Our high-performance computing facilities are highlighted here.
ECMWF's multi-petaflops supercomputer facility is designed for operational resiliency featuring two Cray XC40 systems and independent Cray Sonexion storage systems. The system comprises two independent subsystems located in separate halls. It has separate resilient power and cooling systems to protect against a wide range of possible failures. Each subsystem consists of 20 Cray XC40 cabinets equipped with Intel Broadwell processors and around 3,600 dual-socket compute nodes per system, a number of Cray Development and Login nodes and around 10 petabytes of Lustre storage with the ability to cross mount the Lustre file systems between the halls.
Ecgate is a Linux cluster, available exclusively to registered users. It is a general-purpose facility to process computing jobs that are not suitable for the supercomputer.
Data handling system
We operate a large-scale data handling system (DHS) where our users can store and retrieve the data they need to perform weather modelling, research in weather modelling and mining of weather data. The DHS supports both our powerful Meteorological Archive and Retrieval System (MARS), and our file storage system (ECFS) for non-MARS files.
The local area network (LAN) connects all our computing systems. The LAN is divided into a high-performance network (HPN) and a general-purpose network (GPN). The HPN transfers large amounts of operational data between the supercomputer and the data handling system. The GPN provides connectivity to the supercomputer, the data handling system, other large servers, user desktops and laptops and via firewalls, to the outside world.