Kaminario secures $75 million funding round

All-flash array pioneer Kaminario kicked off 2017 with a cash infusion of $75 million to accelerate global expansion and fuel product support for non-volatile memory express (NVMe) technologies.

Kaminario’s fifth funding round increased its overall total to $218 million since the company launched in 2008. Waterwood Group, a private equity firm based in Hong Kong, led the latest financing effort. Additional investors included Sequoia, Pitango, Lazarus, Silicon Valley Bank and Globespan Capital Partners. Kaminario’s most recent previous funding came in January 2015.

Founder and CEO Dani Golan said Kaminario has more than doubled revenue in each of the last four years. Global expansion and go-to-market efforts will focus on eastern and western Europe, Asia Pacific and the Middle East.

Kaminario will concentrate on incorporating NVMe technologies into the company’s K2 all-flash array. Kaminario currently uses SATA-based 3D TLC NAND flash drives. NVMe-based PCI Express (PCIe) solid-state drives (SSDs) can lower latency and boost performance, and NVMe over Fabrics (NVMe-oF) can extend the benefits across a network.

Golan said Kaminario does not support NVMe SSDs yet because “the price is too high.” He added that the NVMe-oF technology “is not mature enough to run in mission-critical and business-critical environments.”

A handful of new companies are starting to ship products with NVMe drives, but Golan said Kaminario’s NVMe support will probably wait until 2018.

“The ecosystem is not there yet,” he said.

Golan said startups that currently support NVMe use drives directly attached to servers. But, with a mature array platform on the market, Kaminario needs “to drive a full storage software stack over NVMe Fabrics,” he said.

“The big gain is [going to be with] NVMe over Fabrics, because NVMe drives are just media. That’s not interesting. The interesting part is NVMe over Fabrics and NVMe shelves,” Golan said.

Kaminario’s architecture allows customers to add controllers or shelves in any combination, scaling compute separately from storage, Golan said.