Introduction: Solid State Storage

An Industry Commentator’s Delight

Solid state storage seems to be going from ‘zero to hero’ almost overnight. Those steadfast providers that have been promoting some variant of the technology for years may be sighing as everyone else eventually ‘finds religion,’ but they are no doubt also cheering at the prospect of an exponentially growing market opportunity. Unfortunately, as with so many trendy terms in this industry, ‘solid state’ can quickly become semantics without substance. There is an unwritten assumption that meanings are clear and no one wants to look as if they don’t know everything. This paper has been written in order to redress the balance: with few assumptions, it covers what solid state actually is, why it has both IT and business value, and where it is best used. If you believe that solid state is a new development that is synonymous with SSDs, that it is ludicrously expensive, and it is only for the lunatic fringe (or as thumb-drive give-aways used to promote vendors’ other storage devices!), then this paper may well prove to be a worthwhile read.

Solid state storage has been around for decades, but there is currently a renewed focus—and even a certain excitement—regarding the latest technologies that the term now encompasses.  Readers with a historical bent will point out that ‘once again’ might be a better descriptor than ‘currently.’ This time, the degrees of interest, investment, and adoption (at least by the systems manufacturers, if not by enterprise users just yet) suggest that this is a technology for which the time has come. However, it can be hard to determine what is smoke and what is fire—and, as with all new technologies that portend the ‘dawn of a new era,’ the hyperbole can reach intense levels and lead to some bizarre deductions.

Right now, there is more written about solid state than on it. It is an industry commentator’s delight. That will inevitably change; solid state storage is simply too effective, the case for it too logical, and the backing behind it too broad for there to be any other result. While it is that promise that has made it so newsworthy, in a few years we will look back and see what all the fuss was, indeed, about. For now, it is worth the time to look behind the headlines to get clarity on the meaning, value, and potential uses of solid state storage.

Solid State Storage, SSD, and the Potential for Value

First, solid state storage is not just solid state disks. Although Solid State Disk (SSD)—usually flash-based[1]—is the most well-known implementation, it is not the only way for solid state storage to be implemented in data processing environments. And there are indeed also different types and classes of SSD, with widely ranging attributes. However, so glib are some of the discussions around ‘SSD’ that a casual observer might be forgiven for thinking that the acronym stands for a ‘Specific Storage Device!’ The truth is that flash memory enables a wide range of storage devices and solutions, of which SSD per se is only one.

Whatever else is known about solid state storage, some things are clear; yes, it is extremely high-performing, but invariably, today’s offerings are relatively small in overall capacity (when compared to complete data center installations) and are relatively expensive (at least in standard cost-per-capacity terms[2]). Is this a combination that users are looking for?  ESG regularly ‘takes the pulse’ of enterprise storage users; the latest ESG research results[3] indicate that the main challenges users face with respect to storage are:

• Keeping pace with overall storage growth (number 1 of 19 mentioned challenges)
• Storage system costs (number 2 on the same list)

Viewed superficially against this background, the opportunity for solid state storage looks to be a non-starter, with some people relegating it to a high-performance niche; however, and even though it seems counter-intuitive, it is exactly these sorts of mainstream challenges that solid state can help to mitigate and address. Remember that solid state storage devices are not traditional [mechanical] storage products and therefore should not be measured in traditional terms. This is a good motivation to reconsider the role of storage; it may seem obvious (and perhaps that’s why it gets forgotten or overlooked), but there’s no inherent value whatsoever in just storing a bunch of data. All the value is in access and, more precisely, in contributing to positive business goals and outcomes by getting useful work done, both quickly and efficiently. And that is where solid state storage comes into the picture. Used appropriately (more on this later) it provides high performance that can ‘turbo-charge’ applications and underpin a more refined, dynamic, and efficient enterprise storage hierarchy.

Solid State Storage: Origins and Uses

From Past to Present

The idea of non-mechanical storage has a much longer history than a quick glance through recent magazine articles would suggest. The first enterprise solid state device was shipped as far back as 1978, which began a movement that received a major boost in 1984 when ‘flash memory’ arrived on the scene. Unlike earlier media (mainly DRAM), flash is non-volatile, which means that it retains stored data without power having to be continually applied, thus making it suitable as a storage medium. The attractions were obvious—small size, low power, high performance—and the technology was rapidly adopted in a host of consumer devices in order to help drive costs down and capabilities up. In enterprise IT environments, a niche market for solid state has remained over the years for certain extreme and demanding applications. Now, however, solid state storage is a part of just about every vendor’s strategy—and more vendors are actually shipping every month. The reason for the change from ‘propeller headgear’ to ‘general purpose tool’ can be summarized as the interaction of supply and demand factors that combine to make both operational and business sense across a range of product implementations.

Supply Side: Over multiple generations of flash technology, specifications have improved to the point that the technology (today, NAND is the common raw chip) is suitable for general purpose computing in terms of reliability, data accuracy, and capacity. Additionally, the absolute cost of flash media has dropped while the available storage capacities have also increased—partly due to the regular march of progress and partly because the manufacturing plants that produce such products as iPods and USB storage (amongst others) can spread their fixed costs and apply the advantage to high-end products as well. Meanwhile, ‘regular’ HDDs (hard disk drives) have not kept pace with the rate of performance improvement in processing platforms; this is mainly due to disk’s inherent mechanical limitations and has created a wide—and ever-increasing—‘storage-to-server’ performance imbalance; an imbalance that can, of course, only be bridged if a new, higher performance storage layer that better matches the performance needs of those computing platforms is introduced.

Demand Side: The (continuing) growth of online computing and of its associated capabilities (multimedia, for instance) create the need for an ever-increasing ‘instant’ response combined with increasingly sophisticated applications and services behind that response. In order to mask the relative (and expanding) disparity between the demands and capabilities of processing platforms compared to those of traditional mechanical disk subsystems (with their constrained ability to deliver on increasing IO performance demands), users have adopted sophisticated ‘workarounds’ such as short-stroking (whereby only a small percentage of a disk’s capacity is actually used because that saturates its capability to serve IO requests). This is certainly effective, but not at all efficient—and the current focus on more efficient use of resources (whether driven by a ‘green’ consciousness or compelled by the economic downturn) means new approaches are needed.

The Products: Given this growing operational gap, a technology that can bridge it is clearly an attractive proposition. The singular ‘a technology’ in the prior sentence refers to flash, but there are choices in both type and implementation. In terms of the raw media, Single Level Cell (SLC) and Multi Level Cell (MLC) refer to the number of bits that can be stored in a given amount of media. To date, SLC is generally preferred for business use as the applied charge is stronger and less subject to ‘disturbance’ across the cells of the media, which equates to improved data longevity, better reliability, and much improved read/write cycles to the media. However, the technology and implementations are improving so rapidly that MLC is already also being packaged for enterprise use and this is expected to increase as a function of the much larger consumer product market. There are many extended—and also different—media technologies at the R & D or early product development phase. However, the media is only one aspect; how, and as what, it is packaged also varies significantly.

‘Tier,’ ‘Turbo,’ ‘Total:’ The three main usage models are:

1. ‘Solid State Disk’ – in this case, flash memory is packaged into some standard disk form-factor (usually 3.5” and 2.5”) so as to fit into available disk bays in existing disk systems.
2. 2. ‘Flash Array’ – packaged as a high density, standalone solid state storage device targeted to deliver extremely high levels of performance using an extremely small amount of space.
3. Server Component – here, solid state media is built into the server, providing anything from ‘expanded cost-effective memory’ to straightforward storage without the need for a network.

You can recall and refer to these implementations of solid state storage as (respectively) ‘Tier,’ ‘Turbo,’ or ‘Total.’

So, although flash and SSD are not synonymous, they are both about performance. Of course, as the technology moves on, there will inevitably be higher capacities, even higher performance, and lower costs. That is the way of our world. The technology and market are poised to make the leap to where—both soon and for the foreseeable future—high performance needs (IO) will most likely be best addressed by solid state storage and high capacity requirements will be addressed by mechanical devices. Supporting this assertion is ESG’s research: a majority of customers tell us they are either evaluating or have an interest in solid state technologies. What makes this more intriguing is the correlation between SSD adoption and other crucial current IT infrastructure initiatives. As Figure 1 shows, organizations with flash-based SSD adoption are more likely (in some cases, significantly so) than organizations that do not have flash adoption to be pursuing other efficiency and effectiveness initiatives. Put simply, solid state is not marginal technology, but is at the very center of things for data center improvement.

Comparing Flash and HDD: First and foremost, the raw IO performance that flash-based storage can provide is outstanding. Whereas RAM (main memory) response is measured in nanoseconds, that of HDDs is measured in milliseconds (3-4 at best, often 10-20). Solid state NAND flash sits in between the two and ‘fills the performance gap’ with just microsecond response time. To put this into daily ‘human-scale’ terminology, if you imagine that a CPU takes a second to respond, a spinning disk read is months and a flash read is hours. The intractable laws of physics mean that HDD access times are unlikely to improve. In terms of throughput, a typical HDD can hit around 300 read IOPS, whereas for a flash based SSD, the range is in the tens of thousands of IOPS. Furthermore, solid state storage operates all-but-silently, is extremely compact, naturally robust, and requires very little power.

In summary, the main point is that solid state flash can be an effective method of delivering the storage performance needed to bridge the ‘storage to server performance gap’ created by mechanical disk drives that cannot match the IO demands CPUs make of them. Indeed, today’s powerful servers can often be waiting on storage to the extent that even adding CPU power can be precluded from producing improved application performance by an IO and latency bottleneck to the storage system. And the other typical current option— ‘throwing more disk spindles at the problem’ even in association with more DRAM and short-stroking—is costly, inefficient, and not sustainable in a world looking to make optimum use of resources. However, this better approach also means that the metrics used to evaluate ‘value’ need to be updated.

Most users have grown up in a world where the only metrics that count are very straightforward—cost per GB being a prime example. This still tends to be used, even when data centers do things like short-stroking, which means they are really buying (and thus limited to) the IO capacity of that particular drive. And yet the price (and supposed value) will still be inferred from—and expressed as—cost per GB! Clearly, flash memory requires a new set of metrics. If you are buying performance, then performance should be measured. In other words, the measurement should be something like IOPs/$. After all, IOs are the reason to have storage. Given the environmental (a.k.a., OPEX) benefits of flash, users are also beginning to look at IOPS/Watt and IOPS/rack or IOPS/floor-space. And the idea that users cannot look beyond cost per GB is clearly erroneous given that we don’t all run everything on tape; most certainly as disk usurped tape in the data center decades ago, it was not based on the price per MB—it was based on having a more useful performance tool. It’s hard to remember a world where punch cards and tape were significant production tools; years from now, it may be hard to remember when solid state was not.

Themes and Challenges

The motivators which will determine the adoption—or a lack thereof—of solid state storage technologies will obviously vary from user to user. ESG Research posed both the ‘pro’ and ‘con’ questions to determine what users are thinking and the results are shown in Figures 2 and 3.

Figure 2 shows both the most important considerations for users and also the ranking of individual considerations (since respondents could have multiple responses). There is consistency across both sets of rankings; it is no surprise that performance tops the list. Interestingly, the second most important consideration driving the usage of, or interest in, flash-based storage is improved reliability and Mean Time Between Failures (MTBF). This can most likely be explained by the fact that users are indeed looking at flash storage for their key performance applications—and although we have many tools (RAID being the prime one) to overcome most of the negative impacts that result from continuing mechanical disk failures, most of those tools still carry a performance penalty. Integrity and availability is retained at the cost of performance. But solid state removes much of this concern. The more expected factor of power and cooling efficiencies rounds out the top three.

Figure 3 looks at the reasons that a subset of users stated no interest in solid state disk. The leading factor, by a wide margin, is price. This may be true for users that really don’t need high performance or, more accurately, for those that do not feel constrained in this area. However, over time, it is likely that most users will have at least some small percentage of their capacity served by solid state to handle the bulk of their IO. This will happen because the financial logic will become better understood (it’s not just sticker ‘price’) while simultaneously, more and more vendors will be building some percentage of solid state into all their offerings.

As the applicability and ubiquity of flash storage grows, so it will become clear that terms such as ‘constrained budget’ are actually factors in favor of flash rather than against it. The inefficient use of multiple spindles of disk will be seen to be a false economy (in much the same way that buying cheap carpets that wear out quickly can cost more in the long run!).

Rather than perhaps being seen as goods of ostentation for more profligate times, appropriately implemented solid state storage will be seen as a core part of almost any efficient data center. Natural inertia and skepticism in IT may extend this process, but the ultimate success of flash storage—not replacing HDDs, but complementing them—is a result of the simple logic of the storage hierarchy. While the ‘end game’ is clear, the rate of flash capacity increase and flash price decline will jointly determine the speed at which it is achieved. And that in turn is a function of the ‘storage hierarchy’—and its very reason for being.

Why Does the Storage Hierarchy Exist?

The storage hierarchy makes a lovely schematic, but its essence is not about elegance, nor even the desire to have tiers of storage—in reality, it is purely and simply a pragmatic response to the economics that are involved with storage. After all, were it not cost prohibitive, everyone would have everything in main memory all the time. We do not—indeed, cannot—do so because it would be astronomically expensive. And so, everyone makes cost/performance trade-offs about what storage tier makes economic sense for any given user, state-of-the business, or application.

With this understood, it is easier to appreciate solid state storage—not just as a ‘new’ tier, but potentially as a number of tiers in the storage hierarchy. Depending on what form of solid state technology is used (DRAM and Flash being the main current alternatives, although some implementations even combine the two) and where it is placed (as SSD in the storage array, as a ‘cache’ within the controller or network, or even in the server layer) it is clear that solid state storage can have multiple manifestations with the potential to add multiple tiers to the traditional hierarchy (each, as with any tier, offering a unique combination of performance and cost).

As these new tiers are added, and as the IT world learns to measure value via more attributes than has been the ‘norm,’ so the cost and value across the hierarchy will be best if measured differently. Once again, this means not just cost per GB, but cost/IOPS, cost/Watt or /power, cost/foot of space, and so on. Solid state technologies invariably deliver the optimum combination of the highest performance with lowest cost and power use. The only questions are, how much of a user’s data deserves the ‘top of the hierarchy’ treatment and, at the same time, how much can be economically justified?

Key Application Areas and Deployment Models for Solid State Storage

Put simply, there are two main applications areas for solid state memory: any IO intensive application can typically benefit from flash as can any application that demands the fastest possible (non main memory) response time across as much of its capacity as possible (i.e., as can be afforded). Thus, some typical use models might be for high performance computing, web services, e-mail, financial and telecom applications, and database acceleration. But the list is not exclusive—whereverwherever performance is crucial and the volumes can justify the price, solid state storage can provide benefits. As mentioned in the prior section, solid state can be deployed in servers, in arrays/appliances, and in controllers in addition to the ‘straight’ SSD implementation in storage systems. Reports from storage system vendors already shipping solid state products have shown a very general applicability for the technology—matched at times by an intensely detailed desire to know, for example, which precise tables in a database are best placed on solid state.

The optimum deployment model depends on what users are trying to achieve. Clearly, persistent data should go on a flash-based storage device (a ‘tier’) whereas an overall storage performance boost can best be achieved by using flash in a cache/controller/appliance model. Like other storage, solid state can be applied in DAS, SAN, and NAS infrastructures, integrating with and benefiting from the management and policy tools available in each.  Whatever the implementation, one of the key benefits is that there is likely to be an associated improvement in other parts of the hierarchy—in other words, once the major IO activity (which is almost always from a small percentage of a user’s overall capacity) is being served by solid state storage, the remaining, less active capacity can safely reside on less IO capable higher-capacity drives. Put simply, FC and fast HDDs can be replaced by slower, higher capacity, and more economical HDDs. The economic benefit of this move (a significant reduction in TCO derived from using less power, cooling, space, physical disk drives, and even software licenses) may more than compensate for the costs associated with installing solid state storage, even aside from the value of the performance benefits for which it was installed.

In macroeconomic terms, John Maynard Keynes talked about “the paradox of thrift” whereby everyone cutting expenditure actually makes things worse—to borrow that phrase, this use of solid state could represent “the paradox of extravagance” where what looks profligate actually saves money! Recent ESG research[4] has indeed shown that users are very focused on maximizing efficiency and will spend money on tools that help to improve it. This research found that driving OPEX down is far and away the number one consideration for IT when it comes to justifying investments over the next 1-2 years. As we have seen, a well planned and implemented addition of solid state storage can provide precisely these things—improved storage efficiency and reduced OPEX—to the majority of data centers today.

Solid State’s Impact on IT and Business Challenges

Perhaps the best aspect of this section is that it can be stated so succinctly and clearly: solid state storage represents a proven method to address some very pressing IT and business challenges.

IT Challenges: Data center managers are being asked to provide consistent, if not improving, levels of access to growing amounts of data. Doing so with standard tools is problematic because:

• Increases in data volumes far exceed any budget increases.
• The laws of physics dictate that contention only increases as HDD capacities increase.
• HDD performance has improved many times less than that of processors and applications.
• User expectations and the needs of modern online/multi-media services are demanding ever-higher performance and system reliability.

Business Challenges: While IT demands are increasing, so is the pressure on data center managers to meet them at a lower overall cost with improved energy use. Terms such as ‘green’ and ‘efficiency’ are challenging IT professionals to evaluate new approaches and the demand for effectiveness (‘get the job done’) that has driven IT for decades is now joined by a demand for efficiency (‘get the job done at least as well, and preferably better, than before, but do it using as few resources as possible’).

Solid state storage devices actually offer the compelling ability to deliver simultaneously against both IT and business challenges. By using an optimum amount of capacity to serve the highest possible percentage of IO from a device that needs no ‘unnatural acts’ (such as short-stroking or over purchasing of spindles) in order to do so and is highly resource-efficient, solid state storage represents a potential gain in IT performance allied to better business metrics (whether those are derived from the economy, mandated, or moral obligations). As such, its benefits can be enjoyed across a wide spectrum of applications. ESG research in Figure 4 shows that enterprises concur and see broad applicability for the technology, both today (12%) and down the road (61%).

Considerations When Evaluating Solid State

There is some suggestion in the marketplace (from less aware commentators and the minority of vendors that are not yet adopting solid state) that flash-based SSDs, in any manifestation, are only for ‘edgy’ and very specialized use models. ESG research belies this; when asked which applications users believe require the performance boost of flash-based SSDs, most of the top applications mentioned by respondents were extremely standard and were headed by database/OLTP applications. And indeed, in general, any IO-intensive application can benefit from solid state flash, if that is what a user requires. That said, the first question to ask when considering solid state is: what are you trying to achieve?

Just as important, be open to thinking differently and asking multiple straightforward questions, such as:

a)      Do you have performance challenges? Increasing expectations?

b)      Are server and application performance limited by storage latency and IO bottlenecks that compel you to ‘throw’ additional and expensive disk spindles or DRAM at the problem?

c)       Do you know what your energy costs, usages, and objectives are?

d)      Is your current storage infrastructure under-utilizing (a.k.a., wasting) resources?

e)      How are you mitigating that situation now?

And, then, apart from the obvious product and pricing questions, what should you be looking for and asking of any vendors whose solid state you are considering? Given the rapid maturation of the technology, by far the most important area of investigation has to be to discover how any prospective vendor plans to make your solid state storage useful; how they will (indeed, can they) integrate it with your systems and optimize its consumption and usage; how do they (or you?) determine what data gets put on solid state and how long it stays. Are they selling you (to use the earlier vernacular) a tier, a turbo, or a total? This information is crucial. Poorly used solid state is no better than any other poorly used storage—and this is worthy of a little more examination.

Most solid state is all about performance—lots and lots of IOs really fast.  As we have seen, this is useful for both its raw power (for the most demanding applications) and also for the benefits it bestows over the entire storage system (by freeing up under-utilized HDD assets). Irrespective of its other advantages (low noise levels and miserly power, cooling, and space needs), as the relative performance capabilities of HDDs and processors continue to diverge and as the relative capacity capabilities of SSD and HDD continue to converge, so the adoption of solid state storage is assured wherever application performance and response time are concerns. The only other choice is to continue to procure more (and ever-more) badly-utilized HDDs to meet increasing scale and service demands. Easily implemented, solid state storage offers a way to break the vicious circle

However, the significant IO performance that solid state can deliver is potentially limited by the system it operates within. One could imagine a world where all cars are race cars capable of carrying high numbers of passengers—but without the correct infrastructure (capacious roads) to support their capabilities, they would be stuck in the same commuter jams and endless trips to the store as the rest of us. The analogy can be extended; first, you also need to have sufficient performance demands for a race car in the first place (or else a regular sedan or SUV would do); and second, without the need and the infrastructure, all you have is a very expensive car that can cost a lot to run—much of the potential is wasted. Conversely, with both the needs and the infrastructure, then you have a vehicle that can move all the busiest people extremely fast while exhibiting stunning miles-per-gallon results!  Some form of  infrastructure-and-traffic manager (to extend the analogy)—is just as important as the flash-based storage hardware.  Check with any potential solid state supplier that you are considering to find out how it makes solid state ‘usefully used.’ What data gets placed onto its solid state? How and why? What controls do you have? What automation is there? A choice of some manual and some policy-driven automation is what would be best for most users, with an emphasis on the latter being the most effective model for most users most of the time (much as automatic transmission is most useful for most drivers most of the time; but having a stick shift option can be handy occasionally).

Aside from the application suitability, a couple of other implementation considerations deserve mention:

1. Reliability: There is much talk of the ‘write endurance’ limitations of NAND flash. Suffice it to say that there are multiple ‘fixes’ that are already being successfully used to address and overcome this issue. These methods to make solid state ‘enterprise class’ are no different in nature and focus from the sorts of things already done to make existing storage types enterprise class. To suggest otherwise is just to propagate FUD (fear, uncertainty, and doubt) and you will only find it coming from those few vendors with either a limited or no solid state strategy. From ‘wear-leveling’ to deliberate over-provisioning, to ECC and RAID, this is an issue that is being worked around and rapidly becoming relegated to a hang-over perception. Indeed, many current enterprise-class SSD/Flash products have MTBF specifications that exceed those of traditional disk drives.
2. Price: When ESG asked “at what price point relative to HDD technology do you believe your organization would consider increased deployment of flash-based SSDs,” 85% of respondents answered with a statement that they would adopt more at a price increment of anything from 1% to over 50%. A further 10% indicated they would move more at price parity, while the largest number (35%) chose a premium of 11-24% for SSDs compared to HDDs as their ‘tipping point.’ Although a level of pricing much more comparable than currently exhibited between SSD and HDD is eventually possible, the main take-away from the answer is that most users are not yet applying the right metrics to judge ‘price.’ In other words, flash SSD is an IO-centric device and its price should be measured as a function of that; it should not be measured in the traditional terms of capacity. With HDDs, most users ‘abuse’ capacity in order to support the IO traffic they need. What this means, assuming reasonable utilization in either case, is that while the cost per GB of storage remains far cheaper on HDDs and will for the foreseeable future, today, the cost per IO is already far less on flash storage.

Finally, many users want to know about the long term applicability and scalability of any technology decision. Although the roadmaps for solid state storage show order of magnitude growths in performance and capacity (as well as lower prices and the likely removal of the write endurance hurdle) over the coming years, some have a complete offering today. Much as new generation HDDs can slot into existing storage systems, some implementations can implement future solid state iterations as they become available. Remember it’s not one thing—there are SSDs, flash arrays, and  server-based flash storage so that solid state is not just a ‘tier,’ but it is an IO-centric storage technology that can and should be implemented in a number of ways.

The Bottom Line

It is not too strong a statement to say that non-volatile solid state technology has the potential to usher in a new era for storage architectures by delivering extremely high performance and scalability at a far lower per/IO cost. Although common in consumer applications, the solid state storage market is in its early stages for enterprise computing. There is too much operational logic, economic value, and vendor investment behind solid state for it to fail. Indeed, right now, it appears that there’s a ‘perfect storm’ in favor of solid state storage—the technologies themselves have improved[5] and are appearing in a world where HDD advances are slowing, CPU performance is outpacing the ability of those HDDs to deliver their IO needs, the rate of data growth is exceeding the rate of storage price decline, and where the consequent drive to better utilize all resources is bolstered by a desire to be less profligate with the world’s resources (a.k.a., green).

Some commentators say the adoption rate is slow, and there is some truth to this. IT is riddled with understandable and cautious inertia. Users need to learn a different way to compute value (IO rather than pure capacity).  There are no standards. There is misunderstanding (Q: “When do SSDs completely replace HDDs?” A: “They’re not designed to anytime soon!”). Consumption and usage models must be learned. BUT the values— both operationally and in business terms—of solid state technology are so compelling that adoption will occur at an increasing rate. Flash and SSDs will become significant within storage infrastructures—not as a percentage of capacity stored, but as a percentage of IO handled.

Solid state storage is moving from the edge to the mainstream. It is a whole new technology—not one tier, not one product, but a choice of implementations and applications focused (for once and at last) on what really matters to active applications: IO, not storage. The value is already in the products; while this will improve, the fact is that many users could benefit from some level of implementation immediately. Waiting for the ‘right moment’ or the ‘perfect price’ is like trying to apply an absolute point to a relative target—given that the economics of the storage hierarchy make solid state sensible for most users to some extent, the discussion today should be around ‘how much’ and ‘how fast’ rather than ‘whether.’ Both the extent and the speed of adoption are likely to pick up pace over the next 1-3 years (and beyond) as a swathe of quite remarkable products hits the market.

The state of solid state storage? Solid.

[2] Although even this fact /perception (it all depends on to whom you’re speaking!) is poised to change quickly as solid state devices based on the latest, and coming, technologies are expected to demonstrate dramatic increases in capacity at the same time as reductions in cost.

[3] Source: ESG 2008 Enterprise Storage Systems Survey, November 2008. Unless specifically mentioned otherwise, all research statistics and graphics are from this survey.

[4] Source: ESG Research Report, 2009 Data Center Spending Intentions Survey, April 2009.

[5] This largely refers to NAND, but also to the software tiering and migration prerequisites that enable flash to be utilized well.

• Share