Hedvig came out of stealth this week with distributed storage software that aims to virtualize any type of commodity...
hardware and can achieve petabyte scale for cloud deployments.
Hedvig founder and CEO Avinash Lakshman invented the Cassandra distributed database system at Facebook and helped invent the Amazon Dynamo NoSQL database. He spent the past two-plus years raising $12.5 million at Hedvig and working to bring distributed systems capabilities to storage. Lakshman calls the Hedvig Distributed Storage Platform "a modern approach to software-defined storage."
"I took a look at virtualization, storage and the cloud, and I felt there was no fundamental innovation in storage over the last decade at least," Lakshman said. "Most of the innovation that happened was incremental. I felt I had accumulated a unique set of skills having built the kind of systems I've built, and I felt that if that could be stitched together in an intelligent way, then a big industry could be disrupted fundamentally."
The goal for Hedvig software-defined storage is to turn commodity servers into petabyte-scale block (iSCSI), file (NFS) and object (Amazon Simple Storage Service and OpenStack Swift) storage. The vendor claims its software connects to any hypervisor, runs in any cloud, and provides enterprise storage capabilities such as inline deduplication/compression, thin provisioning, server-side caching, snapshots/clones, auto tiering and wide striping.
Hedvig Distributed Storage is expected to become generally available around mid-year, but the vendor said early customers are using the software. Those include Intuit, Van Dijk Education, and Paul Hastings LLP.
Hedvig says its software can support hyper-converged or hyperscale workloads. It scales compute and storage independently in a hyperscale deployment, and scales them together as a hyper-converged system.
Think of DataCore or EMC ViPR created for distributed cloud workloads -- Hedvig software replicates to data centers and clouds, and self-heals by automatically re-creating failed node data on another node in the cluster.
Lakshman said Hedvig will sell the application as standalone software, or package it on an appliance for customers who prefer a turnkey approach. He said there is no theoretic limit on how many server nodes can be managed in a cluster. Pricing will be based on capacity.
No flash is required, but any virtual disk can be pinned to flash as well as hard disk drives.
Hedvig was originally called Quexascale but Lakshman said he wanted a non-traditional sounding company name. "We sat around and put together a list of terms that you can take as attributes of next-generation storage," he said. "We wrote down hyperscale, elastic, distributed, virtual, intelligent and granular. That is Hedvig."
With brash claims, Hedvig software-defined storage has a lot to prove
It's hard to believe a startup with fewer than 30 people can be the disruptive force Hedvig claims it will become. Instead of promising to be the right storage for practically any type of deployment, Hedvig will likely have to narrow its focus to succeed.
Mark Peters, a senior analyst at Enterprise Strategy Group, said Hedvig software-defined storage fits the basic definition of the term, but must find the right market.
"Everyone is trying to claim software-defined storage just like they claimed cloud five years ago," Peters said. "And just like we've had different forms of cloud, there will be different forms of software-defined storage."
"Hedvig's product looks comprehensive. I do think they're talking about the purist's version of software-defined storage -- software that you run on anyone's hardware. But they're not the only people who have tried that. You had the DataCores and Falconstors of the world who, for whatever reason, never made the leap to the big time," Peters said.
Peters said Hedvig also must compete with virtual storage products from EMC, Hitachi Data Systems, IBM and NetApp.
Like any startup, Hedvig will have to prove it can work better than the established vendors.
"You need to create credibility to make a change in the market," Peters said. "You need to have a product that works, and it takes time for people to accept that a product works. Then it will make a difference if they get some big users on board."
Software-defined storage broadens access to functionality
Complete guide to managing your software-defined storage architecture