HPC cluster Baobab
What is Baobab?
Baobab is the new HPC cluster, available to all researchers from the university of Geneva.
With ~2400 compute cores Intel Sandy Bridge, 10TB of RAM, it is a very powerful cluster for massive parallel computations: Its computation nodes are interconnected by high speed infiniband network links.
Baobab was purchased end of 2012 with a special funding from the University's rectorat, and is operational since early 2013.
The initial size of Baobab was ~1000 cores, but it is regularly extended with the addition of "private" compute nodes.
The cluster is managed by the DiSTIC. If you have any question, feel free to send an e-mail to
Baobab is composed of :
- a master node in charge of administration and backup ;
- 74 compute nodes, each with :
- 2 Sandy Bridge Intel(R) Xeon(R) CPU E5-2660 @ 2.20GHz cpu with 8 cores each
- 64 GB of RAM,
- 78 compute nodes, each with :
- 2 X5650 CPU @ 2.67GHz with 6 cores each
- 48 GB of RAM,
- for a total of 936 cores;
- 1 compute nodes, each with :
- 4 CPU Sandy Bridge Intel(R) Xeon(R) CPU E5-4640 @ 2.40GHz with 8 cores each
- 512GB of RAM,
for a total of 32 cores;
- three storage node running FraunhoferFS (BeeGFS) of 20 TB each for a total of 60 TB (extension December 2013) ;
- InfiniBand 4xQDR (40Gbit/s) connectivity between master, nodes and storage.
Interested in Baobab? Request access!
- Your UNIGE official email;
- The faculty you are affiliated to;
- The UNIGE email of your Primary Investigator (research group leader);
- A short description of the intended use of the cluster (with at least the research area, the software used, and the expected job duration)
Sponsoring an extension of Baobab
Interested research groups can sponsor the extension of Baobab by purchasing a compute node. Purchasing a compute node grants you valuable advantages:
- A waiting time of 12h maximum before running your job on your node.
After this initial waiting time, you have an access to your node until you stop using it.
- Submitting jobs of up to 7 days
For normal users, the duration of one job is limited to 4 days. As an owner of a node, you can submit jobs that last up to 7 days. And upon request up to 14 days.
- A professional management of the node, with regular software upgrades
Stored in Geneva, at Uni Dufour, the cluster is professionnally managed by our teams and upgraded regularly
- Serving the community
When you purchase a node, you also bring value to the community: When you do not use your node, it is available for other researchers, for jobs of up to 12 hours.
Consider approximately 5'000 CHF TTC for one 16 cores node, 64GB RAM, infiniband interface, 240GB SSD hard disk.
Contact us for more information
Click here to access the end-user information.