The next IGT HPC Work Group meeting

The next Israeli Association of Grid Technologies High-Performance Computing Work Group meeting 13:45-16:00, December 9th , 2009 IGT Office Maskit 4 Hertzeliya Agenda: 13:45-14:00 Opening by Avner Algom and Guy Tel-Zur 14:00-15:00 Prof. Tilo Wettig, Ph.D., University of Regensburg, Germany - The QPACE supercomputer (Live via Webex) 15:00-15:15 Coffee break 15:15-16:00 Dr. Guy Tel-Zur, IGT – Challenges in teaching Parallel Computing Program: 14:00-15:00 Prof. Tilo Wettig, Ph.D., University of Regensburg, Germany - The QPACE supercomputer Prof. Tilo Wettig will give an overview of the QPACE supercomputer, which was developed by a collaboration of several academic institutions and the IBM R&D lab Boeblingen, Germany. QPACE is a massively parallel architecture, allowing a single application to run on thousands of nodes. Individual nodes consist of a PowerXCell 8i processor and an FPGA that functions as network coprocessor. The nodes are connected in a 3-dimensional torus. The target application is quantum chromodynamics (QCD, thus the name QPACE = QCD parallel computing on the Cell), but the machine is also suitable for other mesh-based applications. QPACE is currently #1 on the Green 500 list, i.e., it is the most power-efficient supercomputer in the world. 15:15-16:00 Dr. Guy Tel-Zur, IGT – Challenges in teaching Parallel Computing We are now in the Multi-Core era; however there are not enough graduates from Computer Science and Computer Engineering departments who luck an elementary knowledge in Parallel Computing. The talk will cover challenges in teaching an introductory course to Parallel Computing. Such a course is given by the speaker at the Ben-Gurion University in Beer-Sheva to Electrical and Computer Engineering students. The Syllabus and the infrastructure for teaching this course will be described and a few demos will be presented. To reserve your place, please send your details to: info@grid.org.il

Date Dec 9, 2009 13:45 16:00 Location IGT Offices

Parallel Computing on Windows

There are many ways to use MPI. In this post I show how to install and use the DeinoMPI implementation which looks very cool and is free.
The screen captures in this post were taken using the nice SnapIt tool.

Part 1: Installing DeinoMPI

After downloading and starting the msi file:








Part 2: Configuring DeinoMPI
2.1 Start the deinoMPI daemon:


2.2 Credentials


Part 3: Testing
In this part I show how to compile a MPI program with the free Bloodshed DevC++ IDE.
I use the famous cpi.c code


This installation is "Local Only" (perhaps in one of my future posts I will show how to use this tool with more than one node):


The Windows Firewall is noticing the new player:


Here is the execution and output window:


It is nice to see the Task Manager showing the 10 parallel running processes:


Part 4: A little-bit Mathematics

The integration in cpi.c is done for f(x)=1/(1+x**2) between 0 to 1.
It is nice to replace this function by f(x)=sqrt(1-x**2) also between 0 to 1 (the unit circle).
Strangely enough, integration of either of these functions between 0 to 1 is equal to Pi/4.
The Sage Symbolic Mathematics tool is a good way to show the difference between the two functions:

[mpich-discuss] Errors while testing MPICH2 on Vista

Guy Tel-Zur gtelzur at gmail.com
Sat Oct 24 14:39:26 CDT 2009

• Previous message: [mpich-discuss] MPI-3 Fortran feedback
• Next message: [mpich-discuss] MPI function to test a machine alive or not.
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

---

I am trying to test a fresh installation of MPICH2 (version 1.2) on Vista32
and get the following errors:

c:\temp>"c:\Program Files\MPICH2\bin\mpiexec.exe" -n 4 cpi.exe
Please specify an authentication passphrase for smpd: <*I type here the
passphrase, e.g. behappy*>
Error while connecting to host, No connection could be made because the
target machine actively refused it. (10061)
Connect on sock (host=user-PC, port=8676) failed, exhaused all end points
Unable to connect to 'user-PC:8676',
sock error: Error = -1

I will appreciate any comment.

--Guy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20091024/49beeb2c/attachment.htm>

---

• Previous message: [mpich-discuss] MPI-3 Fortran feedback
• Next message: [mpich-discuss] MPI function to test a machine alive or not.
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

---

More information about the mpich-discuss mailing list

IGT HPC WG Meeting

The next IGT HPC Work Group meeting

14:00-16:30, November 4^th , 2009 IGT Office Maskit 4 Hertzeliya

Agenda:

14:00-14:15 Welcome talk Avner and Guy

14:15-15:15 Mr. Lior Amar and Mr. Tal Maoz – Clusterlogic: CLiP & Infod -- HPC Cluster Management and Monitoring

15:15-15:30 Coffee break

15:30-16:30 Mr. Nadav Rotem, Haifa University - Introduction to Reconfigurable Computing

Program:

14:15-15:15

CLiP & Infod -- HPC Cluster Management and Monitoring
Mr. Lior Amar and Mr. Tal Maoz
Abstract:
ClusterLogic (www.clusterlogic.net) specializes in providing advanced solutions for HPC Linux clusters. In this talk we will present two of our products, aimed at making the task of cluster management and monitoring easier:

CLiP
CLiP is a Cluster Installation and Management system which enables the rapid installation of diskless (and diskfull) clients. It is very flexible and enables the administrator to setup multiple desired configuration with only a few clicks.
The client nodes boot using the desired configuration and in a few minutes the cluster is ready for use.

In this talk we will perform a live demonstration of CLiP.

Infod

Infod is a gossip based information dissemination system, designed to give a robust cluster-wide information exchange platform. Nodes running this system routinely exchange information with other (randomly selected) nodes in the cluster. The outcome is that each node maintains a partial (or full) view of the cluster. This view can be used for
load balancing or monitoring and is constantly updated.

In the talk we will present the concepts behind the system and a live demonstration of the system will be performed.

15:30-16:30
Introduction to Reconfigurable Computing

Mr. Nadav Rotem, Haifa University

Abstract:

Reconfigurable computing is a paradigm which combines the flexibility of software with the high performance of hardware. Recently we see more and more FPGAs in High Performance Computing. Problems which were previously solved by either software or custom hardware machines are now done by hybrid reconfigurable systems. In this talk I will introduce FPGAs, present their architecture, their programming tools and design process. I will present commercial hybrid systems which combine reconfigurable components with traditional clusters. I will present High-Level Synthesis, which is a method for programming reconfigurable devices using high languages such as C. I will present several reconfigurable applications and compare their performance with traditional cluster implementations.

To reserve your place, please send your details to: info@grid.org.il

Schrödinger's cat - 2009 Edition

What are the chances that you will be found alive if you entered an Israeli train?
Well, according to the recent Israeli trains accidents it is not high.
Here are photos I took today of the train that went off the tracks near Kfar Menachem:

Cloud-based Services and Applications

Chairs:
Prof Mark Baker, University of Reading
Dr. Marcel Kunze, Forschungszentrum, Karlsruhe

Abstract

Cloud computing is increasingly being used for what was known as “on-demand” and “utility computing”. The services provided, the APIs and the applications that can be hosted by these Cloud providers have superseded the use of the grid, and are increasingly becoming popular with users. There are obviously two sides to the services that are provided by Cloud providers: those that are supplied by commercial entities, such as Amazon and Google, as well as those that are open-source systems, such as provided by OpenCirrus and Eucalyptus.

In this workshop we wish to examine and explore the services, interfaces and types of applications that can be executed on Cloud systems. In addition, we are interested in the interfaces used to access the underlying services, the pros/cons of using virtualisation, the range and scope of applications that can be executed, the security used by these services, and aspects such a service level agreements and quality of service provided.

Topics areas of interest:

• Security Policies and Mechanisms,
• Possession and ownership of data,
• Scheduling on Clouds,
• Cloud benchmarks
• Cloud-based services and protocols,
• Cloud Interoperability,
• Storage and file systems,
• Cloud scalability and performance,
• Fault-tolerance,
• Application development and debugging tools.
• Business models and economics of cloud services
• Vituralization of hardware and software resources,
• Performance monitoring and prediction of cloud performance,
• Capacity planning and resource allocation,
• Service Level Agreements and Quality of Services,
• Cloud architecture.
  

The format of submitted papers should follow the guidelines for IEEE conference proceedings format (8.5" x 11", Two-Column). Paper should be submitted in PDF format to the IEEE e-Science conference submission system.

The proceedings of this workshop will be published by IEEE Computer Society Press.

Submitting a paper to the workshop means that if the paper is accepted, at least one author should attend the workshop to present the paper.

Submission Deadlines:

Short contributions (up to four pages), which may state a position or describe a development or research outcome are invited, using the normal IEEE paper formatting guidelines.

First drafts: 25th September 2009

Notification of acceptance: 2nd October 2009

Final papers due: October 14th 2009.

All papers should be sent to Prof Mark A. Baker (mark.baker@computer.org)

Programme Committee:
Mark Baker, SSE, University of Reading

Marcel.Kunze, Forschungszentrum, University of Karlsruhe

Lee Gillam, Department of Computing, University of Surry

Nikolaos Antonopoulos, School of Computing University of Derby

Matthew Dovey, JISC

Garry Smith, SSE, University of Reading

Dan Katz, Computation Institute (CI), University of Chicago

Guy Tel-Zur, Ben-Gurion University of the Negev

David Abramson, Faculty of Information Technology, Monash University,
Jeremy Cohen, Imperial College
Terance Harmer, The Queen's University of Belfast

URL: http://www.oerc.ox.ac.uk/ieee/workshops/cloud

IGT GPGPU Meeting - Performance Primitives on the GPU

Presenter
Mordechai Butrashvily
CEO
Company for Advanced Supercomputing Solutions Ltd.

Abstract

This meeting will showcase NVIDIA library implementing various performance primitives on the GPU for various purposes, from image/video processing to other tasks.

The library can be considered as the GPU equivalent of Intel IPP, although with functionality that is added with time.

In the meeting we will overview the set of supported functions, test and discuss performance topics with regard to the GPU.

Using this library, users can experience between x10 to x50 performance increases when compared to running on the CPU.

This meeting will be followed by another, discussing upcoming LAPACK library implementation on the GPU.

To reserve your place, please send your contact details to: info@grid.org.il

Condor at the IGT

Miron Livny talking at the IGT event, July 15th, 2009.

The presentation of Dhruba Borthakur about Hadoop and Condor is available here.