bluebay2014 - Fotolia

Layer 3 network integration from Datera uses nodes as endpoints

An updated version presents Datera Elastic Data Fabric storage nodes as a cluster of L3 network endpoints in anticipation of increased container adoption.

Storage startup Datera upgraded its Elastic Data Fabric software with Layer 3 network capabilities to help support containers, and it added adaptive compression across heterogeneous server nodes.

Datera storage helps organizations develop cloud-native applications. EDF 2.2 natively integrates Layer 3 (L3) network management tools in the Datera scale-out control plane. The feature presents Datera EDF storage nodes as a cluster of L3 endpoints.

Enhancements in EDF 2.2 mostly target hosting providers and large enterprise deployments of Docker containers. The Datera workload chassis allows multiple services to be launched and managed on a central plane.

"We will behave like L3 endpoints in a cluster. Think of a standard rack in a large data center where a native L3 network extends through the data center all the way to edge nodes," said Ashok Rajagopalan, head of products at Datera, based in Sunnyvale, Calif. "In the world of containers, this architecture is going to become more and more popular."

How a Layer 3 network affects storage

Layer 3 refers to the third of seven layers in the conceptual Open Systems Interconnection reference model. An L3 device serves as a combination switch and router. The switching capabilities connect devices at wire speed that share an IP subnet or virtual LAN. Intelligent IP routing directs the flow of network traffic.

In the world of containers, this architecture is going to become more and more popular.
Ashok Rajagopalanhead of products, Datera

Container mobility remains a drawback to widespread enterprise adoption within a traditional L2 network topology. As a security measure, traditional L2 network topologies do not permit an IP address to migrate from the rack on which it was created. Scaling an L2 cluster across multiple racks requires special iSCSI commands and an overlay tunnel to cross to an L3 switch.

Datera takes out the encapsulation layer and associated network management. A dedicated Datera EDF node advertises across the Layer 3 network whenever a target port IP address needs to migrate across racks for failover or load balancing. Each node in the Datera cluster then participates instantly in all network route changes.

Modernized data centers with rack-level L2 subnets are expanding L3 capabilities, often using L3 as the default network to support dense container clusters, Rajagopalan said.

"Container-only networking can be mapped directly to our storage. We make it easy to deploy and scale containers," Rajagopalan said.

Goal: Emulate Amazon EBS, but on premises

Datera sells branded block storage built on standard x86 servers. The vendor claims its storage provides a local version of Amazon Web Services' Elastic Block Store services.

Scalable, self-provisioned storage is a market with no shortage of competitors. It includes incumbent vendor products such as Dell EMC's ScaleIO and startups such as Hedvig, as well as object storage providers.

All-flash Datera nodes range from 10 TB to 20 TB. Hybrid flash models range from 48 TB to 96 TB. All totals are raw storage. Customers can mix flash and hybrid media to scale a Datera cluster to 50 nodes and 5 PB capacity.

EDF may be implemented as software-only reference architecture on qualified Dell, Hewlett Packard Enterprise and Supermicro servers. The vendor also sells Datera EDF storage as a service to hyperscale data centers.

A Datera Intel Optane SSD node is in the works that would start at 3 TB. Optane is Intel's branding for devices made with its 3D XPoint persistent memory technology. Intel began limited shipments of Optane SSDs in March.

Jeff Kato, a senior analyst and consultant at Taneja Group Inc. in Hopkinton, Mass., said the Layer 3 network integration gives Datera a chance to differentiate.

"I haven't found any other storage vendors natively adding L3 yet," Kato said. "Datera is professing it can cover a pretty wide range of block workloads through configuration management, node management and moving workloads from all-flash to hybrid. The EDF nodes can have different quality of service to meet your different performance levels."

Kato said Datera's strategy of selling to cloud providers "seems like a natural fit versus trying to displace VMware vSAN or ScaleIO."

EDF adds adaptive compression per node and storage media

Previous EDF software versions supported compression, encryption, multifactor replication, snapshots, thin provisioning and volume resizing. Data deduplication remains a roadmap item.

In addition to Layer 3 network integration, EDF 2.2 added adaptive compression across distributed nodes. The Datera compression algorithm adapts to run either inline or out of band, based on a node's capabilities and the application's performance template.

Datera fortified its data placement capabilities to handle profiling and management of heterogeneous server clusters. That allows servers from multiple vendors to coexist within the same cluster.

The EDF agent knows the compute power, memory and storage media on each server and automatically places an application on the node that corresponds to its required service level.

Next Steps

Comparing scale-out and object storage approaches

Persistent container storage makes inroads in enterprise

Datera locally managed block storage resembles AWS

Dig Deeper on Storage architecture and strategy

Disaster Recovery
Data Backup
Data Center
and ESG