Already home to three Cray XC40 systems (the last one deployed in 2016), the Met Office, a leading weather center in the U.K., has now added Cray’s Urika-XC suite of AI and analytics tools according to a Cray announcement today. Like many system suppliers, Cray has been pushing hard to gain a foothold in delivering AI/analytics enabling technology. This latest win is evidence growing traction in that market.
The Met Office is able to take in 215 billion weather observations from all over the world every day and uses an advanced atmospheric model to create tailored forecasts and briefings that are delivered to a wide range of customers such as governments, environmental agencies, the military, and businesses.
“As in many industries, we are challenged with increasing data volumes and are turning to large-scale analytics, machine learning and deep learning applications to drive new insights and innovation,” said Charles Ewen, director of technology and CIO at Met Office in the official announcement. “The Met Office already has one of the world’s largest Cray XC supercomputing systems. Now with our implementation of Cray’s Urika-XC software, we are applying AI and analytics to deliver ever-more accurate and detailed weather forecasts and climate change analyses, while also developing new commercial products.”
Per Nyberg, Cray VP of market development, AI and cloud, said, “Cray and the Met Office share a long, productive and successful relationship. The Met Office’s decision demonstrates its confidence in Cray’s innovation. We believe big data analytics, modeling and simulation are converging into new workflows leading to powerful insights for customers.”
The Cray Urika-XC suite includes:
- Big data analytics tools with Apache Spark and the Spark ecosystem running at supercomputer scale. Includes languages (R, Scala, and Java) and libraries (MLlib, GraphX, and the Intel BigDL deep learning framework for Spark).
- Data science tools: Python-based tools and libraries for data scientist productivity (Anaconda and Distributed Dask) and Jupyter notebooks for end-user productivity.
- Deep learning applications like TensorFlow enhanced to run in a distributed fashion using 1,000 CPUs or GPUs, with simplified setup and administration.
- Graph databases using Cray’s powerful Cray Graph Engine (CGE)for big data graph pattern recognition and discovery applications. For the most demanding graph problems where size and performance matter, CGE has been tested against a 400-billion-tuple collection and shows a nearly 100x performance speed-up over Spark-based GraphX.