Flash storage is evolving on three vectors. Performance continues to be the key vector. Most data centers buy flash because of its performance attributes; high IOPS and low latency. The second vector is density. Flash has the ability to provide dozens of TBs of storage in standard drive form factors. The third vector is drive connectivity, the connection between the device and CPU/memory is reducing its latency and improving its performance. During the Flash Memory Summit, Seagate will make announcements in each of these vectors.
The performance of the NAND flash is typically not the limiter to overall performance. It is the material that surrounds the NAND flash that impacts performance. This means the software that interacts with the flash is an area for improvement. Essentially, the smarter and more efficient the software the better the overall performance and endurance of the drive.
Seagate is demonstrating an upcoming technology called Nytro Q-Boost, which provides hyperscale application acceleration. It uses IO determinism to isolate the IO patterns of specific application workloads and adjust how the flash responds to them. The result as much as a 16X reduction in latency.
Q-Boost also support multi-streams which keeps like data together during the program and erase cycle. Doing so can reduce write-amplification by 4x. Write amplification is a phenomenon associated with flash memory where the actual amount of data written to the flash can be an order of magnitude greater than what is originally written. If data is inefficiently written to flash then during program erase cycles, more data has to be swapped around. Multi-stream better organizes data placement and as a result reduces this swapping which in turn leads to higher performance (3X) and increased NAND endurance.
Seagate will demonstrate a 64TB full-height, half-length NVMe Add-In-Card. The card provides enough controller capability via its PCIe Gen 3 x 16 interface to deliver excellent performance at this capacity. Seagate is claiming 13 gigabytes/s for sequential reads, 2 million random reads, and 800,000 random (70% read, 30% write) sustained IOPS on the card. The card leverage 3D TLC and QLC to optimize costs. At 64TBs per drive that means a 16 drive system could deliver a PB of capacity.
NVMe is a solid state memory specific IO protocol that increases queue depth and command count to reduce latency and improve performance. Seagate is fully behind NVMe and at the show will announce the Nytro 5000 NVMe SSD is available for sampling now and will move into production this quarter. The Nytro 5000 is a M.2 PCIe based SSD ideal for storage system builders who might use it in tiering or caching applications or for data centers that want high speed direct attached performance. Seagate is claiming 143,000 random (70% read, 30% write) sustained IOPS on the card and 245,000 random reads.
The Nytro 5000 makes for a perfect compliment to Seagate’s other announced drive the Nytro 3000. The new Nytro 3000 is a SAS-based SSD that is able to saturate dual 12Gbps links and should be available in Q1 ’18. It will be available in capacities ranging from 400GB to 15TB. It provides optimizations to improve OLTP performance, security and reduced write application.
Seagate has obviously moved well beyond being the hard drive company and offers a very robust portfolio of flash storage solutions that service data center needs of all types. They are pushing the leading edge in performance, capacity, density and affordability. Server vendors, storage vendors and data center should give the Seagate flash portfolio careful consideration for inclusion in their upcoming plans.