3PAR (NYSE Arca: PAR), the leading global provider of utility storage, today unveiled a five-point checklist to help organizations evaluate the most effective thin provisioning technologies for their storage infrastructure.

Thin provisioning allows IT organizations to safely over-allocate physical capacity while consuming physical capacity as applications truly require it on an as-needed basis. Those taking advantage of 3PAR's thin technologies have been able to deploy just 1 terabyte of capacity where 2.5 terabytes of traditional capacity had been required. Each terabyte of disk drives consumes about 1,500 KW-Hours of electricity for power and cooling and produces a metric ton of CO2 emissions each year. The net green effect of thin technologies is a reduction of the energy and carbon footprint from data center storage by approximately 60%.

However, not all thin provisioning technologies are created equal. Identifying which thin provisioning technology to implement may not always be as straightforward as some storage vendors claim. In fact, there are several pitfalls of which organizations need to be aware should they fail to evaluate the available thin provisioning technologies carefully. As a pioneer of thin provisioning, 3PAR has developed a user-friendly 5-point checklist to help organizations avoid these pitfalls in achieving virtualized storage environments that offer both technical and financial advantages.

"In most cases, thin provisioning is excellent for addressing organizations' needs to ease data management and free up both human and financial resources," said Claus Egge, Research director, EMEA storage systems research at IDC. "It does not come as a surprise that as thin provisioning has started to appear on the must-have list for more and more datacenter virtualization projects, more and more vendors have looked to add thin applications to their portfolios. Consequently, it has become critical that organizations evaluate these thin applications on the strength of their underlying technology offerings and not just be lured in by the buzzwords."

At the core of 3PAR's thin provisioning checklist are several key criteria that organizations should consider while evaluating available thin provisioning technologies:

Pitfall #1: Large Allocation Unit Size
As data is written to thin provisioned volumes, different implementations of thin provisioning will consume varying amounts of capacity to accommodate those writes. Where the unit of consumption is much greater than the size of the write, the efficiencies of thin provisioning are diminished, and can be lost altogether. With some thin provisioning implementations, scores of megabytes are dedicated upon even the smallest of writes (e.g. 8 KB is a common write block size). A multi-megabyte, coarse unit of space allocation can mean that the simple creation of a file system on a thin provisioned volume can completely fill the thin provisioned volume, eliminating any value of thin provisioning before the first piece of application data is ever written to the file. With coarse allocation, users must concern themselves intimately with the behavior of every file system, operating system, and application they may wish to support with thin provisioning.

On the other hand, some implementations allow fine-grained allocation where capacity is dedicated in kilobytes, not megabytes. For example, 3PAR Thin Provisioning dedicates capacity in just 16KB increments, which is more granular than the increments of other recently-introduced thin provisioning technologies. Capacity savings are maximized, and thin provisioning can be applied broadly and easily to host operating systems, files systems, and applications.

Pitfall #2: Reservation-Based Implementations
With reserved dedicate-on-write implementations, physical capacity is pre-configured for and reserved up-front into specific pools, a tremendous waste; committed silos of capacity are always the enemy of efficiency. For example, in these “chubby” provisioning implementations users must size, configure, and pre-dedicate disk drives, RAID groups, and numerous other logical definitions. This must be done separately, and often irreversibly, for every data service level they intend to provide (RAID level, disk type). Separate pools may be required for snapshots and other functions as well.

Other implementations, including 3PAR's, are reservationless dedicate-on-write approaches where capacity is drawn and configured in fine increments from a single free space buffer with no significant pre-dedication of any kind. Multiple service levels are served from the same free space buffer because service-level configuration occurs on-demand based on write volume. When dedication is fluid, there are no reservations for any volume types or service levels and users must concern themselves only with replenishing a single highly leveraged free space buffer. Risk is reduced and efficiency flourishes.

Pitfall #3: Manual Provisioning
In traditional or "fat" storage environments, the reason capacity is dramatically over-provisioned (usually by three times or more) is to avoid the time, complexity and disruptions associated with manual re-provisioning activities. Unfortunately, manual thin or chubby provisioning implementations retain much of the original complexity, and can even add to it. In many implementations, the on-going reservation and configuration of replenishment capacity into specific pools represents incremental, manual complexity. The planning, decisions, and configuration must be completed on a pool by pool basis. In situations where decisions cannot be easily undone, pool provisioning may be done quite conservatively and therefore quite often, severely mitigating the overall benefit of thin provisioning. Lengthy whitepapers challenge users to understand how to manually set up and monitor myriads of data structures and related reservation settings over time. In cases where thin provisioning pools are trapped behind certain controllers, current and future load balancing concerns must be factored in. Bottom line: when pool replenishment is a manual activity, users will forego thin provisioning or will compensate by over-replenishing - materially decreasing capacity utilization and efficiency.

Conversely, when thin provisioning is implemented autonomically, capacity is dedicated and configured naturally, without human intervention and just-in-time. User time and effort to plan and configure storage are eliminated, and so are any compensating inefficiencies.

Pitfall #4: Dual-Controller Architectures
Thin provisioning is inherently about making capacity promises that may have to be fulfilled in the future. Hence the ultimate scalability of an array platform to meet the capacity and performance needs of applications were they to fully utilize thin provisioned capacity, is of utmost concern. Dual controller architectures, which usually operate on an active-passive basis (by volume), are disadvantaged in this respect. Prudent planning and application safety suggest thin provisioning should be employed conservatively with such architectures, since in such environments liberal use of thin provisioning is an invitation for abundant and disruptive data migration activities.

Conversely, highly scalable systems are ideal for thin provisioning since they are already architected for storage consolidation and growth within a single system. Examples include monolithic systems and highly virtualized clustered modular systems such as 3PAR's.

Pitfall #5: Bolt-on Virtualization
In some cases, thin provisioning is an added feature to a pre-existing hardware and software architecture. The special demands that thin provisioning can place on an array that is not specifically architected for it can mean unhappy tradeoffs in other areas and functionality. For example users may expect diminished performance, inability to replicate thin provisioned volumes in thin or snapshot form, and an inability to thin provision all available physical capacity. In other cases, traditional reporting capabilities are inadequate, imposing a heavy pro-active monitoring burden on administrators. If users must sacrifice other features, functionality, and performance to gain Thin Provisioning, then overall value is diminished.

Meanwhile, other hardware and software architectures were built from the ground up to support thin provisioning. In 3PAR's case, Thin Provisioning is fully integrated with all other functionalities like Dynamic Optimization (for changing service levels on the fly) and all copy technologies (snapshot, clone, and remote). As evidenced by Oracle testing, performance with or without thin provisioning remains high and predictable. 3PAR Thin Provisioning is fully instrumented for safe, confident operation with minimal requirements for pro-active monitoring.

"Thin provisioning can both save you money and promote environmental responsibility at the same time," said David Scott, President and CEO at 3PAR. "But watch the pitfalls.  Without close attention, your company can be fooled into buying ‘chubby’ implementations masquerading as thin provisioning.  Apply this checklist and make sure you get the benefits you were expecting.”

Thanks to John D'Avolio for the news!