PRACE, the Partnership for Advanced Computing in Europe
Partnership for Advanced Computing in Europe (PRACE) is a persistent, pan-European Research Infrastructure (RI) for High-Performance Computing (HPC) managed as a single European entity.
Supercomputers are indispensable tools for solving the most challenging and complex scientific and technological problems through simulations. To guarantee that the European science community will maintain its international competitiveness, it has to have access to the world-class supercomputer systems. Through PRACE, European scientists and technologists can use world-leading supercomputers, the capacity of which is equal or superior to those in the USA, China or Japan.
There are about 30 European supercomputer sites from 25 countries participating in PRACE. Each site dedicates part of their supercomputer resources to PRACE usage and those resources are divided into two tiers. Tier-0 is formed by four of the biggest supercomputer sites, and all the rest of them form Tier-1, which is a continuation of the DEISA EU project (2004-2009).
More general information about PRACE can be found on the PRACE-RI web pages at http://www.prace-ri.eu/.
PRACE and CSC
CSC participates as a Tier-1 site. CSC allocates 6 % of it's CRAY XC40 supercomputer, Sisu (sisu.csc.fi), for PRACE usage; the users are researchers throughout Europe, for the most part outside Finland.
PRACE Tier-1 usage consists of Distributed European Computing Initiative (DECI) projects, which are grand challenge type of HPC computing projects by leading European researchers. When projects are accepted as a DECI projects, each of them assigned to the most suitable supercomputer site (the execution site). Since the whole idea is to use most suitable architecture for the projects, usually Finnish researchers have projects which are assigned to some other site (than CSC and Sisu) as the execution site; this makes it possible to run a project on a supercomputer architecture that is not available in Finland.
In addition, Finnish users can also run their projects at Tier-0 sites. That way Finnish researchers can run their projects on Europe's biggest supercomputers.
Applying for PRACE resources
You may apply for PRACE resources through a procedure called "the calls", which are issued several times a year and they are announced on the PRACE-RI website at Call announcements (Tier-0) and the DECI access (Tier-1) pages. If you project is one of the awarded projects, below is described some practical information for the services and facilities that are most likely required; no project needs everything.
Getting more information and help
The PRACE Helpdesk is the first point of contact if you have any issues with or questions about PRACE services and facilities. It is recommended that the Helpdesk functionality is accessed through the dedicated web interface. If you do not have access to a web browser or prefer to use e-mail you can also send questions and requests by email. For instructions on contacting the Helpdesk, see the PRACE Helpdesk guide. The frequently asked questions are listed on the PRACE User FAQs page
A personal GRID certificate is usually needed
An important remark at this point is that almost all PRACE services use grid certificates for authentication, not username/password pairs. One example is the interactive access to HPC resources, you can use a certificate authentication even to login to supercomputers. Some sites do not offer other methods. It is strongly advised to acquire the certificate as soon as possible. This will expedite the start of a PRACE or a DECI project. See the Certificates page for instructions on how to get the certificate.
It is strongly recommended (read: required) that you do not store your certificate's private key on publicly used servers, but use the MyProxy service instead. With MyProxy you can store and retrieve a copy of your proxy certificate without having to transfer your private key to a networked system. For instructions on how to use MyProxy, see the documentation of the PRACE MyProxy service.
Running compute batch jobs and transferring data
The most common way to run jobs on execution supercomputers is to login (see below) to them and use the local batch queuing system to submit jobs. The alternative method is to submit the jobs from your own workstation using the UNICORE software. As for the data transfers, SSH or GSI-SSH is a good method if you have a fairly small amount of data to transfer. If you have more, PRACE offers more powerful tools for data transfers: GridFTP and gtransfer. For information about and availability of UNICORE, GridFTP and gtransfer, see the user documentation pages on the PRACE-RI website.
Interactive access to supercomputers
Interactive access is provided to the supercomputer for which the access is granted as part of a PRACE or a DECI project allocation. The available access methods vary from site to site. Many sites offer only GSI-SSH, which is a variant of SSH that does not use passwords or public keys for authentication but grid certificates. For instructions on using GSI-SSH, see the GSI-SSH clients documentation on the PRACE-RI website.
For information about which access methods are available at your execution site, see the Interactive access to HPC resources page on the PRACE-RI website.
The primary service for getting information about your CPU time usage is DART, the Distributed Accounting Reporting Tool. Local usage reporting services are also available on many PRACE sites. For information about DART, see the documentation of the Distributed accounting reporting tool on the PRACE-RI website.