Product and service reviews are conducted independently by our editorial team, but we sometimes make money when you click on links. Learn more.

Flash And 3D XPoint Are Better Together, Plexistor Launches Software Defined Memory

By - Source: Plexistor
Tags :

Plexistor aims to bring the inherent benefits of software-defined architecture to the storage and memory tiers, including bleeding-edge NVDIMM and 3D XPoint products, with the release of its Software Defined Memory (SDM) architecture.

The era of software defined everything is upon us, and it is increasingly utilized to abstract away complexity and cost from underlying hardware while simultaneously increasing performance. The continued maturation of numerous software-defined initiatives (on every front) is altering compute, storage and networking design.

Plexistor is a scrappy start-up from Mountainview, CA that managed to bring its product to market quickly; it only has 20 employees and was established in 2014. The company, and its Israel-based R&D team, were able to develop a new technique that dynamically tiers data over several layers of memory and storage. The company has 12 patents pending and a head full of steam as its first product comes to market.

The concept of combining storage and memory to boost performance is not new by any means, there have been several implementations in the past. However, many of these implementations primarily relied upon system RAM for caching. Plexistor differs from other approaches through its incorporation of NVDIMMs for data caching. The Plexistor multi-tier SDM cache algorithms also leverage host RAM as part of a four-tier architecture that melds the best aspects of RAM, NVDIMMs, NVMe SSD and HDDs into one cohesive layer. This allows administrators to employ both in-memory and traditional enterprise applications in a single infrastructure without the need for dedicated clusters.

SSDs are the go-to performance boosting resource, and utilizing them in tandem with HDDs as a layer of speedy cache is a common tactic. Plexistor expands upon that approach by utilizing a combination of RAM and NVDIMMs to cache data for the underlying flash and HDDs. Plexistor provided us with the chart above to highlight its internal test results, which show an increase of 118x to 800x with 4K random 50 percent read/write workloads (measured in FIO), and an incredible performance increase of 2.5x to 4x with MongoDB NoSQL. The company indicates that SDM reduces SQL database latency by 24 percent. 

The use of NVDIMMs provides a layer of safe data storage that has non-volatile characteristics and user data is not lost if there is an unexpected host power loss. The company utilizes Micron and SMART Modular NVDIMMs, and the emerging pmem driver ecosystem. SDM will support 3D XPoint and other next-generation NVM products as they come to market. Plexistor also designed the software to work well with more traditional NVMe SSDs and upcoming NVMe over Fabric (NVMeF) implementations. Perhaps most surprisingly, the SDM solution supports both on-prem and AWS EC2 deployments.

The result of combining multiple tiers into one pool is bleeding edge performance in a wide variety of log analysis, DevOps, cybersecurity and financial applications, among others. Plexistor made its first version of the product free for users, and the community edition can be found here. The company will offer a subscription-based version once the development is complete for roughly $2,000 per server.

Bringing the speed of a customized multi-tier software defined memory product to the cloud is a game-changer, and one that will have us following Plexistor closely as the company moves forward.

Paul Alcorn is a Contributing Editor for Tom's IT Pro, covering Storage. Follow him on Twitter and Google+.