Today's life scientists need to have advanced High Performance Computing (HPC) facilities at their disposal. For that reason, and commissioned by NBIC, SURFsara places powerful computer clusters at the local sites of interested universities and academic hospitals. These clusters can then be used for experiments. Together with the user community, we take care of support.
For some scientific disciplines, such as high-energy physics and quantum chemistry, High Performance Computing (HPC) is part of the standard toolkit. For other scientific disciplines, for example the Life Sciences, this is not yet the case. Yet, HPC can be of big interest here as well.
To outsiders, HPC and especially Grid technology may seem daunting. With the Life Science Grid project, SURFsara aims to close the gap between the Life Sciences and HPC facilities. Accordingly, computer clusters are built at various locations throughout the Netherlands. We maintain these clusters from a distance, allowing users to fully focus on their research.
Currently, the Life Science Grid is able to run more than 2700 jobs concurrently and together with the BiG Grid central facilities (accessible for the Life Science Grid users too) this adds up to almost 10000 potential job slots. Clusters have been placed in Amsterdam (at the AMC, NKI and VU), Nijmegen (Radboud), Wageningen (WUR and KeyGene), Leiden (LUMC), Utrecht (Hubrecht), Groningen (RUG), Delft (TU), Rotterdam (Erasmus MC) and Eindhoven (HTC). All clusters are interconnected by the fast SURFnet network as well as optional dynamic light path for secure and guaranteed-bandwidth connections between clusters. Locally, every cluster can be utilized as a traditional computer cluster, thus leveraging significant computational power and data storage.
The true force of the Life Science Grid, though, is that the clusters together form a Computing and Storage Grid, a virtual supercluster. Applications running locally on the cluster can also be executed in a distributed fashion on the entire Grid.
By placing the computer clusters on site, we facilitate tight integration with local ICT infrastructure and present measuring equipment. Participants may share their data, but can also protect their data from others and keep it within the premises. On their local cluster, participants have extra permissions. E.g., one can log in on the local cluster, which in some cases greatly simplifies debugging of (self-built) applications.
Currently, twelve clusters are in place at the following sites:
A typical cluster has 128 compute cores, 512 GB of system memory, 10 TB of scratch space, and 18 TB of storage and staging area.
The Life Science Grid is available for all life scientists working in the Netherlands. Each cluster has its own login node that is tailored for the local users to prepare, test, launch and manage their jobs. A central login node to the Life Science Grid is available for all other life scientists (at institutes other than those mentioned above), including life scientists working in commercial organizations (provided that the infrastructure is used for pre-competitive research only).