How Cray Climbed the Weather Forecasting Market Charts

Although weather forecasts are nested into the fabric of our daily lives, it can be easy to forget that to arrive at accurate, long-term predictions takes massive amounts of computing power.

Most countries have government-sponsored weather centers that are tasked with prediction and regional climate research and these require Top 500-class supercomputers to produce.Of course, it takes more than a bevy of powerful nodes to deliver spot-on forecasts across broad geographic areas–it requires extensive tuning of the legacy codes on an architecture that can match the low latency needs of weather prediction models.

While weather forecasting centers generally do not tap high performance computing accelerators for prediction models or have other exotic system requirements for their forecasting codes, they do need large, finely-tuned systems that are optimized around the limited, but complex codes these centers use for their production forecasts. This means they need architecture, system, software, and application-centric expertise—a set of demands that few supercomputing vendors have been able to fill to capture massive market share in the way Cray has managed to do over the last ten years.

Cray has historically been a very research and development focused company, which is the top reason why they have managed to succeed in specific areas like weather, says the company’s Barry Bolding. Ten years ago, the company made its first significant investment in weather experts to hone the systems for numerical weather prediction and other models. The efforts of this team kicked off one of its first large weather system deals at the Korean Meteorological Administration (where they are still supplying supercomputers), and has grown to include an ever-increasing share of the weather systems market.

At the time of the Korean weather system, Cray had a ten percent slice of the weather market worldwide. A decade later, they are hands-down the top supplier of weather forecasting machines with what Bolding says is between a 60-70% share—a fact that continues to drive their investments in continued research and development in that area. “Our investors understand that research and development is what really differentiates us,” he said. And although it took ten years to climb to the peak of the weather market with its large system contracts, the investments in this space are beginning to bear more fruit–and often outside of its native U.S.

There is another reason why system deals in the weather forecasting are desirable for any supercomputing vendor. One contract is closer to the price of two–what is notable about the weather machines is that large centers for forecasting and climate research generally always buy two duplicate machines to ensure the continuity of forecasts and to have a secondary research cluster to run background workloads in parallel with the backup forecast ensembles. For instance, one of Cray’s largest weather supercomputers to date at the European Centre for Medium-Range Weather Forecasts (ECMWF) occupies both the numbers 38 and 39 on the Top 500 (even though it is a matching set) and in South Korea, a dual-weather cluster rests at numbers 216 and 217 to power the Korea Meteorological Administration’s forecasting efforts (along with a slightly larger, newer cluster, the Uri machine, which is an XC40).

Earlier this year, Cray announced another significant weather win in the United States with the a win at NOAA in the U.S., another for the Met Office in the UK for $156 million, and today topped another milestone by announcing a contract worth up to $53 million for a new Cray XC40 supercomputer at the Bureau of Meteorology in Australia. The center has had a range of supercomputers over the years to support its numerical weather prediction models, including two earlier generation Cray machines (an X-MP and Y-MP), an NEC SX-6, and an Oracle/Sun system. Although no petaflop capabilities were given today, this is not a small machine—the Cray Sonexion storage system alone will provide more than 12 petabytes of storage and the overall system will allow the center to run nearly eight times the number of forecasts it can do with its current machines.

When asked about why this is when there are plenty of other vendors who are offering supercomputers at lower price points featuring the same processors, he said it is a matter of expertise and tuning across the entire system. Indeed, while others might all have the same Knights Landing or Haswell processors, the optimizations and weather application expertise extends across the entire system.

This expertise is driven by weather application experts on Cray’s research and development team who can fine tune around their unique architecture against the couple of main weather forecasting and prediction codes that are available. While the codes might be limited, Bolding says there are various plugins and model differences based on certain geographies that create variance, thus require tuning on the Cray side to lend performance.

Cray has been on a roll over the last couple of years in particular, garnering an ever-larger number of systems on the Top 500 as well as in enterprise. As of the most recent Top 500 list, the company claimed 71 machines, including the #2 system, Titan, at Oak Ridge National Lab and two others in the top ten.

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now

Be the first to comment

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.