“Being able to create a picture in 200 cities or settlements in a short amount of time is dramatically different from what we can do with our existing compute resources,” said Brian King, coordinator of CGIAR’s Platform for Big Data in Agriculture.

“Since the impacts of COVID-19 are unfolding differently by country, our ability to look at the situation both at the aggregate level and from an on-the-ground, local view is incredibly valuable. That capability enables a different way for us to operate as a research organization. But generating high-frequency insights across multiple distinct contexts at once demands compute power to support it and more compute capacity than we had. The Memory-Driven Computing Sandbox appeared at just the right time.”

While CGIAR has high-performance computing clusters at several of its Centers, it is seeing increased need to develop timely, localized information and analysis across an array of food security contexts in light of the pandemic, and this is beyond its existing compute resources. The Memory-Driven Computing Sandbox sets itself apart by giving every processor (up to 64 sockets) in the system access to a giant shared pool of memory – up to 48 terabytes – which is a sharp departure from today’s systems. Typically, relatively small amounts of memory – just a few terabytes – are tethered to each processor; the resulting inefficiencies limit performance. By having all of the massive, diverse data sets available at one time in memory, users can clear computational bottlenecks that hinder research and discovery.

With access to the Sandbox – so named for the controlled environment it offers customers to experiment with advanced compute resources – CGIAR is building a cross-cutting, high-frequency views of food systems linking crop modeling – including weather records and how crops performed and what the yield was, by year and location – survey data, and overall economic activity (e.g. movement of goods and people). CGIAR is monitoring emissions from up to 1,000 points across India and East Africa using public satellite data from space agencies. Changes in emissions indicate changes in economic activity that give researchers important context for understanding how food security challenges are unfolding by location. Equipped with this dynamic, unfolding picture, CGIAR is able to compare with crop and survey data to monitor how that individual crop will impact the broader food supply.

“With CGIAR, we saw the opportunity to apply innovative technologies, like HPE’s Memory-Driven Computing Sandbox to drive faster insights and help address this incredibly complex challenge.” – Janice Zdankus

Insights from this data will enable CGIAR to see and increasingly predict how food security challenges are unfolding from the COVID-19 crisis to inform policymakers, food relief actors, and other stakeholders. Using CGIAR’s existing technology, emissions analysis on one point on the Earth could take four to five hours to run. Today, CGIAR can run multiple analyses over multiple points with sufficient frequency to inform timely action on food security.

“At HPE, our purpose is to advance the way people live and work and we are committed to applying technology to help address some of society’s toughest challenges,” said Janice Zdankus, VP, Innovation for Social Impact, HPE. “One of our focus areas is world hunger, inspired by results from Purdue University’s 1,400-acre research farm and its application of precision agriculture to increase crop yields while drastically conserving resources. With CGIAR, we saw the opportunity to apply innovative technologies, like HPE’s Memory-Driven Computing Sandbox to drive faster insights and help address this incredibly complex challenge.”

While the pandemic has caused immediate and near-term issues that must be addressed, the future of food security must be evaluated on a longer horizon as well. Mapping and predicting climate risk, vulnerability, and adaptation options are top priorities for CGIAR.