Pure Storage leverages Internet of Things with new Pure1 META AI platform

The first product of this technology, designed to understand complex workload relationships with machine learning, is the Pure1 Workload Planner.

SAN FRANCISCO — On Tuesday at their Accelerate customer event here, Pure Storage announced Pure1® META, its new Artificial Intelligence platform, which leverages the Internet of Things to deliver on what it defines as its vision of self-driving storage.

“Self-driving storage comes from a starting point where we have already automated a lot – so where else can we go,” said Jason Nadeau, Director, Product and Vertical Marketing at Pure Storage. “We are fundamentally an Internet of Things company, and pull in a ton of data, over a trillion array telemetry points per day in a real-time global sensor network.”

Pure1 META represents a major breakthrough in enterprise artificial intelligence and machine learning, analyzing this data to generate a new Workload DNA, which predicts both capacity and performance and provides intelligent advice on workload deployment, interaction and optimization.

“The hard part of a storage array is trying to understand the complex interrelationship of workloads,” said Sandeep Singh, Director, Product Marketing at Pure. “With Pure1 META we are learning from everyone’s arrays and everyone’s workloads.”

“Capacity is easy to predict, but for humans, performance need is not,” Nadeau said. “Many interrelated things make calculating performance very complex. The result is that companies overprovision for performance, working from the sensible position that paying more to overprovision is better than things falling down because you underprovision.”

Last year, Pure took a step towards addressing this issue.

“We take our role as an IoT company very seriously here at Pure,” Singh stated. “Last year, we introduced the Global Predictive Engine. It gets smarter day by day, but the hard part was truly understanding the workload. For years, we as an industry have been terrible at this, and any attempt to try to size for performance has been a real finger in the wind.”

Pure1 META is designed to address that dilemma.

“We are solving that problem, by bringing analytics to the table to do it,” Nadeau said. “Pure1 META combines the global sensor network with the new META AI engine. It analyzes a data lake of over seven petabytes of data to generate both Issue Fingerprints and Workload DNA. Fingerprints are something that we have been determining manually. This technology looks at over 1000 measures of performance, a figure far greater than where it breaks down for humans in predicting workloads. We will put META on this.”

The tool which turns META’s Workload DNA into actionable data is the new Workload Planner.

in Pure1. This new capability will allow customers to answer questions about new workload deployment, interaction, performance and capacity growth, and workload optimization, helping reduce risk, increase consolidation, and provide better visibility to plan for upgrades or expansions.

“The Workload Planner will show how array performance and capacity will grow over time,” said Sergei Zhuralev, ‎Software Engineering Team Lead at Pure Storage. “It looks at your workload, and every other one we have ever seen, and understands what yours looks most like. That’s just the beginning. . Whether putting new workloads on, moving them, or deciding what should go where, over time META will be able to answer all those questions.”

The META Workload Planner is the first productized version of this technology, but it won’t be the last.

“META will be expanded into broader parts of the environment,” Singh said. “With META, we are well on our way to delivering self- driving storage.”

Pure also announced Pure1’s new Global Dashboard.

“From a single dashboard, it’s a nice easy way to get key aggregate metrics from an entire fleet of arrays,” Nadeau said. It shows the total data reduction and average load performance for all arrays, along with performance trends for the last 24 hours, to enable better capacity consumption prediction.

“We think these kind of metrics are still a differentiator in the enterprise space,” Nadeau stated. “This isn’t table stakes yet. It’s basically just us and Nimble Storage who have them, and with this, we have matched Nimble and gone well beyond them.”