Tiered storage buying guide
With storage needs spiraling upward, it's impractical to keep every piece of corporate data on expensive, high-performance Fibre Channel disks. The practice of tiered storage reduces this expense by keeping mission-critical data on the fastest disk platforms, while relegating less valuable data to slower and less-expensive storage such as SATA disk or even tape.
Most organisations will establish several storage tiers. But organizing corporate data into storage tiers is hardly a straightforward process. First, a storage administrator must classify data to
understand the relative value of each file or file type. Data types that are infrequently accessed, such as email, are often indexed and stored with automated archiving tools. Data then has to be migrated between storage platforms so it can be stored according to its classification.
Each chapter in this Buying Guide lists a set of purchasing criteria for a specific product area, as well as product specifications, to help readers identify prospective tiered storage products. Before we look at specific product categories, let's identify the general core concerns.
Establish data retention policies up-front. Data retention is a part of tiered storage. Most organizations are obligated to retain certain data types in order to meet compliance and litigation needs, and then delete that data once its retention period expires.
But retention policies are not set automatically -- no software can tell you how long you should keep a piece of data. To define a retention period, you must know the value of each data type and its relationship to compliance. You can then use data migration and archiving tools to enforce the retention policies across your storage tiers.
Formulate a data categorisation plan in advance. Data classification is not automatic. No software can determine the actual value of data to your organization, so don't rely on data classification tools to make important classification decisions. The data classification process involves input from managers in other departments such as HR, finance and legal. Once you've established the relative value of each data type, then data classification tools can then find the data, apply metadata and enforce established retention policies. Note: The most successful tiered storage deployments occur when categorisation and retention policies are already proven.
Additional storage systems will have a management impact. Tiered storage involves the use of multiple storage systems, which can add to a storage admin's burden. For example, Tier 1 storage may reside on a Fibre Channel SAN, Tier 2 storage may reside on a SATA NAS, and Tier 3 storage may exist on a content-addressed storage (CAS) archiving system, VTL or tape library. The added cost and management overhead imposed by multiple storage tiers can outweigh the benefits. In fact, some large data centers are abandoning tiered storage in favor of a single tier.
Set realistic savings expectations. While the ultimate objective of tiered storage is to save money, the actual savings often fall below expectations. Tier 2 storage should cost 20% to 30% less than Tier 1; Tier 3 storage should cost 50% to 60% less. In actual practice, however, most companies only see a savings of 10% to 15% at Tier 2, and 30% to 40% savings at Tier 3. Below-expectation savings are another reason why some companies revert to a single storage tier. Tiered storage initiatives are sometimes incorporated into other storage consolidation plans.
Buy the tool that fits your needs. You'll find countless migration, archiving and data classification tools to automate your own tiered storage environment, but they differ in complexity, cost and capabilities. Some, like Brocade's Data Migration manager, are relatively simple; others, like Symantec's Enterprise Vault, provide a sophisticated suite of functions. You may select a tool intended for a specific storage platform -- such as using EMC's SRDF tool with their Symmetrix storage platform -- or opt for a more heterogeneous tool such as Incipient's Network Storage Platform. Some tools address one application, such as Mimosa Systems' NearPoint for Microsoft Exchange Server; others support a range of applications. Focus on meeting your minimum performance criteria rather than getting distracted by unnecessary value-added features.
Is there potential for improving storage performance? Tiered storage can enhance storage performance. For example, when all storage is located on a single tier, the contention for user and application access can hurt storage system performance. That contention can be eased by spreading access across multiple storage tiers. Even though SAS or SATA drives offer lower performance than FC drives, reduced competition for access may actually allow for good performance at the SAS or SATA tier. And since this also reduces the number of requests arriving at FC drives, top-tier performance may also improve.
Consider storage systems that support multiple tiers. A popular means of implementing tiered storage is to opt for storage systems that can support multiple tiers within the same array. One example is the Universal Storage Platform V from Hitachi Data Systems, which supports both Fibre Channel and SAS disks in the same box. Not only does this approach promise simpler management, it also reduces the number of storage systems. Tiered arrays are sometimes selected during normal technology refresh cycles.
Evaluate the need for Tier 0 storage. Solid state disk drives (SSDDs) provide unprecedented storage performance for demanding enterprise applications. Offerings such as the 64 GB Universal Solid State Disk from Solid Access Technologies LLC or the 256 GB ZeusIOPS Solid State Drive from STEC Inc. are gaining the attention of storage managers. Some storage professionals are even opting for a solid state tier -- often dubbed "Tier 0" storage -- eliminating the mechanical latencies and speed limitations of traditional hard drives. Users with stringent performance needs may consider adding Tier 0 storage resources.
Weigh the role of storage virtualisation in tiering. Virtualisation adds a layer of software abstraction between the applications and the storage systems. This allows storage resources from multiple storage systems to be pooled and allocated without regard for physical locations -- a great benefit when tiering. With virtualization, a storage administrator can create and allocate a particular tier using storage from across many physical systems, yet present the application with a single storage pool. In most cases, virtualisation improves storage utilisation.
05 Mar 2008