Storage in NVMe-oF mode at Vast Data
To support large accounts using large amounts of data, Vast Data relies on the simplification of storage with decoupling between controllers and bays without sacrificing performance.
Founded in 2016 in Israel by Jeff Denworth, Shachar Fienblit and Renen Hallak, Vast Data worked two years in secret mode before bringing its first high-end storage array to market in 2019. The team, which includes DDN Storage alumni, Kaminario and XtremIO have developed a platform based on an architecture called DASE (disaggregated, shared-everything) which not only provides improved robustness, but also greater scalability with servers in container mode at the front and storage bays at the back. QLC/3DxPoint hybrid NAND flash storage attached via NVMe fabric over Erhernet or Infiniband. The array actually combines NVMe QLC SSDs, for deduplicated data, with Intel Optane components, for metadata storage, all in an NVMe-over-Fabrics environment to ensure optimal data access over the network.
Vast Data’s architecture moves controller intelligence from arrays to dedicated servers to make them easier to update. (Credit Vastdata)
The main advantage of this architecture is that all the intelligence of the controller resides in the containerized storage servers exploiting Governors. It is the latter who manage the requests forE/S, data migration between flash storage tiers, erasure coding, reduplication and encryption. According to Jeff Denworth, vice president of product, latency between servers and mass storage is always under 10 microseconds, and supported protocols range from NFS to S3 to SMB. A standard VAST Data cabinet is based on four 2U servers, an NVMe Ethernet or Infiniband fabric and 400 or 675 TB storage arrays. VAST Data offers two ways to deploy its solution: turnkey servers and storage cluster appliance (up to tens of petabytes in the same rack), or simply storage arrays with Vast container software which then runs on customers’ machines or VMs.
A full cabinet with optimized flash capacity through deduplication. (Credit Vast Data)
Vast Data also announced today a Series C fundraising of $100 million for a now estimated valuation of $1.2 billion (x3 since February 2019). CEO Renen Hallak said he would use the investment to drive expansion and set the stage for the start-up’s future growth. Customers include General Dynamics, Ginkgo Bioworks, Harvard Medical Schoo, National Institutes of Health, Squarepoint, and Zebra Medical Vision, Square Point.