The Helvetios cluster has just been delivered (December 7th, 2018) and we are working with the vendor to put it in production.
Helvetios is an Intel Xeon based cluster, which will be available to the EPFL research community around February 2019.
- Peak performance Rpeak: 763 TFLOPs
- 10’368 cores total
- Total RAM: 54TB
- Storage (scratch): 200TB all-flash
- Infiniband EDR 100Gb 2:1 (1:1 on each 24-node chassis)
The cluster is composed of
- one front end,
- two master nodes in charge of administration and backup,
- two dedicated data movement/transfer nodes,
- 287 compute nodes, each with
- 2 Skylake processors running at 2.3 GHz, with 18 cores each,
- 192 GB of DDR3 RAM,
- 800GB NVMe local drive
Production will be preceded by an early-adopters phase, during which the system will be available for large calculations. To sign up for the early adopters program please respond to this message by letting us know the following:
- The scientific nature of your runs
- Maximum number of CPU cores for your runs
- Application you are going to run
- Modules used
- Maximum wall clock time
- Estimated core hours needed in total
Since this is a pre-production phase, we warn you that we are not ensuring any level of service. This means: the cluster could be rebooted, some modules may not work. In addition to this, we will ask you to report any issues you may encounter and provide us with a report on your runs.