Our institute has access to several computers or supercomputers. Each computer is dedicated to a particular usage. It is important to identify what solution suits you the best. Here is a description of these different systems, so you can determine which one is the best for you.
- Piz Daint: this is one of the largest supercomputer in the world (ranked #3 in the November 2017 www.top500.org list). It is operated by the Swiss supercomputing centre (CSCS). Each professor in the institute has an account on the multicore partition of Piz Daint. These projects are labelled uzh1, uzh2, uzh3... Ask your supervisor to get you a login for the corresponding project. Piz Daint is dedicated to massively parallel jobs, so your code needs to be MPI-parallel. Piz Daint also provides dedicated nodes with GPU acceleration, but this is restricted to the CSCS competitive allocation schemes (more info here).
- Hydra: this is a collection of large memory nodes. Each node has 3TB of memory and 64 shared-memory cores. It is operated by the Science Support and Service for IT (S3IT) team of the University of Zürich. Access to these nodes is limited, you have to ask your supervisor. Hydra is dedicated for large analysis jobs requiring massive amounts of memory. Ideally, your codes have to be MPI- or OpenMP-parallel.
- Science Cloud: this is a large ensemble of small memory nodes that can be used in "cloud computing" mode. You can very easily set up your own instance and manage it as if you were the administrator. This instance can be saved and re-used later. Each group at the University has access for free to a minimum amount of resources. Again, ask your supervisor how to access these resources.
- Vesta: this is a small GPU cluster operated by S3IT. Access is restricted to very specific applications. You code must be written in CUDA or OpenACC.
Each of these computers is meant to be used for a certain type of applications (MPI, OpenMP, CUDA). These are large systems used by thousands of users, so you need to be careful and respectful of the other users. Training on how to use these systems is provided by S3IT and the ICS every spring during the HPC course (ESC 401).
