Infinidat Now Using Machine Learning to Optimize Storage Performance

InfiniBox takes a software-optimized approach, using machine-learning algorithms to extract high performance and reliability out of low-cost hardware.

Infinidat.logo

Infinidat, which specializes in petabyte-scale data storage systems, on Nov. 14 launched the latest release of its InfiniBox core operating software, v4.0, that now uses machine-learning algorithms to produce high performance.

The rise of enterprise digital transformation requires new ways of thinking about information storage, the company said. Infinidat offers a new storage architecture that is fundamentally different from the consumer-centric cloud storage platforms of the last decade and earlier legacy enterprise storage systems.

Unlike conventional enterprise storage systems that rely on flash hardware for performance, InfiniBox takes a software-optimized approach, using machine-learning algorithms to extract high performance and reliability out of low-cost hardware. This includes the same ultra-high capacity drives widely employed by Google, Facebook and other hyperscale cloud operators for large-scale data storage, the Waltham, Mass.-based company said.

System Consolidates Legacy Storage

Infinidat began selling its flagship InfiniBox solution in 2014, and its customers deploy InfiniBox to consolidate large numbers of legacy enterprise systems onto a more efficient, higher-performance platform. Each InfiniBox system manages up to 5 petabytes (5 million gigabytes) of data and provides industry-leading performance and reliability.

“As we continue to advance and adapt our solutions to meet rapidly evolving enterprise customer and market demands, we do so with very disciplined oversight on quality, reliability, and usability,” Chairman and CEO Moshe Yanai said. “InfiniBox 4.0 represents our most feature-rich release to date, and incorporates capabilities that will further enhance our business continuance and service provider support.”

Among the key features in InfiniBox 4.0 are an advanced, performance-optimized synchronous replication capability that delivers RPO=0 at ultra-low latency and nominal bandwidth consumption, as well as a fully configurable automated fallback to asynchronous replication in the event of a network failure.

A new quality-of-service feature in InfiniBox 4.0 allows InfiniBox customers to explicitly control resource allocation and consumption at a granular level while providing a non-disruptive governance model based on tokenization of resources, and including a “burst allowance” for optimal SLA management.

As with all features, enabling and managing QoS is a one-click operation within the InfiniBox UI, the company said.

InfiniBox R4 is available starting Nov. 14 for select deployments and will be generally available later this month. As is standard with all Infinidat releases, all new features are included in the base purchase price, with no additional license fee attached.

Infinidat Founder a Legend in Data Storage Annals

You might know Yanai’s name from his past experience in the storage business—he has been instrumental in the development of some of the most advanced and successful storage technologies in the industry.

Yanai began his career in the 1970s, building IBM-compatible mainframe storage based on minicomputer disks. He went on to develop high-end storage systems for Nixdorf, and in the late 1980s, joined EMC, leading the team that developed Symmetrix (today's DMX Series).

As vice president of EMC's Symmetrix group, Yanai was responsible for design and development of all hardware and software products for EMC's industry-leading enterprise storage systems. EMC's Symmetrix group grew from one employee in 1987 to more than 3,500 under his leadership. It is still a mainstay of EMC's core storage business.

Yanai founded Infinidat in 2011.

Chris Preimesberger

Chris J. Preimesberger

Chris J. Preimesberger is Editor of Features & Analysis at eWEEK, responsible in large part for the publication's coverage areas. In his 12 years and more than 3,900 stories at eWEEK, he...