Startup WekaIO Scales Storage Performance, Capacity Across On-Premises And Cloud

WekaIO has introduced what it calls the first cloud-native scalable file system aimed at using flash storage technology to scale storage across both on-premises and public cloud infrastructures.

WekaIO, which last week officially exited stealth mode after raising a total of $32 million in funding, is founded by three members of the original team from high-performance storage developer XIV, which IBM acquired in 2008.

Storage offerings typically address either performance, simplicity or scalability, but not all three, said Barb Murphy, chief marketing officer for the San Jose, Calif.-based vendor.

[Related: Emerging Vendors 2017: Cloud Startups You Need To Know]

id
unit-1659132512259
type
Sponsored post

"We do software that combines all three across on-premises and public clouds," Murphy told CRN.

While public clouds have done a fantastic job when it comes to networking technology, storage in public clouds is still in one of those three silos, Murphy said.

"This means data is fundamentally never shared," she said. "WekaIO brings storage to the compute layer so that as performance increases, customers get more scale. It supports up to eight storage targets from on-prem to Amazon Glacier. The key is, everything is presented to the application as a global name space."

The WekaIO software deploys on industry-standard servers. The company already has reference platforms for Hewlett Packard Enterprise Apollo and Dell EMC servers, and is working with Supermicro and Lenovo on future reference architectures, Murphy said. The software runs on bare-metal servers or virtual machines or in containers, she said.

"We're completely software-defined," she said. "We're providing the next generation of easy-to-deploy software with the performance of all-flash arrays."

The software scales from six nodes to 240 nodes with almost no impact on latency, Murphy said. A 30-node cluster can address up to 2 petabytes of storage with up to 1.8 million IOPS of performance and up to 60-GBps bandwidth, she said.

While WekaIO's initial engagements have been primarily direct with customers, the company expects that by next year about 70 percent of its business will come via channel partners who can integrate the software with customers' own hardware in the field.

"We're just a component of a successful sale," Murphy said. "It doesn't make sense for us to sell the software and hope the other components fall into place. And a lot of our partners bring us an important vertical focus."

The data is protected using n+2 or n+4 redundancy, meaning the data remains available even if two or four of the nodes simultaneously fail, Murphy said. If a node does fail, it can be rebuilt in as few as 15 seconds in a 60-node cluster.

"We don't assume the customers have battery backup," she said. "You can use the simplest, dumbest servers available."

WekaIO's software-defined file system meets the needs of customers looking to store, protect and access file, block and object storage data, said John Downey, founder of NextGen Storage, a New York-based solution provider and WekaIO channel partner.

The WekaIO front end acts like all the most common storage protocols, including Amazon S3, NFS (Network File System), SMB (Server Message Block) and HDFS (Hadoop file system), he said.

"It snaps right into legacy storage and next-generation storage," he said. "And the performance is linear. The larger the cluster, the better the performance, all in a predictable manner."

WekaIO not only offers high-performance unified storage on premises, it also fills a critical performance gap related to file sharing in the cloud, Downey said. "And it's certainly a good solution for customers who don't want to refactor their applications with cloud APIs for performance," he said. "For applications that are not refactored, it's hard to use the cloud for high-performance data at an economical cost."

The WekaIO software has been shipping for about six months, and the newest release just started shipping, Murphy said.