On January 12 IBM announced some very serious upgrades to its DS8000 series of storage arrays. Until this announcement only the top-of-the-line model, the IBM System Storage DS8888, was all-flash while the less expensive DS8886 and DS8884 sported a hybrid flash + HDD approach. The new models of the DS8886 and DS8884 are now also all-flash.
But that’s not all: Every model in this product family has been upgraded.
The original DS8000 systems used a module called the High Performance Flash Enclosure (HPFE) for any flash they included, while these newer models are all based on HPFE Gen 2. While the original HPFE was limited to a maximum capacity of 24TB in a 1U space, the larger 4U HPFE Gen 2 can be configured with as much as 153.6 TB, for more than six times the storage of the previous generation. By making this change, and by optimizing the data path, the Gen 2 nearly doubles read IOPS to 500K and more than triples read bandwidth to 14GB/s. Write IOPS in the Gen 2 have been increased 50% to 300K, while write bandwidth has been increased by nearly 4x to 10.5GB/s .
This kind of performance opens new opportunities. While IBM tells us that it’s difficult to do real-time analytics in a shared-nothing installation using Hadoop, that’s exactly how the world’s largest bank is using its DS8000 system to detect fraud in near real time.
The company’s approach to flash is not simply to replace HDDs with SSDs. Although this method achieves significant performance gains, a lot of the flash chips’ performance is lost thanks to delays in system interfaces and software. Instead, IBM re-architected the system, interfaces, and software to make all these elements complementary to flash so that they could squeeze the most performance out of their customers’ flash investment. IBM’s spokespeople told us that the company is: “Typically the champion,” when response times are compared.
IBM attributes an important part of its high performance to the fact that these arrays are all based on the company’s POWER8 processor rather than more standard server chips. The company says that this not only helps these systems to better integrate with IBM z Systems and Power Systems servers, but it also allows IBM to use the same internally-developed storage management software across all three models, enabling all models in this line to provide latencies of less than 1 millisecond.
Although they are all based on the POWER8 processor, there are differences in the three models’ controller hardware: The “Business Class” DS8884F, aimed at midrange enterprise installations, provides dual 6-core controllers with 256GB of DRAM cache, and 32 Fibre Channel ports, while the “Analytics Class” DS8888F gives users 48-core controllers, 2TB of DRAM cache, and 128 Fibre Channel ports. The enterprise-class DS8886F falls between these two with 24 cores, the same cache and port count as the DS8888. Prices for the DS8884F start at $95,000.
With this announcement, all of IBM’s storage systems, ranging from the more economical Storwize systems, through the FlashSystem line and DeepFlash, to the top-end DS8000 series, are available in all-flash configurations. The DS8000 series is aimed at business critical applications with replication to as many as four sites and what IBM calls: “Industry-leading reliability”. At “Six Nines” (99.9999%) uptime that’s a pretty solid advantage, and it’s one that allows IBM to boast that 19 of the world’s 20 top banks already use the DS8000 family.
Nothing succeeds like success, and that’s what IBM is aiming for with its flash storage. The company has found that its focus on becoming a leading force in the flash array marketplace has provided it with competitive advantages in its higher-end arrays that can be economically ported to its lower-level systems to provide powerful features to smaller companies. These smaller organizations make up a much larger number of data centers, a fact that should benefit IBM very well over the longer term in terms of sales volume.
The company has told The SSD Guy that it will continue to improve the design of its flash arrays to keep pace with new chips being offered by their flash chip suppliers. By their estimate, upgrades will be announced about every 6 months. At the most fundamental level this means that new storage arrays will boast ever-increasing capacities as NAND flash chips continue to do the same, but experience has told us to expect more than that from IBM.
In August The SSD Guy wrote an analysis on IBM’s product refresh at that time, with new offerings that included array controllers with higher processor core counts and the availability of better software across more models. It safe to assume that the company will keep pushing out newer and better flash support to keep pace with these increasing capacities.
In fact, IBM’s flash R&D is bent on achieving the highest performance that the technology can offer, and the company has redesigned both hardware and software to deliver the highest performance flash arrays possible. Over time Objective Analysis expects to see more of this dynamic, with the same software and controller hardware being merged into all levels of IBM’s offerings, to provide consistent management and performance across its flash-based storage systems.