Cern using OpenStack to avoid running out of computing capacity


Cern, the scientific organisation which runs the Large Hadron Collider, popularly known as the atom smasher, is using OpenStack to create a public-private hybrid infrastructure.

Cern’s computing infrastructure has becoming increasingly complex and difficult to manage, and IT managers are considering their public cloud options.

LHC is one of the largest generators of data in the world. The tunnel is which the atoms are smashed together is approximately 27 kilometres long, and there are currently at least four experiments running there – generating around 9 gigabytes per second.

That data is sent to Cern’s on-site data centre, referred to as “Tier 0”, which processes approximately 1 petabyte of data per days.

The Tier 0 data centre has around 11,000 servers with 100,000 processor cores.

Beyond Tier 0, Cern operates what it calls its Worldwide LHC Computing Grid, which links more than 170 data centre facilities in more than 40 countries.

After outgrowing its own on-site computing capacity, Cern awarded T-Systems a cloud services contract last year to create a private cloud, called the Helix Nebula Science Cloud.

Currently, approximately 8,000 scientists around the world have near-real-time access to LHC data. They work on many different but interconnected experiments.

T-Systems’ solution is based on OpenStack, the software that enables the management of public and private cloud infrastructures.

Cern’s private stack currently consists of 8,500 servers with 280,000 processor cores, which run at least 30,000 virtual machines.

But, having tried to keep everything on-site and then in its private cloud, Cern now says it is outgrowing its private cloud and is looking to integrating public cloud.

In comments reported on IT Online, Jan van Eldik, Cern’s resource provisioning team leader, says: “We realised that we are soon going to run out of CPU capacity, and have started to look at extending the private cloud into the public cloud.

“Public cloud is attractive to boost resource provisioning.”

T-Systems is working with Huawei to develop the public cloud service for Cern. Other companies involved in the project are HPE, IBM, Advania, Sixsq, and Indra.