Recently, IBM announced a $1 billion initiative intended to improve the overall flash storage market and integrate flash storage in the company’s line of enterprise technology equipment, including servers, storage, and other products. The company feels that flash-based storage is an a tipping point in the marketplace and is poised to become much more widely used, thanks to the incredible performance gains offered by the technology. Further, as is the case with any technology, as it approaches a critical mass point, the overall costs of the technology begin to drop and this is certainly happening with flash storage. There are also other significant cost benefits to flash-based storage, such as reduced power consumption. At scale, such power savings can be real and significant.
The changing face of technology
The rise of flash storage has coincided with the rise of big data analytics, hyperscale computing, massive databases, and other I/O intensive workloads. In fact, I see some workloads as inextricably linked to the kind of storage upon which they run. For example, a major analytics application is heavily reliant on the underlying flash storage that powers it. Rather than storage simply being the repository in which data is housed, for certain applications, the storage technology has become an integral part of the overall application and, without it, the application can’t function.
Introducing IBM FlashSystem
As a part of the announcement, IBM also gave airtime to the company’s new IBM FlashSystem products. The product models, the FlashSystem 710, 720, 810 & 820, each provide the customer with an all-flash solution, but in slightly different ways. The 710 and the 810 each provide customers with the best possible performance by including single-level cell (SLC) flash storage. The 710 provides 6.9 TB of raw (5 TB of usable) capacity while the 810 provides 16.5 TB of raw (12.4 TB usable) capacity. The 720 and the 820 use enterprise grade multi-level cell (eMLC) flash storage which, while not providing quite the performance of SLC, provides greater capacity capability. The 720 provides 10 TB of usable enterprise eMLC flash storage (13.6 TB raw capacity) while the 820 provides 20 TB of usable RAID 5 protected eMLC flash (24.7 TB usable RAID 0, 33.0 TB raw capacity) storage. If you’re not sure about the differences between SLC, eMLC and MLC flash storage, read my flash storage primer here at Wikibon. If I’m reading IBM’s list price sheet correctly, the 710 and 810 will run anywhere from around $52,000 to $165,000, depending on the type of drives that are chosen. For the 720 and 820, list price runs anywhere from $190,000 to $350,000. Bear in mind that these are list prices and many options are a la carte.
All of IBM’s new products carry with them major performance benefits, including variable stripe RAID, which enhances performance without sacrificing reliability, and latency figures measured in microseconds. In addition, these system provide hundreds of thousands of IOPS worth of performance capacity, making them suitable for even the most demanding tasks. IBM has joined the ranks of many all-flash plays on the market, but carries with it a brand that many people trust implicitly and that is backed by a vast array of products and services.
At present, I find no evidence that IBM provides built-in data reduction features, such as deduplication. The FlashSystem lineup is geared at raw performance.
First of all, I applaud IBM for taking this bold step. These significant efforts will almost certainly bring down the overall cost of flash storage for the masses, making the technology more accessible to those that need it or are beginning to need it. This includes even SMBs and midmarket firms, which are discovering the need for solutions that include flash storage in order to ensure that workloads perform as desired.
I agree that there are market verticals and projects — trading, big data, hyperscale, large VDI environments – that can gain significant benefits from all-flash systems. In such environments, low latency figures are in high demand and there a big I/O needs. System such as IBM’s FlashSystem lineup fit these low latency, high IOPS situations very well and can provide performance levels not possible with other solutions.
As a part of the company’s announced, IBM indicated that many customers actually pay more for their rotational storage than they would for an all-flash system. This is due to some of the techniques that people use – such as short stroking – to gain as much performance as possible out of rotational disk system. This is a storage practice in which data is written only to the outermost rings of a disk, thereby minimizing the distance that read and write heads have to travel. While this can yield performance gains, those gains come at the cost of capacity as customers simply don’t use a good chunk of their storage.
For mainstream CIOs running typical mainstream workloads, though, I still believe that it’s too early to take an all-flash approach. Although the economics are getting better, the fact remains that the dollar per gigabyte figure is far higher in an all-flash environment than it is in a rotating disk environment, particularly since many organizations aren’t using things like short-stroking. I remain convinced that the near-term opportunity for all-flash storage still remains in outlier territory when compared against mainstream needs and that, for affordable performance gains, CIOs should consider hybrid storage players that provide very good performance at very good prices. Obviously, for CIOs in extremely I/O intensive industries, the flash market is getting more competitive every day and all-flash solutions are becoming reachable to more organizations as prices continue to drop. I believe that we are still a few years away from seeing significant mainstream adoption of all-flash arrays, particularly as hybrid devices gain more and more features and performance. That said, and as I mentioned, I believe IBM’s foray into all-flash provides significant validation for this space in the long term and that IBM will bring to the market engineering breakthroughs and eventual cost reductions that will accelerate the pace of a switch from hard disks to all flash.