customers have the power to override the system. The default time to evaluate whether or not data should be moved between different storage tiers is 12 days.
This movement of inactive data automatically to different tiers is an instantiation of familiar Hierarchical Storage Management (HSM). The advantage of this approach is that, for example, new e-mails can be kept on a higher storage tier and older e-mails, which over time are a much larger volume, can be moved to a lower tier. This saves money, since lower tiers offer a better cost per unit capacity than higher tiers, without causing any problems to users since the e-mail application can read all e-mails regardless of where they are physically stored.
Now theoretically old data, which follows a long-tail distribution of access, such as a Michael Jackson interview, might become highly active again, but even SATA drives can handle a decent workload and overrides can be put in place in rare situations to accommodate that kind of demand. Special circumstances like these make manual data movement too difficult, time consuming and expensive for companies to consider. That is why companies, such as EMC with FAST, are now moving to automated storage tiering. But Compellent can well claim that with Storage Center automated storage tiering is built in and not added on, as well as being at the sub-LUN level rather than the LUN or volume level.
LegosLand Building Block Approach
Compellent uses a building block approach where the modular components, such as disk controllers and disk enclosures, are backwards compatible, allowing new items to be integrated seamlessly. For example, with the newest version of Storage Center SAS disk shelves (i.e., disk enclosures) are now available but customers can still keep their old Fibre Channel (FC) or SATA disk enclosures. Moreover, the disk enclosures are cost effective, just a bunch of disks (JBODs) solutions to which Compellent adds software RAID for necessary data availability.
Now "religious" arguments can be had about software and hardware RAID, but the pragmatic approach is to go with what works and software RAID seems to work just fine for Compellent and its customers. The company doesn't force a product swap-out when the customer wants to take advantage of the functions and features of new-generation solutions. That not only offers investment protection to the customer, but also simplifies the whole process of upgrades. Compellent rightfully claims that the future is built into its solutions.
This is not to say that Compellent does not offer new things with each release, but features tend to be additive to what is already available. One example is Portable Volume. When setting up a disaster recovery site, the problem is how to move the data from the original local site to the remote site for the first time, a process which requires that all the data be replicated. After that initial process, only incremental changes are needed which can be handled over a communications link, but the initial replication of terabytes or more of data might be too time consuming and expensive over a standard network. Portable Volume is really about copying data to removable disk (i.e., the Portable Volumes) at the local site and then physically moving them to the remote site to cost-effectively facilitate the replication process.
In addition to Portable Volume, SAS drive enclosures that scale from 6 to 384 SAS drives with a choice of 450 GB drives at 15K RPM or 1 TB drives at 7200 RPM. The Dynamic Block Architecture uses storage virtualization so all storage (SAS, SATA, FC, and SSD) can be mixed in a single, virtualized storage pool.
Storage Center 5 also adds RAID 6, which can handle two instead of one nearly simultaneous disk failures before rebuilding completes, but with a twist called Data Progression. Incoming data is initially assigned to a high-level tier with RAID 10 (disk mirroring with striping for performance), and then when the data becomes inactive it is stored on a lower tier using RAID 6. Note that RAID 6 is new to Compellent, but Data Progression has been a capability for some time.
Storage Center 5 also offers some new capabilities, one of which is the ability to manage consistency groups within its Dynamic Replay Architecture. Consistency groups are necessary to be able to do time-consistent recovery of data that has been spread over multiple volumes. Another capability, called Server Mapping, enables the use of rapid thin provisioning of volumes for multiple virtual servers simultaneously. Finally, new Virtual Ports extend virtualization to physical ports for improved bandwidth utilization while still providing high availability.
