Tutorials

HPCS 2010 TUTORIALS


T1.
Simulation for Large-Scale Distributed Computing Research
3 Hours

Martin Quinson
University of Nancy, France

T2.
Wireless and Ubiquitous Broadband Real-time Multimedia Social Networks
3 Hours

Andrea Tarasconi
President & CEO, Essentia S.p.A., Italy

T3.
Observing and Declaring Program Speedups with Rigorous Statistics
3 Hours

Sid Touati
INRIA laboratory - University of Versailles Saint-Quentin en Yvelines, France

T4.
Security in Vehicular Ad hoc Networks: Security and Mobility
2 Hours

Abderrahim Benslimane
University of Avignon - France

T5.
Introduction to the Partitioned Global Address Space (PGAS) Programming Model
2-3 Hours

David E. Hudak
Ohio Supercomputer Center, Columbus, OH, USA





T1.
Simulation for Large-Scale Distributed Computing Research
3 Hours

Martin Quinson
University of Nancy, France


Martin Quinson photo

TUTORIAL DESCRIPTION
This tutorial will provide attendees with clear perspectives on the challenges for experimental research in the area of
parallel and large-scale distributed computing, and on current technology for conducting experiments with real-world testbeds, emulated testbeds, or simulated testbeds.  The first part of the tutorial will present and contrast current experimental methodologies, giving attendees in-depth understanding of the scientific and technological issues at hand.  The second part of the tutorial will focus on simulation, giving a state of the art of current simulation technology and discussing challenges for the development of sound simulation models.  The tutorial will use the SimGrid simulation framework as an example since it implements sophisticated and validated simulation models. The last part of the tutorial will focus on an in-depth presentation of the different simulation approaches enabled by SimGrid, each with its specific range of applications and goals.  SimGrid has been used to obtain results published in over 50 research articles and has thus emerged as one of the key tools for simulation in the area of parallel and large-scale distributed computing.  Tutorial attendees will have the opportunity to gain some hands-on experience with SimGrid, by witnessing step-by-step development of small simulation projects.  By the end of this tutorial attendees should have a clear understanding of current technology and best practice for experimental parallel large-scale distributed computing research, and in particular on the use of simulation. 



T2.
Wireless and Ubiquitous Broadband Real-time Multimedia Social Networks
3 Hours

Andrea Tarasconi
President & CEO, Essentia S.p.A., Italy


Andrea Tarasconi photo

TUTORIAL DESCRIPTION
Recent developments in the IEEE 802.11 based wireless technology open an innovative scenario where users can have ubiquitous access to network broadband services and to multimedia resources of their interest. However, providing suitable solutions to ever growing users demands and expectations is a quite challenging endeavor that requires both the availability of innovative networking solutions, and deep understanding of market needs and trends. Based on long-standing industrial experience in the field, this Tutorial will cover various state-of-the-art wireless technologies and will discuss market developments and users' expectation changes over the last decade. In addition, the presentation will draw a research agenda for the years to come, by keeping into account the emerging requirements for energy efficient wireless solutions, the popularity increase of network-intensive mobile applications, along with the new opportunities opened by wireless mesh technology allowing for self-healing and self-configuring easy network deployments.



T3.
Observing and Declaring Program Speedups with Rigorous Statistics
3 Hours

Sid Touati
INRIA laboratory - University of Versailles Saint-Quentin en Yvelines, France



Sid Touati photo

TUTORIAL DESCRIPTION
Code optimisation methods are usually experimented with by repeatedly measuring the execution times of the initial and optimised code versions. Given a fixed input for the program, speedup is defined as the ratio of the initial and optimised execution times. However, even with a fixed execution environment, running times vary, especially for toy/kernel benchmarks. With the introduction of multi-core architectures, this variability is becoming increasingly unstable. So different kinds of speedups may be reported, based on average execution time, minimal execution time, median, etc. Many published speedups in the literature are observations of a set of experiments that do not guarantee reproducibility. To improve upon that, this tutorial presents a rigorous statistical methodology regarding program performance analysis. We rely on well-known statistical tests (Shapiro-Wilk's test, Fisher's F-test, Student's t-test, Kolmogorov-Smirnov's test, Wilcoxon-Mann-Whitney's test) to study whether observed speedups are statistically significant. By fixing a desired risk level, we are able to analyse the statistical significance for both average and median execution times.



T4.
Security in Vehicular Ad hoc Networks: Security and Mobility
2 Hours

Abderrahim Benslimane
University of Avignon - France



Abderrahim Benslimane photo

TUTORIAL DESCRIPTION
The tutorial focuses on security issues specific to the vehicular ad hoc environment.  Two main categories are exposed in the talk: security in Ad hoc networks and security in vehicular networks.  The presentation highlights new security requirements raised by the VANET paradigm and outline the security schemes proposed in the literature that cope with the particular characteristics of a self-organizing network.  The tutorial ends with an analysis of the current trend in security solutions for the VANET environment and draws recommendations for an appropriate approach in solving open problems that are not covered by the literature.


T5.
Introduction to the Partitioned Global Address Space (PGAS) Programming Model
2-3 Hours

David E. Hudak
Ohio Supercomputer Center, Columbus, OH, USA



David E. Hudak photo

TUTORIAL DESCRIPTION
The Partitioned Global Address Space (PGAS) programming model serves as the foundation for a set of software libraries and programming languages intended to improve the productivity of scalable parallel programming. PGAS is a refinement of shared memory programming in which a logically shared global address space is divided into parts that can be mapped to individual physical memories. This memory model supports the creation of distributed data structures and relies on one-sided communication mechanisms.  PGAS is available to existing C and FORTRAN programs via a library (the GA Toolkit) and both commercial and open-source implementations exist forMATLAB.  PGAS languages have been developed, notably UPC and Co-Array FORTRAN.  In addition, the PGAS model is central to the DARPA HPCS languages, including X10.

In order to address petascale science, HPC system architectures are simultaneously increasing in complexity along several dimensions:  increasing CPU core count in the processors; GPU accelerators on the motherboard and multi-level interconnects between nodes.  In order to effectively exploit these systems, applications will have to support multi-level parallelism and aggressively overlap communication and computation.   Developers currently address these issues by combining programming models, e.g., OpenMP within a node and MPI across nodes. PGAS tools and languages seek to provide an abstraction that enables developers to reason about algorithm correctness while also supporting performance tuning at the application source code level.  Solutions using the PGAS programming model will be an increasingly attractive alternative to OpenMP for intra-node solutions and may be extended to provide a unified model for both intra-node and inter-node solutions. 

This tutorial provides an overview of the PGAS programming model and compares applications in pMatlab and UPC.  The X10 programming language is introduced as an example of a new generation of PGAS programming languages.  Detailed examples in X10 will demonstrate the language’s new features. 


 

# of hits: 51518 Copyright © 2007-2017 CIS Build:2017.09.24:15.27.50 EDT