Vexata partners with Fujitsu
Specializing in NVMe and Optane storage subsystems, Vexata offers an architecture that leverages FPGAs to offload x86 controllers from certain tasks and help speed up data processing. And to expand sales, they announced a partnership with Fujitsu.
The following was translated from original article posted in French on Lemonde Informatique 12/18/18
Ashish Gupta, CMO of Vexata, welcomed the partnership with Fujitsu. (Credit SL)
Live from Sunnyvale . Founded in 2013 and out of the stealth mode in 2017, the start-up Vexata is no stranger to the readers of Le Monde Informatique, since we had already met its leaders in San Jose a year ago . The company, which has already raised $ 54 million from Lightspeed Ventures, Intel Capital, Mayfield and Redline Capital, offers a storage solution for demanding applications based on their core architecture. The latter has been specifically designed to reduce the bottlenecks that slow subsystem performance and slow response times, which has a direct impact on the user experience.
Database, analytics, and machine learning environments generate large amounts of structured and unstructured data, but bottlenecks – in performance and I / O – slow down real-time processing. In almost all cases, according to Vexata, the bottleneck is the result of the architecture of the storage controller, especially with full flash arrays. At the same time, as multi-core computing performance increases, the number of virtual machines supported by hypervisors continues to grow. This can result in hundreds of virtual machines per rack in a data center. This type of traffic load can easily overwhelm the full flash storage infrastructure, because a single multi-core server can generate more throughput than most full-flash bays can handle. This forces architects to distribute, load, and acquire more subsystems than is needed to improve performance.
Separating data from control plans is a practice that has been implemented for years in high performance network products. (Vexata credit)
The support of the NVMe-oF if needed
Vexata’s storage systems reduce the latency by separating the control plane from the data plane via the Vexata Operating System (VX-OS) software coupled with a hardware architecture resting in the bays on a “FPGA to avoid going through the processor” said Ashish Gupta, CMO of the start-up. “The latency goes down to less than 10us. Our primary system uses NVMe SSDs and we support the NVMe over Fabrics (NVMe-oF) protocol if needed, but there is no standard on the market.” By reducing the latency, users can take full advantage of NVMe and 3D-XPoint component performance, which means that artificial intelligence, machine learning, real-time analytics, and other limited workloads can now be unleashed.” It should be noted that the VX-Manager software provides visibility, management and analysis of the applications running on the system, as well as the analysis of blockages and other delays.
This year, Vexata also launched an optimized version of its solution for VMware environments and forged a strategic partnership with Fujitsu for the resale of its storage arrays. With the latter, the start-up intends to convince more customers. We can already mention Tata Consultancy, Oath, Pacific Data Center and Sanmina .