Cray has been awarded contracts with Eclipse Holdings Ltd. to upgrade the supercomputing system operated by the South African Weather Service (SAWS). SAWS has been using a Cray XC30 supercomputer and Lustre file system to perform its operational weather forecasting and related analyses since 2014.
SAWS fields requests from government agencies, aviation operators, and other industries. Over the last few years, South Africa has faced a number of extreme weather events, including droughts, floods and storms. The ensuing requests for impact analysis and higher-resolution forecasts effectively doubled the burden on SAWS’ computing capacity. Beyond enabling SAWS to improve their forecasting, the new upgrades will facilitate the handling of increased workload demands. This is especially important since SAWS is the only World Meteorological Organization (WMO)-designated long-range global modeling center in Africa.
“SAWS has experienced great success in the last three years using its existing Cray XC30 supercomputer and ClusterStor storage system,” said Nico Meintjes, CEO at Eclipse Holdings, Ltd. “As SAWS began to receive more tasks from the South African government, doubling computing needs, this put significant strain on system performance. With these growing needs, expanding and upgrading its Cray solutions was clearly the strongest choice.”
The XC30 system – which was Cray’s first-ever installation in Africa – was expanded from 88 Ivy Bridge nodes to 172 Ivy Bridge nodes, adding an infusion of 2697v2 Xeons (12-core @ 2.70 GHz) to the original 2695v2 (12-core @ 2.40 GHz) SKUs. In addition to doubling its compute power, SAWS tripled its storage capacity: going from .52 petabytes @ 12 GB/s with the 2014-era Sonexion 1600 to 1.8 petabytes @ 27 GB/s with the new ClusterStor L300 product.
The weather forecasting system, which is air-cooled, expanded from 1.5 to 3 cabinets (housing 48 blades) and utilizes Cray’s Aries interconnect. The upgrades were installed in March 2018 and the system was relocated to a new datacenter the following month.
According to Ilene Carpenter, earth sciences segment director at Cray, the choice was made to upgrade (rather than replace) the system to minimize downtime. The system had never experienced downtime while in service, and with no application porting or revalidation work to tackle, the upgrade process limited service interruptions to just four days – a crucial factor, considering the urgency of weather forecasting.
“It’s an exciting time in weather forecasting as organizations embrace artificial intelligence to more accurately predict forecasts and weather patterns,” said Carpenter. “As weather simulations grow more advanced and comprehensive, generating weather predictions is an increasingly data-intensive task. With the Cray system upgrades in place, SAWS has the storage and compute resources needed to handle an increasing number of hydrometeorological observations and to run higher fidelity weather and climate models.”