Mopic - Fotolia

News Stay informed about the latest enterprise technology news and product updates.

Startup Datera stretches Elastic Data Fabric with all-flash nodes

Datera Elastic Data Fabric qualifies Dell PowerEdge servers, adds all-flash nodes, and increases scaling to 50 nodes and 5 PB raw capacity per cluster.

Software-defined storage startup Datera is expanding its Elastic Data Fabric capabilities with all-flash nodes, role-based multi-tenancy, encryption and compression.

Elastic Data Fabric is designed to help companies and cloud providers develop born-in-the-cloud applications. Datera uses its scale-out iSCSI block storage software to stitch together commodity-based hardware nodes. The vendor claims its elastic data plane can scale symmetrically to provide on-premises data center block storage that resembles public Amazon Elastic Block Store in the cloud, a model other vendors are also pursuing.

Datera came out of stealth in April, launching 50 TB hybrid nodes built on 2U Supermicro x86 servers and NVMe flash. Version 2.0 launched this week, and, now, customers can purchase all-flash nodes from Datera resellers built with Dell PowerEdge R630 servers. The 1U, 10-drive flash systems are available in 10 TB and 20 TB capacities using generally available solid-state drives. Version 2.0 also supports Dell PowerEdge R730 servers for hybrid nodes.

Datera increased scaling from 20 to 50 nodes per cluster, and up to 5 PB of raw capacity -- 100 TB per node. Customers can mix and match Datera hybrid and flash nodes.

Datera: Companies want storage that is easy, elastic

Datera CEO Marc Fleischmann said data center customers -- particularly cloud service providers -- expect scale-out storage that doesn't hog bandwidth, yet is simple to manage and provision. Datera reduces latency for cloud-based workloads by distributing data management and data services across a cluster of storage servers.

Datera rates its nodes to deliver an average of 100,000 IOPS and submillisecond latency. Datera flash and hybrid nodes feature an NVDIMM or NVRAM layer to absorb writes at memory speed, before acknowledging writes to the host and flushing them to NVMe flash or SAS-based HDDs on the back end.

Datera Elastic Data Fabric software profiles how an application is used and maps it to the most appropriate storage. A configurable policy engine automates dynamic tiering across cloud, physical and virtual storage.

"Our innovation is that we can drive latency down on every node in a cluster," Fleischmann said. "Our control plane is able to extend the same performance and latency characteristics of a single node to a set of nodes that are dispersed across generic IP networks."

Aside from the flash nodes, Datera 2.0 adds role-based multi-tenancy that allows discrete workloads to be isolated and managed on the same Datera cluster.

"The Datera system understands application requirements at a reasonably high level, all the way up to their business intent," Fleischmann said. "We also learn the data center capabilities and constraints to enable smart data placement. And we self-optimize and balance workloads all the time."

At-rest encryption, compression added; dedupe coming

Datera already supported instant multifactor replication, "zero overhead" snapshots and clones, and thin provisioning. Version 2.0 adds data compression, encryption of at-rest data and live instant volume resizing. Fleischmann said data deduplication is on the roadmap.

Datera 2.0 expands on Docker Swarm and Google Kubernetes container storage orchestration by integrating support for Apache Mesosphere. Other additions include VMware vSphere Storage API certification and plug-ins for vSphere and vRealize.

Jeff Kato, a senior analyst at Taneja Group Inc., in Hopkinton, Mass., said Datera's data-placement process could help separate it in the crowded software-defined storage market.

"Their secret sauce is multi-tenant adaptability to the workloads," he said. "It's a data-placement layer they say can guarantee quality of service. They are claiming you can have an architecture that spans tier-one to tier-three workloads based on how you configure it. Typically, that's a hard thing to do.

"Datera is putting a lot of emphasis on covering all the workloads and going after next-gen workloads that are born in the cloud. Scale-out architectures are going to take the bulk of the workloads going into the future, which is why there is such a huge battle between software-defined storage vendors."

Datera sells its appliances through resellers, starting at a minimum configuration of three nodes per cluster and a perpetual license for bundled Datera Elastic Data Fabric software. Customers with existing qualified Dell and Supermicro hardware can purchase Datera as a software-only installation from resellers.

Fleischmann said proofs of concept are underway with several cloud service providers, including Comcast and Google.

"We are working on a joint go-to-market [strategy] with Google for Google-like on-prem clouds with Kubernetes orchestration," he said.

Fleischmann said Datera hopes to finalize partnerships with Cisco and Hewlett Packard Enterprise to qualify its software on their hardware.

Next Steps

Scalability vs. elasticity: What's the diff?

Datera storage fabric targets cloud

Scalable storage approaches expand

Is elastic storage stretching it?

Dig Deeper on Hybrid cloud storage