Finnish Grid Infrastructure

FGI: Distributed Computing Capacity for Science

In Brief

The FGI is the Finnish Grid Infrastructure, a national distributed computing environment and the National Grid Infrastructure (NGI) for Finland. The idea was to join a number of computing clusters at different places, which are logically combined into a grid. The total aggregated peak capacity of the infrastructure is 154 TFlop/s, or which 51 TFlop/s is CPU capacity and 103 TFlop/s is GPGPU capacity.

The project originally started in the winter of 2004 resulting in the M-grid which has been extended in 2011 to form the FGI. The Academy of Finland has participated in the funding through a grant in its FIRI 2010 Infrastructure call. The FGI consists of CSC and the following nine universities: Aalto University, University of Helsinki, Lappeenranta University of Technology, Tampere University of Technology, University of Eastern Finland, University of Jyväskylä, University of Oulu, University of Turku and Åbo Akademi University.

The FGI hardware procurement was done in 2011 as a joint effort between the participating universities and CSC. System administration of the clusters/Grid is done in co-operation between CSC and the university partners. The FGI will be available for researchers in early 2012.

Purpose

The available resources are used both locally by CSC/University users and through the grid middleware.

The grid middleware provides scientists with the computational capacity for serial and parallel jobs, as well as, jobs that require huge amounts of memory. Via the FGI, users are able to run their applications flexibly on computing resources hosted by different institutions. In addition to self compiled applications the FGI has a selection of pre-installed and centrally managed applications. The grid middleware automatically chooses a suitable resource.

Submitting grid jobs and retrieving results is not instantaneous. The FGI is ideally suited to calculations which take from tens of minutes to up to 24 hours. Job related data is transferred with the program to the cluster on which it will run. Results can be retrieved when the job has completed.

Jobs that run on the FGI are not billed from the users CSC project budget.

Grid jobs can also fill the unused capacity on the clusters.

Help and further information can be requested from grid-support(at)csc.fi.

Instructions are available on the FGI User guide and FGI user pages