The Split Provisioning feature can separate scaling of host storage speed and persistent capacity. The software supports scaling from one to 128 compute nodes, or input/output (IO) bandwidth up to 200 Gb/s, and up to 10 data nodes with capacity up to 1.7 petabytes in a single system.
Datrium’s previous product supported just a single data node.
The compute nodes run all workloads in local commodity flash, which results in lower latency. Each compute node writes persistent data to network-attached and managed data nodes. With the compute nodes operating as stateless, data availability is not impacted should some of them fail.
“Datrium’s premise has always been that their ‘open convergence’ method delivers better overall scalability, along with independent scalability of performance and capacity, compared to hyperconverged architectures – in essence, private clouds with virtually limitless performance and scale with the ease of public clouds,” said Arun Taneja, founder and consulting analyst at Taneja Group, in a statement.
Datrium’s open converged infrastructure (OCI) is different from hyperconverged infrastructure (HCI) systems that can maintain persistent data on every server, according to the company. HCI systems are also limited to up to three simultaneous server failures before a data outage.
These limitations require HCI users to run smaller clusters of between eight and 16 nodes, according to Gartner. To support larger data storage needs, companies would need to deploy multiple clusters, which requires increased configuration and management.
“Performance is on compute nodes with flash, and persistence is erasure coded across a cost-optimized object store pool (data nodes), as nature intended,” said Datrium CEO Brian Biles. “So it doesn’t suffer HCI’s erasure coding performance problems for hot data.”
Pricing for such capabilities is set at $12,000 per compute node.
Datrium last month began offering new storage support for various open source platforms. This included Red Hat Enterprise Linux (RHEL) servers and kernel virtual machines (KVM) to its core software platform that supported VMware servers and vSphere virtual machines.
The additional support garnered applause from analysts, who noted benefits for enterprises being able to manage multiple environments from a single source.
Mike Matchett, senior analyst at Taneja Group, said the Datrium approach is a “good option” for enterprises “to consider in order to accelerate and open up existing compute clusters.”
“It represents a safe, forward-looking investment even if you are focused on just VMware today,” Matchett said. “Given that storage refresh cycles are traditionally three to five years, it’s a big risk these days to not build in some agility and adaptability into a current refresh.”
Matchett did note the firm would have to show it can scale to the needs of larger enterprise customers.
“Datrium will need to prove out large-scale scalability and protection claims, but significant performance improvement with server side flash is almost assured,” Matchett said.
Datrium, which launched in 2012, late last year scored $55 million in Series C funding, boosting its capital haul to $110 million since its founding.