10 questions to ask when purchasing an automated tiered storage solution

When Compellent first introduced automated tiering, few people anticipated the impact it would have on the industry, revolutionizing high capacity storage by making it more efficient and affordable.   Now fully integrated into Dell's SC Series arrays, Compellent auto-tiering has been enhanced and accelerated by nearly a decade of development and iteration across multiple Dell product generations.

Although other vendors have since claimed varying degrees of tiering capability, delivering some real benefits in specific areas, Dell remains the gold standard, with unrivaled solution breadth and refinement.

In this post, I’ll give you a detailed checklist of must-have features to knock off when considering an auto-tiering array.  Each has tangible business value – so if you really want to enhance datacenter efficiency and reduce costs, be sure you can answer yes to every question.  Competing solutions may offer one or more of these advantages, but only Dell nails them all. 
 

  1. Can the solution you are considering automatically tier multiple layers of flash? 
    Flash provides a fresh catalyst for tiering benefits by dramatically increasing the selection of drive capabilities and price points available for building arrays.  Tiering harvests advantage from diversity – and applied to flash, can synthesize the best attributes of diverse new SSDs into a balanced, unified system.  Multi-tier all-flash arrays (AFAs) perform like high-end single-tier systems, but at a much lower cost.
     
     Yet…few vendors can actually support multiple flash layers.  Most traditional storage players can’t use the super-affordable “read-intensive” SSD drives, and many new flash-centric startups can’t tier at all, offering only single-tier systems at a very high solution cost.  Not surprisingly, these vendors ignore or downplay the benefits of tiering.  Only Dell’s SC Series combines two of the most significant storage advancements of the past decade (tiering and flash) to produce a multi-layer AFA that is uniquely performant and cost-effective.
     
  2. Does it also allow hybrid configurations – saving additional cost by automatically moving cooling flash data to inexpensive spinning disks?
    Despite the claims of “flash-only” vendors to cost-effectively store all data (hot and cold) on flash, spinning disk remains up to 20x cheaper for bulk storage.  IT managers, well aware of the tradeoffs, are now struggling to determine how much of their data they can afford to accelerate with flash.
     
    Dell’s answer is simple: you should accelerate all your hot data – i.e., all the data your users actually interact with – while letting low-cost bulk storage do what it does best for data that does not impact workload performance.  The same optimization methods that make Dell’s multi-tier AFAs so cost-effective – combined with the ability to move cold data entirely off flash – make Dell’s advanced hybrid solutions the preferred choice for a majority of customers.   Whether deployed as a multi-layer AFA, or a cost-saving hybrid, SC Series’ exclusive flash tiering helps SSD investments pay off quickly.  This was recently confirmed by IT Brand Pulse 2014 SSD Brand Leaders survey, which named Dell as the overall Price Leader for both hybrid SSD/HDD arrays and all-flash Fibre Channel, iSCSI and NAS arrays.
     
  3. Does this tiering method move sub-LUN data in “right-sized” chunks?
    Dell’s Data Progression algorithms optimize data placement within each tier and across the entire array, moving data at a granularity ranging from 256KB to 4MB.  This block size is considered the “sweet spot” for tiering by Gartner, in contrast with other solutions which use 100MB or even 1GB blocks.
      
  4.  Does it actively track and anticipate data usage over an extended time period?
     If the answer is no, chances are the solution you’re looking at is not auto-tiering at all – but simple caching.  “First in, first out” or “least recently used” caching is less dependable than tiering, and is vulnerable to becoming de-optimized quickly.  For instance, running a different workload on the weekend (a common practice) can cause the data needed for core user applications to not be in SSD cache on Monday morning.  Although cache is designed to adjust rapidly, there is no pre-fetch of cold data, which must be restored one block at a time, limited by the speed of the lower tier.  To avoid user complaints in this scenario, you would need to size your spinning disk performance to cover the gap, negating much of the cost-savings you were trying to achieve in the first place.

    True tiering keeps a “heat history” of all the parts of the volume, maintaining user happiness (even on Mondays) by positioning both frequently accessed and “likely to be accessed” data in the top performance tier, while purging your most expensive drives of data that will not be accessed.
     

  5. Can it automatically provision multiple RAID levels per tier, volume…and drive?
    Competing solutions do not even support multiple RAID levels per drive, let alone configure and manage them for you.  These solutions require you to segregate and lock in RAID levels when drives are first installed, invariably holding large amounts of storage “captive,” which leaves money (and/or performance) on the table.

    Dell SC Series arrays present a fully-virtualized storage pool, configuring and mixing RAID levels on the fly across all drives to treat reads and writes differently.  RAID 10 is used for new writes, while blocks not being updated are converted to read only at parity RAID 5 or 6.  You get the performance benefits of RAID 10, plus the capacity benefits of RAID 5/6.  RAID acts as a flexible “sub-tier” for Dell arrays, optimizing data placement within the main tiers and creating powerful efficiency advantages before other vendors’ auto-tiering even kicks in.
     

  6. Are snapshots integrated with tiering?
    Only Dell’s space-efficient snapshots (called Replays) are integrated with tiering to minimize cost.  Unlike less efficient “copy on write” solutions, which require old data to be moved before a snapshot is written, Dell Replays simply freeze snapshots in place, making them read-only, but otherwise allowing normal use to continue.  Data Progression then intelligently migrates the snapshots to parity RAID or less expensive disks.  This method requires fewer disk operations overall, and frees up more Tier 1 capacity for active data.
     
  7. Can you move data into the host for further acceleration while maintaining full end-to-end snapshot capability?
    Dell’s Fluid Cache for SAN (FCFS) is a shared server-side flash solution that uniquely accelerates both writes and reads, increasing application response times by as much as 99% with full data protection and no single point of failure.  SC Series arrays can be configured to promote hot data into FCFS, so mission-critical IO occurs without waiting on external storage.  A single interface controls both the SAN and the server cache, while “cache-aware” SC Series snapshots ensure the entire range of application data is secure, no matter where it is in the environment.
     
  8. Can you tier within a disk to reduce seek time while using the entire disk’s capacity?
    Dell’s FastTrack feature treats “location on disk” as a separately controlled tier, concentrating hotter data on the outer, faster tracks.  Unlike other tiering solutions which actually recommend keeping disks partially empty to avoid performance degradation, Dell leverages the inner tracks for cold data, giving you the benefit of “short stroking” without wasting capacity.
     
  9. Is tiering preserved and extended non-disruptively when adding/removing disks?
    For some solutions, removing a single disk from the virtual provisioning pool means first removing all data from the entire pool, effectively destroying it.  Be sure the array you choose encourages incremental investment by allowing you to smoothly retire old disks.
     
  10. Can the tiering be easily customized to meet your unique needs?
    Dell arrays feature built-in, automated best practices leveraging almost two decades of storage experience – but they are also highly configurable.  Changes to tiering profiles, snapshot schedules, and many other features are made easily on the live system, with no interruption to applications.  Can your solution “pin” volumes temporarily to an upper tier, or write an entire volume directly to RAID 5/6 when loading a new database for the first time?  With Dell Storage SC Series, these methods may be scripted to repeat at regular intervals along with standard Data Progression, automating optimization for infrequent but recurring events such as monthly or year-end reports.
     

*********

Automated tiering has become one of the fundamental pillars for modern storage systems – and something of an industry buzzword.   But the phrase has different meanings for different vendors, and not all implementations are equal.  How did the solution you are considering fare on the list above?  To be sure you get full return on your storage investment, purchase your array from Dell, the company that invented automated tiering and continues to evolve it for tomorrow’s needs.

About the Author: Marty Holmes