Elastifile leverages its distributed metadata model to produce a scale-out, flash-native distributed file system that provides both storage and data lifecycle management.
Elastifile, a Silicon Valley startup with an Israeli back end, has emerged from relative stealth to announce both themselves and their product, a scale-out, flash-native distributed file system designed to unify data within a single global namespace to deliver hyperscale efficiency.
While Elastifile briefly popped out of stealth a year ago to announce their Series B funding round, for the most part the company has steered clear of the public eye. And while multiple scale-out parallel file systems are on the market today, the company says that Elastifile’s is fundamentally different.
“We have built up very quietly since we were founded in 2013, and today is both our product launch and company launch,” said Andy Fenselau, Elastifile’s VP of Marketing. “We were looking to build something that’s better than just a better, faster cheaper storage silo. So we pulled together a team of leadership, investors and funding to build a next-generation data management and storage solution across true hybrid cloud workflows.”
Their backers include four of the leading global enterprise cloud companies – Cisco is the public one – and three of the largest enterprise flash vendors, with Western Digital being the public one. Over $50 million has been secured in investments to date.
“Our design objective is to bring all the cloud infrastructure and hyperscale advantages to the enterprise – elasticity, easy management, and access for business users, not just admins,” Fenselau said. “This includes cross-cloud capability, but without compromising performance SLAs. The idea isn’t just to maybe do things cheaper in the cloud by tiering some data or apps there, but to have everything as a service and move data and apps around in a way that’s driven by end users, not the storage admins.”
The secret sauce here, which provides the application readiness of file, the scale of cloud and the performance of flash is an architecture based on a patented distributed metadata model and its Bizur specialized key-value consensus algorithm. In conventional consensus algorithms, every operation is written to the log, and can only be applied once a majority of the nodes acknowledge it. This has the potential to slow things down, which Bizur overcomes by allowing operations on independent keys to be applied concurrently. The result is consistent 1-2 millisecond latency at any scale in noisy and heterogeneous cloud environments, which allows any workload to meet any enterprise SLA in the cloud or the data centre.
“We’ve blown away traditional bottlenecks with this cloud scale distributed metadata model,” Fenselai said. “Our distributed file system creates an extended global namespace capability, which makes everything more dynamic than had been thought possible across sites and clouds. Before this, a global namespace across different sites with cloud scale and enterprise SLAs was not an option, but that’s what our system does. You have virtual controllers deployed on every node, which confederates across other nodes, on-prem and in cloud, into a global namespace with 1-2 millisecond application response time. And as you add more nodes, you add more performance.”
The Elastifile solution also features advanced data reduction (global deduplication, compression, and snapshot shipping) and CloudConnect object tiering which lets them tier working data with a cold data tier.
“We finalized our product and had quietly had our initial product release in the fall of last year,” Fenselau said. “This 1.0 release showed that we could deliver on enterprise expectations. We signed up our first dozen customers and our first partners. This is ourbig launch, where we already have real customers and real proof points and a next-gen product with the full cross-cloud and cloud deployment options. This is the version that can be deployed in Amazon and Google, and is the one with CloudConnect.”
While the sweet spot is larger enterprises, it is by no means the only market.
“This kind of technology tends to find its way to the Global 5000 first, and medium to large enterprises are really the sweet spot,” said Sebastiano Tevarotto, President and General Manager at Elastifile. “However, channel partners aggregate services and sell them to smaller companies as well, so the benefits of this extend beyond that Global 5000.”
While Elastifile does sell direct, the projection is that a majority of sales will be channel.
“We expect that a minimum of 80 per cent of business will go through partners, and about half of that will be service providers,” Tevarotto said. “We will have both integrator and managed services partners. We enable the MSP model, around our services that are delivered.”
Elastifile is presently building out a channel, with the primary qualification being strong competence around targeted use cases, and the ability to expand the use cases more.
“Numbers of partners aren’t a concern at this point,” he said. “We have some large companies that can provide the breadth we need, and a significant number of smaller ones. They include traditional partners and service providers, and some that do both.”
Tevarotto said that regardless of partner type, they can all benefit from Elastifile’s ability to address the paradox for the channel of how to transition effectively to it while retaining their revenues.
“Cloud has been the trigger of turbulent change, which has brought about changes in the channel, and in distribution, with some of the traditional role of distributors being replaced by AWS and Google,” he said. “We enable partners to leave storage behind and focus on the data, and lifecycle management of the data. That is where the channel can add value.”