In this forum we would like to discuss the different aspects related to providing an infrastructure for the HPC systems we are operating. In many institutions the factors limiting HPC are not the available personnel or monetary resources but simply the available resources where new systems are to be installed.

Providing an adequate technical infrastructure for HPC encompasses different fields such as overall architecture, electricity, cooling, security etc. This is made more difficult since in most cases it is not possible to build a new data center but we have to adapt our needs to existing buildings.

Key Questions

We would like to exchange our experiences in this field and discuss some of the following points:

• What is the actual situation at our centers? Expressed in available square meters, maximal power usage, cooling technologies, PUE (Power Usage Effectiveness), etc.
• What are our forecasts (3 to 8 years) in the increase of technology of HPC?
• How will the infrastructure requirements for HPC systems grow in the next future? On what do we base our extrapolation? What will be the limiting factors?
• What are today’s best practices and problems in running an infrastructure for HPC?
• What are the trends in this field? What will our infrastructure look like in a decade?
• What problems are often experienced, such as unreliable power, cooling infrastructure problems or polluted air?
• Essential, unthought of needs, such as cable management, inventory management, configuration management, etc.

Location / Hosting

The Topic Forum will take place at Empa Akademie; Empa, Ueberlandstrasse 129, 8600 Duebendorf

Please refer to this link to get information about how to reach Empa »

Chairmanship

Daniele Passerone, Empa, tel 058 765 4877
Carlo Pignedoli, EMPA, tel 058 765 4203
Michele De Lorenzi, CSCS / hpc-ch, tel 091 610 82 08

Agenda

10:00 – 10:30 Coffee and registration

10:30 – 10:40 Greeting and introduction

Daniele Passerone (Empa)
Michele De Lorenzi (CSCS / hpc-ch)

10:40 – 11:00 Internal keynote presentations

• Pushing the borders of science: in‐depth studies and their price – In this talk, I will focus on two cases within the field of fluid mechanics, where HPC computing became a necessity. These examples draw from my own experience and serve to highlight some limitations of current HPC infrastructure, mainly related to data transfer and storage; Peter Moonen (EMPA)

11:00 – 11:45 Short presentations

• New DC@EPFL : a bet on a “passive” water cooling solution; Aristide Boisseau (EPFL)
• Infrastructure for Schrödinger: Lessons learned; Christian Bolliger (UZH)

11:45 – 12:00 Community Development, Michele De Lorenzi and Sigve Haug

12:00 – 13:00 Lunch

13:00 – 14:00 Guided tour at Empa (two tours of half an hour)

14:00 – 17:00 Short presentations (with a short break at 15:30)

• The new CSCS building; Ladina Gilly (CSCS)
• Energy-driven algorithms; Vincent Keller (CADMOS)
• Experiences from a small-size HPC server room; Vittoria Rezzonico (EPFL)
• Infrastructure for HPC @ PSI; Edgar Barabas and Derek Feichtinger (PSI)
• HPC Infrastructure @ UNIBE; Gianfranco Sciacca (University of Bern)

17:00 Farewell and end of the meeting

Observations have suggested that this shape helps them to abruptly put the greatest distance possible between themselves and their predators.

Petros Koumoutsakos, a professor at the Computational Science & Engineering Laboratory » at ETH Zurich, and two of his doctoral students, Mattia Gazzola and Wim Van Rees, tested this hypothesis for the first time using a supercomputer combining a three-dimensional flow simulation with an  evolutionary optimization algorithm.

The analysis showed that the more water is accelerated by the fish, the more they accelerated themselves. The created vortical structures play a lesser role in the escape process than had been assumed thus far. The fish use their entire body to escape (by means of the C-start), so that they can displace much larger volumes of water, to accelerating themselves.

These results can be applied for example to wind turbines and energy generation devices inspired by fish.

This simulation has been computed at CSCS Swiss National Supercomputing Centre. Read more about the simulation on the CSCS web pages »

Bibliography: Gazzola M, Van Rees WM & Koumoutsakos P: C-start: optimal start for larval fish, Journal of Fluid Mechanics (2012) 698, 5-18, DOI: http://dx.doi.org/10.1017/jfm.2011.558

Advanced Distributed Memory Parallel Programming: MPI-2.2, MPI-3.0 and PGAS

workshop to be held at CSCS in Lugano on  23-25 May, 2012.

The goal of this training workshop is to introduce MPI-2.2 performance critical topics and to provide an overview of MPI 3.0, MPI for hybrid computing and Partitioned Global Address Space (PGAS) languages, Coarray Fortran and Unified Paralell C (UPC). For lab sessions, Cray XK6, a massively parallel processing (MPP) platform with GPUs and a QDR InfiniBand cluster with Intel processors and GPUs will be targeted.

Attendees are encouraged to bring in their own applications and codes for the hands-on sessions. Representatives from MPI 3.0 forum and Cray PE will be present at the meeting for discussions and feedback.  There will be invited talks where presenters share their experiences and discuss issues in using MPI and PGAS on the CSCS systems.

Registration deadline: May 18, 2012.

Event and registration details are available here »

Instructors

• Torsten Hoefler (University of Illinois at Urbana-Champaign)
• Roberto Ansaloni (Cray Inc.)

Invited speakers

• Romain Teyssier, University of Zurich
• Angelino Paolo, EPFL
• Käppeli  Roger, ETHZ
• Will Sawyer, CSCS

Tentative agenda

First Day (May 23, 2012)

09.30 Welcome
09.40 Introduction to Advanced MPI Usage
10.00 MPI data types (details and potential for productivity and performance with several examples)
10.30 Break
11.00 Contd. MPI data types (details and potential for productivity and performance with several examples)
11.30 Nonblocking and Collective communication (including nonblocking collectives, software pipelining, tradeoffs and parametrization)
12.15 Lunch
13.30 User talks and discussion
14.30 Lab (MPI data types, non-blocking and collective communication)
15.00 Break
15.30 Contd. Lab
17.00 Wrap up

Second Day (May 24, 2012)

09.00 Topology mapping and Neighborhood Collective Communication
09.45 One sided communication (MPI-2 and MPI 3.0)
10.30 Break
11.00 One sided communication (MPI-2 and MPI 3.0)
11.30 MPI and hybrid programming primer (OpenMP, GPU, accelerators, MPI
3.0 proposals)
12.00 Lunch
13.30 User talks and discussion
14.30 Lab (Topology mapping, collective communication, one-sided communication)
15.00 Break
15.30 Lab and feedback on MPI 3.0 proposal
17.00 Wrap up

Third Day (May 25, 2012)

09.00 CAF and UPC introduction (portability and performance)
10.00 Cray programming environment and PGAS compilers
10.30 Break
11.00 Cray performance tools for MPI and PGAS code development and tuning
11.30 User talk
12.00 Lunch
13.30 Lab
15.00 Wrap up

If you wish to be considered for the allocation period starting on 1 October 2012, then you need to submit your production project proposal no later than

Friday, 11 May 2012 at 18:00  (deadline).

Please make sure that you have an active username and password so that you can download a submission form.

All details for the proposals submissions are available here »

PRACE Call for Proposals Open until 30th May

If you wish to apply for a resource allocation under this scheme then please make sure that you can fulfill the necessary conditions including the scalability and readiness criteria.

• Opening date: April 17th, 2012
• Closing date: May 30st, 2012, at 12.00 CEST
• Allocation start date: 1st November, 2012
• Allocation period: 1 year for Project Access and DECI, 1+1 year for Multi-year Access
• Type of access: Project Access and Multi-year access (Tier-0) and DECI (Tier-1)

PRACE – the Partnership for Advanced Computing in Europe – makes it possible for researchers from across Europe and the world to apply for access to Tier-0 HPC systems via a peer-review process.

This Call also invites proposals for project access to Tier-1 resources, via DECI, the Distributed European Computing Initiative, providing single project cross-national access to national (Tier-1) HPC-resources

Project proposal submissions are sought from academics who hold a position of senior researcher or higher and who are based either in Switzerland or in another PRACE Association member country or an EU country

Details of the call are available on »

Parallel IO is used by HPC applications to quickly export data to disk, but the time/space taken to save and reload the results for post-processing is often prohibitive and so data-staging is emerging as a necessary and convenient method of holding data temporarily so that reduction operations can take place using post-processing tools. In this talk we discuss staging methods/libraries and how they have been used at CSCS for coupling of simulation and analysis for computational steering, and the future developments planned to cope with the ever growing size of data produced by simulations.

The INCITE program targets large-scale, computationally intensive projects that would not be possible or productive without petascale computing. Consequently applicants for this program must present evidence that their proposed production simulations can make effective use of a significant fraction, in most cases 20 percent or more, of the machine for which resources are being requested – this typically means that the minimum starting point for consideration is to have simulations that can make effective use of the whole of CSCS flagship facility the Cray XE6 “Monte Rosa” or some equivalent system.

Details of the call for proposals are available at the following webpage »

Potential applicants should visit the FAQ page » for the INCITE call.

In particular applicants should pay attention to these questions

4. Who is eligible to participate in INCITE?

6. How are INCITE projects chosen?

12. What constitutes a “computationally intensive” research project?

13. Can I meet the computationally intensive criterion by loosely coupling my jobs?

17. Are foreign researchers and teams eligible to apply?

This year potential interested parties can register for a webinar on the proposal process with question and answer session on April 24th 2012. Details can be found here »

Information on the machines available in this call are provided on this web page » and include a Cray XK6 at Oak Ridge National Laboratory and an IBM BlueGene/Q at Argonne National Laboratory - the combined theoretical peak of these two machines will be in the order of 30 petaflop/s.

This INCITE call expects to award between 3 and 6 billion core-hours or equivalent to successful applicants in 2013, with average project allocations of around 50 to 100 million core hours (with the potential for the largest projects to receive as much as 300 million core hours).

The closing date for this call is 11:59 p.m. EDT on Wednesday, June 27, 2012 (this converts to 5:59 a.m. CEST on Thursday, June 28, 2012).

Neanderthals were used to a cool climate, but still died out during the peak of the last ice age. (Image: Homo Neanderthalensis - Homo Sapiens: A Portrait. Stefan Auf der Maur, Marcia S. Ponce de Leon, Christoph P.E. Zollikofer, 2008)

20,000 to 200,000 years before our time, with Neanderthal man and modern man, at least two types of hominid were alive on Earth at the same time. Whereas modern humans in Africa and Neanderthals in Europe occurred separately from one another, in the Near East Homo sapiens and Homo neanderthalensis lived alongside one another for around 100,000 years. Just why the Neanderthals suddenly died out more than 20,000 years ago remains unexplained to this day.

Christoph Zollikofer, Professor of Anthropology at the University of Zurich is developing with the help of CSCS simulations that could give an explication.

Read the article on the pages of CSCS »

In Manno im Kanton Tessin steht der nationale Höchstleistungsrechner im Swiss Scientific Computing Center (CSCS) der ETH. Installiert für Forscher aus der ganzen Schweiz. Ergebnisse sind beispielsweise ein Medikament gegen Malaria, Erkenntnisse über die Entstehung von Krebs oder aber auch eine detaillierte Hochrechnung der Entwicklung des Weltklimas und dessen Auswirkungen auf die Schweiz.

Der Supercomputer läuft auf Hochtouren, ununterbrochen 24 Stunden am Tag, sieben Tage die Woche. Der Standort ist nun an seine Grenze gestossen. Das Superhirn produziert zuviel Abwärme und es hat zu wenig Strom. Ein Neubau in Lugano bietet jetzt die besten Voraussetzungen: genügend Strom und Kühlungsmöglichkeiten. Das Rechenzentrum hat den Bund 68 Mio. Franken gekostet und ist eins der effizientesten und ökologischsten der Welt.

Listen the broadcast »

We are pleased to invite you to the upcoming hpc-ch forum on

Infrastructure for HPC

to be held on Thursday, May 24th, 2012 from 10:00 until 17:00, kindly hosted by Empa.

Here below you find a description of the forum and a few key questions as guideline for contributions. Additional information is published here »

To register send an e-mail to delorenzi (at) cscs.ch (registration only open to community members and guests).

Setting the Scope

In this forum we would like to discuss the different aspects related to providing an infrastructure for the HPC systems we are operating. In many institutions the factors limiting HPC are not the available personnel or monetary resources but simply the available resources where new systems are to be installed.

Providing an adequate technical infrastructure for HPC encompasses different fields such as overall architecture, electricity, cooling, security etc. This is made more difficult since in most cases it is not possible to build a new data center but we have to adapt our needs to existing buildings.

Key Questions

We would like to exchange our experiences in this field and discuss some of the following points:

• What is the actual situation at our centers? Expressed in available square meters, maximal power usage, cooling technologies, PUE (Power Usage Effectiveness), etc.
• What are our forecasts (3 to 8 years) in the increase of technology of HPC?
• How will the infrastructure requirements for HPC systems grow in the next future? On what do we base our extrapolation? What will be the limiting factors?
• What are today’s best practices and problems in running an infrastructure for HPC?
• What are the trends in this field? What will our infrastructure look like in a decade?
• What problems are often experienced, such as unreliable power, cooling infrastructure problems or polluted air?
• Essential, unthought of needs, such as cable management, inventory management, configuration management, etc.

Location / Hosting

The Topic Forum will take place at Empa Akademie
Empa, Ueberlandstrasse 129, 8600 Duebendorf

Please refer to this link to get information about how to reach Empa »

Chairmanship

Daniele Passerone, Empa, tel 058 765 4877
Carlo Pignedoli, EMPA, tel 058 765 4203
Michele De Lorenzi, CSCS / hpc-ch, tel 091 610 82 08

Agenda (preliminary)

10:00 – 10:30 Coffee and registration
10:30 – 10:40 Greeting and introduction, Daniele Passerone (Empa) and Michele De Lorenzi (CSCS / hpc-ch)
10:40 – 12:00 Presentations
12:00 – 12:05 Community Development, Michele De Lorenzi
12:05 – 13:00 Lunch
13:00 – 14:00 Guided tour at Empa
14:00 – 17:00 Presentations (with a short break at 15:30)
17:00 Farewell and end of the meeting