AccelStor: Unleashing the True Potential of Flash

There comes a time in the lifecycle of every technology when vendors need to supply businesses with more benefits apart from base-level expectations such as cost savings, increased productivity and ease of management. In an era where round-the-clock availability and speed are paramount, businesses need to step up their game and reinvest in innovation to accelerate their IT. Over the last decade, while flash-based hardware revolutionized the speed and efficiency of high performance computing (HPC) applications, the software—used to group solid state drives (SSD) like redundant array of independent disks (RAID) were initially designed for mechanical devices based on spinning media—inducing a bottleneck when utilized in a purely digital flash memory.

Fueling the transformation from conventional disk arrays to modern all-flash storage, AccelStor was born to eliminate this bottleneck and unlock the true potential of SSDs. Founded on a ten-year research project on flash memory technology, the company brings to all-flash arrays, enhancing and fortifying IT infrastructures.

AccelStor’s award-winning FlexiRemap^® technology, a flash software technology, harnesses the latest advancements in processor architecture to achieve unparalleled performance, versatility, and scalability of storage arrays. FlexiRemap^® technology designs SSDs from the ground up with a distinctive algorithm that allows NeoSapphire, an All-Flash Array (AFA) model, to efficiently handle the workload of a flash memory in HPC applications. FlexiRemap^® technology’s ability to “remap” large amounts of random data into sequential before writes direct to disk while maintaining data integrity in real-time without data caching, outperforms a RAID system by a factor of 1100 percent. Replacing the data redundancy of conventional RAID storage with a top-tier all-flash array system to achieve maximum performance, FlexiRemap^® technology breaks down any data block size, be it 16k, 32k, 64k or 128k, into 4k chunks prior to writing them sequentially over the SSD array by distributing the workload evenly across all drives.

Tiering data within multiple SSDs requires a systematic recovery of storage pool—Garbage Collection (GC), which induces a temporary drop in solid-state media performance as SSDs clear space from deleted blocks. To that end, FlexiRemap^® technology’s intelligent GC algorithm efficiently groups the hottest data in the uppermost blocks of each drive, minimizing the Program/Erase cycle impact, while ensuring consistent performance. The NeoSapphire AFAs leverage unique and complete inline and background data reduction technology to intelligently deliver average 5:1 data reduction ratio, with VDI environments gaining more than 10:1, and ensures optimal performance even during peak loads. This effortless solution enables IT administrators to have a better experience and reduced OPEX. For instance, companies could enjoy VDI advantage with NeoSapphire AFAs in lowering the cost per power user desktop less than $30. Additionally, NeoSapphire All Flash Array is based on a “Shared-nothing” architecture that reduces downtime caused by storage controller, networking and storage pool failures without affecting write performance. This enables NeoSapphire AFAs to perform at staggering speeds using SATA SSDs for price performance are as low as $0.09. “The blend of low cost and high performance makes NeoSapphire AFAs ideal for HPC, artificial intelligence (AI) and business intelligence applications,” states David Kao, Vice President.

One of AccelStor’s clients, ECS, deployed 2x 6U hybrid storage systems from a Tier 1 vendor to perform monthly business intelligence reports from a 160GB database. However, the report process through their existing platform consumed over three hours, which led the client to approach AccelStor to reduce the time cost. After implementing AccelStor’s NeoSapphire H710 all-flash array into their infrastructure, the client experienced a significant drop in the operation time, from three hours to 46 minutes, all while consuming a mere 4U of space. With its shared-nothing architecture, H710 delivers real active-active clustering with ZERO impact to data availability and performance in the event of hardware failures, offering total protection right out of the box, all while offering enticing cost per IOPS statistics.

Scripting similar success stories, AccelStor strives to leverage and apply its rich expertise in storage arrays to bring in the latest, exciting technology in flash, and continuous innovation of Non-Volatile Memory express (NVMe) in the ever-evolving flash memory market. “With a new approach to enable computer nodes to access NVMe JBOF, this new technology takes a huge leap in harnessing the true potential of flash memory,” David Kao says. Moving forward, AccelStor is currently spearheading bleeding edge research to develop the next iteration of FlexiRemap^® technology algorithm to exploit the benefits of NVMe and NVMe over fabric (NVMeOF).