At the Flash Memory Summit (FMS) this past August, Darren Thomas, VP of Micron’s Storage Business Unit, gave a presentation about the state of flash usage in the storage market — issuing a call-to-action to the industry. Darren stated that in order to “cross the chasm” to greater user adoption, flash vendors have to bring more value beyond performance to the customer. He said that it’s time to move from engineer-to-engineer speak (AKA, “speeds, feeds, slots and watts”) and start talking about the real-world customer pain points that flash technology solves. Watch a brief overview of his presentation.
Now, coming off of the Gartner Symposium/ITxpo and having talked to several CIOs at the event, I’ve developed a broader perspective of what this change looks like. Because, like all things in technology, change takes place over time, I came up with an analogy to waves. So in the spirit of waves, let’s dive right in!
Wave 1 – Need for Speed
In defining the transition to flash, many companies are still firmly positioned in what I’m calling Wave 1. In Wave 1, flash is used only for the most demanding applications where speed is so desirable that any cost is justified to improve application performance. This is perfectly logical, as flash currently sits at about 5 percent of worldwide storage deployments (in terms of gigabytes). In Wave 1, customers compare flash to its substitute products, like hard disk drives (HDDs), and selectively look to replace them as the situation justifies.
Wave 2 – Enhanced Experience
A few companies have caught the second wave of flash – let’s call this Wave 2. Wave 2 is more about the experience that flash brings to applications, not just the performance. Maybe the easiest way to explain this is by comparing the data center to what’s already happened in the corporate client space. Take the humble laptop: It started by using flash to boot faster. But soon, battery life and storage size became further justifications for flash — to the point where the enhanced user experience that flash brings to the mobile world is now quite obvious.
This enhanced user experience with flash is in its infancy in data centers, but this data center experience will be the mechanism that takes flash from 5 percent to more like 15 percent in terms of adoption. Here’s why: With flash, you can avoid or defer new raised floor space because at scale, solid state drives (SSDs) can save power, run hotter (less cooling) and significantly reduce the number of nodes in a cluster (simplified management and reduced space). Flash will be a cornerstone of the experience that drives large-scale IT further and faster than it can go today. Cost will matter, of course, but the savings at the system and even data center level will make up for any single component cost premiums.
Here’s another way to think about it: As servers become more-disposable elements of a larger solution, the reliability, capacity and power savings will drive thousands of servers to have a five-year operating life, saving companies millions. Does this still make SSDs look more expensive than spinning media in the right applications? No. The experience has changed.
Wave 3 – Pure Application Necessity
Finally, we have Wave 3. Wave 3 is not yet upon us; it’s where all of the technologies come together in just the right mix to make flash both practical and possible. But we can see this new wave forming on the horizon. If Wave 2 was about the experience, Wave 3 is driven by pure application necessity. It’s the perfect wave of data and processing that’s coming from algorithms, machine-to-machine communications and ultra real-time decisions demanded by an automated world. Gartner calls it Mode 2, IDC calls it the 3rd platform; whatever your buzzword is, prepare yourselves as it’s on the horizon.
What does it mean for flash? The first thing to note here is that in Wave 3, we’re no longer talking just NAND flash — it’s broadened to all nonvolatile memory. In Wave 3, these memory solutions combine to create the speed of DRAM and the capacity of the largest flash environments — all in a single system. A system that runs (and stores) massive amounts of data in a nonvolatile memory space is constantly fed by even more massive pools of flash because changes and updates are continuous. It sounds frightening as I write it, but we can already see this happening as we move past the MapReduce solutions of today into even more powerful and sophisticated solutions, like Apache Spark, SAP HANA and VoltDB. Even at this level, the combination of these memory solutions is probably between 15 to 40 percent of the capacity. In Wave 3, flash will become the primary medium for new systems, and spinning disk will be relegated to older, more traditional systems that are secondary and purely archival for forgotten or isolated data.
Which Wave Are You Riding?
No matter the wave, the industry is delivering world-class storage solutions to improve and modernize your data center and your experience.