Seminars

From CD++
Revision as of 11:19, 25 April 2014 by WikiSysop (Talk | contribs) ('''Continuous System Simulation with DEVS''')

Jump to: navigation, search

The Advanced Real-Time Simulation Laboratory , in conjunction with the Ottawa Student Chapter of the SCS (Society for Modeling and Simulation International), organizes varied Student Seminars (within the V-Sim Centre and the RADS Lab, SCE). The seminars will focus both on student research and technical discussion, and they are open to the community. Feel free to invite your coleagues and students to attend.

These Student Seminars are one of the main activities of the Ottawa Student Chapter of the SCS (Society for Modeling and Simulation International).

If you need further information or want to organize a seminar, contact the organizers.

Contents

Continuous System Simulation with DEVS

  • Speaker
Federico Bergero, French-Argentine International Center for Information and Systems Sciences (CIFASIS), Rosario, Argentina
  • When
2:00 pm., Tuesday, April 29th 2014
  • Where
3222 VS (Campus Map here).
  • Abstract
Computer simulation has become an important tool for studying, developing, and evaluating mathematical models of systems of different domains (physics, economics, biology, etc.). Such continuous systems are usually represented by sets of differential equations. Since most differential equations lack analytical solutions, they must be discretized to simulate the continuous systems they represent. To this end, most numerical integration methods approximate the differential equations by discretizing the time variable transforming them into sets of difference equations.

In this seminar we will introduce different approaches for continuous system simulation of differential equations. In particular the Quantized State Systems integration methods will be reviewed and compared to classic numerical solvers.


  • Short Biography
Federico Bergero received a Computer Science degree in 2008 and a PhD degree in Informatics in 2012 from the Universidad Nacional de Rosario, Argentina. He is currently a postdoctoral fellow at the French Argentine International Center for Information and Systems Sciences (CIFASIS) and holds a professorial position at the Universidad Nacional de Rosario in the Department of Computer Science, FCEIA—UNR, Rosario, Argentina. He is the winner of the Best DEVS Ph.D. Dissertation Award 2013. His research interests include discrete event systems, real time and parallel simulation, simulation tools, signal processing, and audio filtering methods.


== > DEVS-based Modeling of Coordinated Multipoint Techniques for LTE-Advanced ==

  • Speaker
Misagh Tavanpour, PhD candidate, SCE
  • When
12:00-1:00 pm, Wednesday October 2nd, 2013
  • Where
3222 VS (Campus Map here).
  • Abstract
Considering the ever-increasing bandwidth demand of the users in cellular networks, there have been ongoing investigations of new standards to support users' requirements and increasing their performance in mobile networks. One of the more promising mobile communication standards for the Fourth Generation (4G) cellular systems is the Long Term Evolution Advanced (LTE-Advanced) standard. This technology provides an order of magnitude higher data rates and improves the users' quality of service by using a number of technologies including Coordinated Multipoint (CoMP) processing. In our next meeting we will see how we have used the Discrete > Event System Specification (DEVS) formalism to model a mobile network using two approaches of CoMP: namely, coordinated scheduling/beamforming and joint processing.
  • Short Biography
Misagh Tavanpour is a Ph.D. student at Carleton University. He obtained a B.S. in Computer Engineering (2006) and a M.A. Sc (EE) at the Department of Computer Engineering, Science and Research University, Tehran, Iran (2009). His thesis focused on topologies and source mapping for QoS in NoC. He is working in ARS-Lab with DEVS since January 2012.

== ==

  • Speaker
  • When
  • Where
3222 VS (Campus Map here).
  • Abstract
  • Short Biography

Using Workflows Of Web Services To Manage Simulation Studies Into The Cloud

  • Speaker
Dr. Judicaël Ribault, Carleton University
  • When
6:00-7:00 pm, November 29th, 2011
  • Where
PA 236 (Campus Map here).
  • Abstract
We propose to drive simulation studies using workflows of web service. On one hand, workflows help us to control the complete simulation process by automating task and/or by guiding users through user interaction and on the other hand web services allows us to execute the simulation into the cloud. These two techniques promote the sharing and the replayability of experiments since no simulation environment must be locally-installed. In this paper, we demonstrate the feasability of such methodology using the Taverna workflow management system coupled with myExperiment wich is an online repository of workflow and the RISE middleware. RISE can make your simulator available from the cloud through a REST API.
  • Short Biography
Judicaël Ribault received a Ph.D. in Computer Science in 2011 in the field of modeling and simulation software (Université de Nice-Sophia Antipolis). He is currently a postdoctoral fellow from the DGA at Carleton University under the supervision of Gabriel Wainer, Department of Systems and Computer Engineering. His postdoctoral research concerns interoperability and the conduct of simulation study using workflows.

Seminar: Web-Languages for Distributed Simulation: from community-based Vocabularies to Web services and X-base Storage of Information Exchange

  • Speaker
Andreas Tolk, Old Dominion University.
  • When
3:00 pm, April 27th, 2011
  • Where
3222 VS (Campus Map here).
  • Abstract
One of the main reasons for interoperability and composability research is the wish to enable the reuse of existing solutions. Why should we invest something into rewriting a solution that already exists? However, most of the modeling and simulation solutions are not designed with reuse and interoperability in mind. The view is often focused on the particular application domain for which the problem has to be solved, not the general applicable of the designed solution in a broader context. Within the broader world of IT, Web services have been used to successfully build bridges between heterogeneous solutions. As such, they are potential candidates to generally support of interoperable solutions. However, model-based solutions require not only interoperability of the implemented solutions, the underlying conceptualizations need to be aligned as well to ensure composability of the models. In other words: the same concepts must be tagged with related terms and represented with structures that can be mediated into each other. The presentation will present the Levels of Conceptual Interoperability Model (LCIM) as a framework to address the various challenges when independently developed model-based solutions need to be executed in a distributed simulation to provide a common solution. It will present model-based data engineering to show how model-based solutions can be aligned in support of a common reference model based on the terms used in the supported community. Finally, it will present a web-service based solution that uses state-of-the-art web technology to implement these ideas. Although the underlying research was predominantly funded by military organizations, the presented principles and findings are not limited to military applications and have been successfully used in the domain of energy, smart grids, transportation, and medical applications.
  • Short Biography
Andreas Tolk is Associate Professor for Engineering Management and Systems Engineering at Old Dominion University in Norfolk, Virginia. He holds a Ph.D. and an M.S. in computer science, both from the University of the Federal Armed Forces of Germany in Munich. He is affiliated as a Senior Research Scientist with the National Centers for System of Systems Engineering (NCSoSE) in Norfolk, Virginia, and the Virginia Modeling Analysis and Simulation Center (VMASC) in Suffolk, Virginia. His research focuses on integratability and composability of model-based solutions and modeling and simulation based systems engineering. He is senior member of the Institute of Electrical and Electronics Engineers (IEEE) and Society for Modeling and Simulation International (SCS) and member of the American Society for Engineering Management (ASEM), Association for Computing Machinery (ACM), Military Operational Research Society (MORS), National Defense Industrial Association (NDIA), and Simulation Interoperability Standards Organization (SISO). He was recognized with the Excellence in Research Award of the Frank Batten College of Engineering and Technology and received the first Technical Merit Award of SISO.

PhD Dissertation Defense: Enhanced Distributed Simulation Interoperability and Algorithms Using Web Services

  • Speaker
Khaldoon Al-Zoubi, Systems and Computer Engineering, Carleton University
  • When
2:00 pm, April 28th, 2011
  • Where
4359 ME (Campus Map here).
  • Abstract
With the expansion of the Internet, the desire toward global cooperation in the distributed simulation technology has also been on the rise. As a result, much research has been devoted to develop middleware interoperability methods on the Web, particularly using purely SOAP-based WS or HLA with SOAP extension. However, such frameworks still have constraints in the structural rules that are placed on the middleware design methods. In this work, these issues are addressed, enhancing the distributed simulation methods on the Internet using SOAP WS and RESTful WS. In particular, the main objective of the methods presented is hiding systems heterogeneity (implementation) in components while allowing composition scalability and dynamicity. To do so, the proposed SOAP-based methods wrap each system in a single WS port with few RPCs while algorithms synchronize simulation via exchanging XML messages. However, the thesis shows how the objective can be better achieved by the RESTful Interoperability Simulation Environment (RISE) middleware, which is the first existing RESTful simulation environment. RISE is a general middleware that serves as a container to hold different components. In RISE, all functionalities are hidden in resources (URIs) that connected to each other via constant uniform virtual channels where all synchronization messages are described in XML. As a proof-of-concept, we RISE-enabled a simulator called CD++, so that it can perform distributed simulation on the Web. This system performance tests have shown promising results comparing to the SOAP-enabled version. Additional methods have also been defined such as algorithms that could be used as a basis for DEVS standardization, and workflow methods to enhance simulation experimentation automation and reusability via the Web.
  • Short Biography
Khaldoon Al-Zoubi is a PhD Candidate in Electrical and Computer Engineering within the Department of Systems and Computer Engineering in Carleton University, Ottawa, Canada. His main research interest is distributed simulation and Web services. He is also a senior software analyst and programmer with over 13 years of industry experience occupying a number of seniority and leadership positions. His industry experience spreads over wide range of areas such as embedded software and mobility, air-traffic software management and telecommunications, and security software for explosive and narcotics detections.


PhD Dissertation Defense: Algorithms for Parallel Simulation of Large-Scale DEVS and Cell-DEVS Models

  • Speaker
Qi Liu, Systems and Computer Engineering, Carleton University
  • When
9:30 am, Sept. 14, 2010
  • Where
4359 ME (Campus Map here).
  • Abstract
This dissertation addresses software development and performance issues that arise in demanding parallel DEVS simulations on both multiprocessor systems and chip-multiprocessor (CMP) architectures. A Lightweight Time Warp (LTW) protocol is proposed for efficient optimistic parallel DEVS simulation on multiprocessor clusters, allowing for purely optimistic simulation driven by only a few full-fledged Time Warp Logical Processes (LPs) while most of the LPs are released from the burden of TW execution. As general-purpose parallel discrete-event simulation on CMP architectures has received a growing interest from the M&S community, this research proposes a computing technique called Multicore Acceleration of DEVS Systems (MADS) for efficient parallel DEVS simulation on the heterogeneous IBM Cell processor, combining multi-grained parallelism and various optimizations to overcome the major performance bottlenecks, while hiding the technical details of multicore programming from non-expert users. Several key concepts and methods derived from this research can also be applied to other CMP and shared-memory architectures.
  • Short Biography
Qi Liu received his B.Eng. degree from Huazhong University of Science and Technology, China (1993). He received a M.A.Sc. degree from Carleton University, Ottawa, Canada (2006, Senate Medal for Outstanding Academic Achievement). He is currently a Ph.D. candidate in the Department of Systems and Computer Engineering at Carleton University. His research interests are centered around high-performance computing in simulation and advanced parallel/distributed simulation algorithms. (see Qi's website here).

Seminar: Modeling Hybrid systems with DEVS and Modelica

  • Speaker
Alfonso Urquia, Universidad Nacional de Educacion a Distancia (UNED), Spain
  • When
3:00 pm, Sept. 7, 2010
  • Where
3222 VS (Campus Map here).
  • Abstract
Modelica is a freely available, object-oriented modeling language that supports the physical modeling paradigm. Models are mathematically described by differential and algebraic equations (DAE), algorithms and discrete equations. Modelica supports a declarative (i.e., non-causal) description of the model. Therefore, the use of Modelica reduces considerably the modeling effort and permits better reuse of the models. A number of free and commercial Modelica libraries in different domains are available. The fundamentals of the Modelica language will be discussed. In addition, the requirements needed to describe Parallel DEVS models using Modelica will be identified and a free Modelica package, named DEVSLib, will be presented. DEVSLib facilitates the description of discrete-event models according to the P-DEVS formalism and provides components to interface with continuous-time models, which can be composed using other Modelica libraries.
  • Short Biography
Alfonso Urquia was born in 1969 in Madrid, Spain. He received his M.S. degree in Physics in 1992 from Universidad Complutense de Madrid and his Ph.D. in Physics in 2000 from Universidad Nacional de Educacion a Distancia (UNED). His work experience includes six years as a R&D engineer at AT&T, Lucent Technologies and Agere Systems. Since 2002, he has been working as an Associate Professor at Dpto. Informatica y Automatica, UNED, in Madrid. (see Alfonso's website here).


Seminar: Some Methodology Issues and Methodology Experiments in the OSA Project

  • Speaker
Olivier Dalle, INRIA, CNRS, University of Nice (UNS), Sophia Antipolis, France
  • When
3:00 pm, Aug. 9, 2010
  • Where
3222 VS (Campus Map here).
  • Abstract
In this presentation, Prof. Dalle will cover 3 topics about simulation methodology. The first is a short discussion about how to code discrete time in simulators. In many simulators, time is coded as a floating point value because it offers a great flexibility and scalability. He will show that, despite the higher difficulties in handling, integer coded values end up being more accurate. The second topic is about reproducibility and traceability of simulation results. After introducing the problem, he will briefly survey a few techniques and tools that might be used to help solving this critical issue. The last topic is about instrumentation of simulation models. He will describe the OSIF Framework, a framework developed in my team that allows for flexible and seamless instrumentation of simulation models.
  • Short Biography
Olivier Dalle is a Maître de Conferences (Associate Professor) in the Informatics Department of the Faculty of Sciences at University of Nice-Sophia Antipolis (UNS). He received his B.Sc. from University of Bordeaux and his MSc. and Ph.D. from UNS from 1999 to 2000. He was a post-doctoral fellow at the French Space Agency center in Toulouse (CNES-CST) doing research on the simulation of multimedia traffic over satellites. In 2000, he joined UNS and the MASCOTTE research group, a joint team of UNS, CNRS, and INRIA. Since then his research is mainly centered on simulation methodology and large scale networking and telecommunications systems (See Olivier's website here).

Seminar: PowerDEVS. An Integrated Tool for Hybrid System Modeling and Simulation

  • Speaker
Federico Bergero, French-Argentine International Center for Information and Systems Sciences (CIFASIS), Rosario, Argentina.
  • When
3:00 pm, Aug. 16, 2010
  • Where
3222 VS (Campus Map here).
  • Abstract
We will introduce a general purpose software tool for DEVS modeling and simulation oriented to the simulation of hybrid systems. The environment, called PowerDEVS, allows defining atomic DEVS models in C++ language that can be then graphically coupled in hierarchical block diagrams to create more complex systems. A remarkable feature of PowerDEVS is the possibility of performing simulations under a RTOS that permits the design and automatic implementation of synchronous and asynchronous digital controllers. Another feature is the interconnection between PowerDEVS and the numerical package Scilab.
  • Short Biography
Federico Bergero received a Computer Science degree in 2008 from the Universidad Nacional de Rosario, Argentina. He is currently a Ph.D. student at the French Argentine International Center for Information and Systems Sciences (CIFASIS), and he holds a teaching position at the Universidad Nacional de Rosario in the Department of Computer Science, FCEIA-UNR, Rosario, Argentina. His research interests include discrete-event systems, real-time simulation, signal processing, and audio filtering methods.

Seminar: Hybrid Control of Quality of Service in Networking Systems: A DEVS–based Methodology

  • Speaker
Rodrigo Castro, Universidad Nacional de Rosario, Argentina
  • When
3:00 pm, Aug. 23, 2010
  • Where
3222 VS (Campus Map here).
  • Abstract
In this seminar, we will present a formal Modeling and Simulation (M&S) methodology for hybrid control of networking systems. The method is designed to be used for analysis, design and implementation of Quality of Service (QoS) control systems in Network Processor (NP)-based applications. The main goal is enabling the application of continuous Control Systems Theory to enforce Admission Control strategies into Discrete–Event network traffic. This represents a hybrid system modeling problem that has to be treated formally to guarantee the applicability of the continuous control theoretical results into discrete–event systems.
By using DEVS (Discrete Event System Specification), in combination with Quantized State Systems (QSS) numerical methods for the approximation of continuous systems, numerous advantages can be exploited. These frameworks provide the means to accurately analyze and design hybrid models for Admission Control. They can be seamlessly integrated into a unified formal framework, and they enable a low-risk transition between the DEVS–based simulation and the deployment of the obtained hybrid models into the target networking platform.
We will review some recent advances in the implementation of the proposed methodology. The PowerDEVS tool is used for the offline analysis and design phases of the QoS controllers, and the ECD++ tool is used to run the obtained controllers into an Intel IXP2400 Network Processor based testbed, running a real time embedded Linux.
  • Short Biography
Rodrigo Castro is a Ph.D. student at the Universidad Nacional de Rosario, Argentina. He is also an adjunct Lecturer at the Computer Science Department, Faculty of Sciences (FCEN), Universidad de Buenos Aires (UBA), Argentina, and a lecturer for the post-graduate course "Information Integration Technologies in the Factory" at UBA. Since 2000, he has worked in several projects for Siemens, Cisco, Hewlett Packard, in the area of networking optimization, performance analysis, IT infrastructure design, and software development. In 2007 he was awarded a Fundación Repsol YPF - José A. Estenssoro PhD Fellowship for his PhD studies. Also in 2007, he was awarded an INGENIO(CSIC-UPV)/OEI grant (Spain) for studies on "Best Practices in the Cooperation University-Enterprise for Development" and in 2009 he received an "Emerging Leaders in the Americas Program" scholarship by the Government of Canada. He is an IEEE Member of Computer Society and Communications Society, and the Society for Modeling and Simulation International SCS.

RADS Presentation

  • Speaker
Prof. Wainer, Majumdar, Kunz, Chinneck, Yee, Petriu, Franks and Woodside (and maybe more), Systems and Computer Engineering, Carleton University
  • When
Wed., May 5th, 1:30 - 3:30pm
  • Where: 4359 ME
  • Abstract

Each faculty member and his/her students will have about 10 minutes to summarize their current goals and achievements. Presentations will be by the faculty member, the students, or a combination.

  • Organization

Refreshments will be provided starting at 1:00pm



ICS Grad Cog Talks - Behavioural Economics and Bounded Rationality

  • Speaker
Dr. Derek Ireland, Arthur Kroeger College of Public Affairs, Carleton University
  • When
Wed., March 3rd, at 12pm
  • Where: VSIM 5220
  • Abstract

Information overload, cognitive dissonance, hyperbolic discounting, salience, little nudges that "encourage us to do the right thing" whether we want to or not. Based on his recent PhD and other research, Dr. Derek Ireland will summarize some of these concepts from behavioural economics and their implications for the preferences we have and the decisions we make.

  • Short Biography

Dr. Ireland, who in 2009 received his PhD from the School of Public Policy, is an instructor at the Arthur Kroeger College, Carleton University.



MITACS Step Workshop Series @ Carleton University

  • Speaker
Various Speakers
  • When
February 5th, February 24th, February 25th and February 26th, 2010, 8:30AM - 4:30PM
  • Where
1501 CTTC Building (Carleton Technology and Training Centre)
  • Schedule
Feb. 05 - Practice Your Presentation Skills
Feb. 24 - Managing Projects
Feb. 25 & 26 - Foundations of Project Management I: A team-based approach
  • Cost
Free


Modeling and Simulation of Continuous Systems with DEVS

  • Speaker
Rodrigo Castro, UNR and UBA, ARgentina
  • When
Wednesday, November 25th. 6:05 pm-7:25 pm (invited talk for the SYSC 5104 course)
  • Where: ME 3190
  • Abstract

The DEVS formalism is a general system theoretic based formalism that can represent all systems whose input/output behavior can be described by sequences of events.

While DEVS can model and simulate any discrete system (discrete-time and discrete-event), continuous systems cannot be represented by the formalism due to the continuous evolution of its variables. Although differential equations can model these continuous systems, in order to simulate them we need to use classic numerical integration methods (i.e., Euler or Runge Kutta) to approximate the corresponding Ordinary Differential Equations (ODEs). However, classic numerical integration methods have problems with the discontinuities found in hybrid systems (caused by the discrete dynamics). These must be properly detected and treated; otherwise, the approximation of the continuous parts can lead to wrong results. Appropriate treatment and detection of discontinuities is in general difficult and computationally demanding, as it usually calls for expensive iterations.

We will show how Quantized State Systems (QSS) numerical methods can provide a discrete-event approximation of continuous systems ODEs by a DEVS model, preserving stability properties and guaranteeing error bounds. QSS/DEVS methods guarantee the avoidance of the state explosion problem that usually arises when a state trajectory is discretized by classic finite automata techniques, provided DEVS handles infinite state sets through the usage of transition functions, instead of transition tables. QSS methods are also able to detect and handle discontinuities in an efficient and straightforward way, which does not require any iteration. Thus, they are particularly convenient to simulate hybrid systems, and some times perhaps necessary when running under stringent real-time constraints.

DEVS in combination with QSS methods offer important advantages: they provide the means to accurately analyze and design general hybrid models for Control applications, and they can be seamlessly integrated into a unified formal framework. This enables the transition from a DEVS-based simulation to the final deployment of the obtained hybrid controllers into target hardware platforms.

In this talk, we will introduce the fundamental concepts of QSS methods and provide a practical application context: the analysis, design and implementation of Quality of Service (QoS) controllers for Network Processor (NP)-based systems. We will see how to apply continuous Control Systems Theory to enforce Admission Control strategies into discrete-event network traffic. This represents a hybrid system modeling problem that has to be treated formally, to guarantee the applicability of the continuous control theoretical results into discrete-event systems. By means of QSS methods, continuous systems are seamlessly integrated with discrete-event systems and provide a very suitable common framework to integrate heterogeneous disciplines in the area of Control, Real-Time Computing and Networking Systems.


  • Short Biography

Rodrigo Castro is a Ph.D. student at the Universidad Nacional de Rosario, Argentina. He is also an adjunct Lecturer at the Computer Science Department, Faculty of Sciences (FCEN), Universidad de Buenos Aires (UBA), Argentina, and a lecturer for the post-graduate course "Information Integration Technologies in the Factory" at UBA. Since 2000 he worked in several projects for Siemens, Cisco, Hewlett Packard, in the area of networking optimization, performance analysis, IT infrastructure design, and software development. In 2007 he was awarded a Fundación Repsol YPF - José A. Estenssoro Ph.D. Fellowship for his Ph.D. studies. Also in 2007, he was awarded an INGENIO(CSIC-UPV)/OEI grant (Spain) for studies on "Best Practices in the Cooperation University-Enterprise for Development" and in 2009 he received an "Emerging Leaders in the Americas Program" scholarship by the Government of Canada. He is an IEEE Member of Computer Society and Communications Society, and the Society for Modeling and Simulation International SCS.



VSIM Presentation Series

  • Speaker
Various Speakers, VSIM, Carleton University, Canada
  • When
12:00 - 13:00, Every Monday starting on 9 Nov. 2009
  • Where
VSIM Room #2221 (Hurricane Room)
  • Schedule
Nov. 09 - Overview of VSIM Activities - Herdman ACE Lab - Herdman
Nov. 16 Advanced Real-Time Sim Lab - Wainer; Science of Imagination Lab - Davies
Nov. 23 Math Cognition Lab - LeFevre; Applied Dynamics Lab - Langlois
Nov. 30 Reality in the Loop Lab - Pearce; Speech Lab - Logan
Dec. 07 CIMS Lab - Fai; TBD
Dec. 14 Cognitive Affective Neuroimaging Lab - D'Angiullli; TBD


Seminar: Efficient Parallel Simulation of Large-scale DEVS Models on the Cell Broadband Engine

  • Speaker
Qi Liu, ARS Laboratory, Carleton University, Canada
  • When
10:00 - 11:00 AM, Tuesday, 18 August 2009
  • Where
IBM T. J. Watson Research Center, Yorktown Heights, NY, USA [1].
  • Abstract
Modeling and simulation (M&S) has become a powerful tool for cost-effective and detailed analysis of natural and artificial systems where a mathematical approach is intractable. A general M&S methodology for describing discrete-event systems is the Discrete Event System Specification (DEVS), which supports hierarchical construction of reusable models in a modular way and reduces system development and testing effort. As the system under study becomes more and more complex, the simulation tends to be increasingly time-consuming and resource-demanding. In the quest for better performance, parallel simulation is widely accepted as one of the best choices to speed up discrete-event simulation systems. The purpose of this research is to fully leverage the potential of the Cell Broadband Engine (CBE) processor in large-scale parallel simulation of DEVS-based discrete-event systems. Towards this goal, new computing models, algorithms and protocols are proposed to address some of the key challenges raised in software development and performance engineering of parallel discrete-event simulation on such platform.
As the monolithic approach to microprocessor design has reached a point of diminishing return due to limitations such as memory latency, power consumption, and difficulty in achieving ever higher clock frequency, a clear trend has been observed in the industry moving towards multicore processor designs. Among them, the IBM CBE processor exhibits tremendous potential for scientific computing and has been used in the Roadrunner project to build a petascale hybrid supercomputer for the Los Alamos National Laboratory. However, parallel computing models for the CBE processor are still in their infancy, and standardized middleware is not even on the horizon. In this research, we are developing the first object-oriented simulation engine on top of the CBE processor using the general-purpose DEVS methodology. Doing so not only allows the broad community of DEVS users to harness the computing power of the CBE processor without being distracted by the technical details of multicore programming, but also provides valuable insights into the various design trade-offs that are of particular interest to application developers targeting the CBE platform.
Instead of taking the conventional coarse-grained parallelization approach, a novel fine-grained parallel computing model is proposed in this research to simultaneously exploit multi-dimensional parallelism at five distinct system levels, including event-level parallelism to achieve concurrent event execution without violating causal consistency, task-level parallelism across nine processing elements, memory-level parallelism to hide memory latency by virtue of multi-buffering techniques, data-level parallelism to optimize code execution using manual or compiler-assisted SIMD vectorization, and device-level parallelism to overlap computation with file I/O operation using a lock-free event caching technique. Built upon a solid theoretical foundation, the computing model adopts a performance-centered modular design, incorporating a variety of optimization strategies in the software architecture and providing the flexibility, configurability, and scalability required for future expansion. This paper presents the design methodology and rationale behind the computing model. Specifically, new forms of event-level parallelism are identified in the simulation process for efficient event execution and effective load balancing among the cores. Different computational kernels are extracted from the DEVS simulator and the issue of kernel allocation is addressed, taking into account the workload distribution and the unique characteristics of the CBE processor. A decentralized event-queue management scheme is introduced to orchestrate the parallel simulation with minimized resource contention. In addition, this paper also discusses how the proposed computing model for CBE platform can be seamlessly integrated with other existing conservative and/or optimistic parallel simulation approaches to combine the benefits of parallel simulation at the cluster level with the advantages of accelerated simulation on each individual multicore processor.
  • Short Biography
Qi Liu received his B.Eng. degree from Huazhong University of Science and Technology, China (1993). He received a M.A.Sc. degree from Carleton University, Ottawa, ON, Canada (2006, Senate Medal for Outstanding Academic Achievement). He is currently a Ph.D. candidate in the Department of Systems and Computer Engineering at Carleton University. His research interests are centered around high-performance computing in simulation and advanced parallel/distributed simulation algorithms.


Carleton University VSIM Speakers Series: Broadband Visual Communication in the Future

  • Speaker
John Spence and Martin Brooks, Communications Research Centre, Government of Canada
  • When
10:00 - 11:30 AM, Tuesday, 02 December 2008
  • Where
2221 VS (Hurricane Room) (Campus Map here).


Seminar: Communicating with FPGA Board using TCP Protocol

  • Speaker
Mohammad Moallemi, ARS Laboratory, Carleton University, Canada
  • When
3:00 pm, October 10, 2008
  • Where
3220 VS (Campus Map here).
  • Abstract
The AP1000 board donated to Carleton University by CMC (Canadian Microelectronics Corporation) is a programmable logic board, known as Field-Programmable Gate Array or FPGA, which consists of memory blocks, microprocessors (soft and hard macros), multiplier and digital signal processing blocks, embedded system IP such as bus architectures and peripheral components, and application-specific IP such as in DSP and telecom.
In this Project, a TCP protocol for communication between the AP1000 board and the IBM host PC was studied. The resulting design uses the high speed Ethernet Mac IP core and the light weight IP (lwip) library. The lwip library contains the implementation of TCP protocol in EDK design tool, making it suitable for developers to send and receive data using this protocol. With this approach, we successfully opened a TCP socket on the board side to establish a telnet connection between the board and the host PC. The board was configured to accept incoming connections from the host PC via the socket, and the status of data transmission was shown in a HyperTerminal window.
  • Short Biography
Mohammad Moallemi received his B.Eng. degree from Shahid Beheshti University, Tehran, Iran (2004). He received a M.A.Sc. degree from Ferdowsi University, Mashhad, Iran (2007). He is currently a Ph.D. candidate in the Department of Systems and Computer Engineering at Carleton University. His research interests include embedded simulation development using the CD++ toolkit.


Seminar: Vesicle-Synapsin Interactions Modeled with Cell-DEVS

  • Speaker
Rhys Goldstein, ARS Laboratory, Carleton University, Canada
  • When
5:30 pm, September 25, 2008
  • Where
4332 ME (Campus Map here).
  • Abstract
Interactions between synaptic vesicles and synapsin in a presynaptic nerve terminal were modeled using the Cell-DEVS formalism. Vesicles and synapsins move randomly within the presynaptic compartment. Synapsins can bind to more than one vesicle simultaneously, causing clusters to form. Phosphorylation of synapsin reduces its affinity for vesicles, and causes the clusters to break apart. Upon dephosphosphorylation, new clusters form. Taking advantage of Cell-DEVS, as opposed to traditional techniques for implementing cellular automata, the model prevents collisions between arbitrarily large clusters using transition rules restricted to a 5-cell neighborhood. Simulation results indicate that, in a qualitative sense, the behavior of vesicles and synapsin in neurons was captured.
  • Short Biography
RHYS GOLDSTEIN received a B.A.Sc. degree (2003) from the Engineering Physics Department at the University of British Columbia (Vancouver, BC, Canada). He then worked in the mining industry, developing data analysis software and leading geophysical surveys in various parts of the world. He is now pursuing a M.A.Sc. in Biomedical Engineering from the Department of Systems and Computer Engineering at Carleton University (Ottawa, ON, Canada). His email is <rhys@sce.carleton.ca>.


Thesis Defence (M.ISS.): Implement Parallel and Distributed DEVS and Cell-DEVS Simulation in Windows Platform

  • Speaker
Bo Feng, ARS Laboratory, Carleton University, Canada
  • When
10:00 am, September 19, 2008
  • Where
4359 ME (Campus Map here).
  • Abstract
DEVS is a modeling and simulation framework based on generic dynamic system theory. Cell-DEVS is a DEVS-based formalism to model complex physical systems as cell spaces. Message Passing Interface(MPI) is a portable, flexible, vendor independent and platform independent standard for messaging on high performance computing. In this work, a parallel simulation tool (called PCD++) is presented. PCD++Win is based on CD++, which is a modeling and simulation toolkit capable of executing DEVS and Cell-DEVS models. Specifically, PCD++Win is following the conservative approach for synchronization, taking the advantage of MPI and moving PCD++ engine from Linux to Windows environments. PCD++Win allows users to setup, configure and execute parallel DEVS and Cell-DEVS simulation with a Windows-based graphical user interface(GUI). Moreover, PCD++Win is exposed as a Web service, which is independent of programming language, platform and device. Finally, PCD++/.NET system is created by combining .NET Remoting technology with parallel CD++ and allowing user to execute distributed simulation with commodity PC machines across a network. PCD++/.NET supports various protocols and formatter. That allows PCD++/.NET Remoting has more flexibility and extensibility than Web service-based simulation tool, which can only use HTTP protocol and SOAP formatter. The performance of PCD++/.NET heavily depends on partition strategy and model behavior. PCD++/.NET can be suitable for the model, which has modest remote message, and presents tolerable performance.
  • Short Biography
Bo Feng is currently a Master student in the Department of Systems and Computer Engineering, Carleton University. His research interests focus on parallel and distributed DEVS and Cell-DEVS simulation. He has joined in Participatory Design Studio project, which was sponsored by CANARIE. He is expected to graduate in Sept 2008.


Seminar: Implementation of a Car Travel Model on IBM Cell/B.E. Processor

  • Speaker
Mahdi Yusuf and Yuri Boiko, ARS Laboratory, Carleton University, Canada
  • When
3:00 pm, August 19, 2008
  • Where
3220 VS (Campus Map here).
  • Abstract
In this seminar, Mahdi will summarize his project on model development in Cell/B.E. environment. The performance of parallel simulation of a car travel model obtained on the Cell/B.E. architecture is compared to that achieved on single CPU execution. The talk will also address the vectorization concepts in Cell/B.E. applications.
  • Short Biography
Mahdi Yusuf is an undergrad student in the Department of Systems and Computer Engineering, Carleton University, Canada (expected graduation in 2010). During the Summer semester of 2008, he has been working with a group of students in the ARS laboratory on projects of Cell/B.E. based simulations. Yuri Boiko is a second year M.Sc. Student in the ARS laboratory working on related projects.


Seminar: Lightweight Time Warp - A Novel Protocol for Parallel Optimistic Simulation of Large-Scale DEVS and Cell-DEVS Models

  • Speaker
Qi Liu, ARS Laboratory, Carleton University, Canada
  • When
4:00 pm, August 15, 2008
  • Where
3220 VS (Campus Map here).
  • Abstract
In this seminar, we propose a novel Lightweight Time Warp (LTW) protocol for high-performance parallel optimistic simulation of large-scale DEVS and Cell-DEVS models. By exploiting the characteristics of the simulation process, the protocol is able to set free most logical processes (LPs) from the Time Warp mechanism, while the overall simulation still executes optimistically, driven by only a few full-fledged Time Warp LPs.
The LTW protocol includes a rule-based event-scheduling mechanism using two types of event queues, an aggregated state-saving technique for optimal risk-free state management, and a new rollback algorithm that recovers lightweight LPs from causality errors without sending anti-messages. The impact on global control mechanisms such as GVT computation, fossil collection, and load migration is also discussed. The basic concepts of the protocol could also apply to a broad range of Time Warp systems under certain conditions and with appropriate control over the LPs.
  • Short Biography
Qi Liu received his B.Eng. degree from Huazhong University of Science and Technology, China (1993). He received a M.A.Sc. degree from Carleton University, Ottawa, ON, Canada (2006, Senate Medal for Outstanding Academic Achievement). He is currently a Ph.D. candidate in the Department of Systems and Computer Engineering at Carleton University. His research interests are centered around high-performance computing in simulation and advanced parallel/distributed simulation algorithms.


Seminar: An Overview of the Existing Parallel CD++ Simulators

  • Speaker
Shafagh Jafer, ARS Laboratory, Carleton University, Canada
  • When
4:00 pm, June 24, 2008
  • Where
3220 VS (Campus Map here).
  • Abstract
In this seminar, we will present new techniques for DEVS and Cell-DEVS models in parallel and distributed environments. The parallel simulators are based on the Time Warp optimistic synchronization protocol, and are developed as new simulation engines for CD++, an M&S toolkit that implements both DEVS and Cell-DEVS formalisms.
Two distinct parallel simulators, namely Purely Optimistic PCD++ and Conservative PCD++, are introduced which use hierarchical and flattened architecture respectively. Different Cell-DEVS models are built in CD++ in order to judge the performance of these two simulators. Moreover, two new algorithms, Local Rollback Frequency Model (LRFM) and Global Rollback Frequency Model (GRFM) are implemented to control optimism of the optimistic PCD++. The LRFM and GRFM techniques are modifications to the WARPED kernel which are applied to the optimistic PCD++. A set of detailed tests are collected to investigate the effect of these approaches on the simulator.
  • Short Biography
Shafagh Jafer graduated with B.Eng. degree from Department of Systems and Computer Engineering of Carleton University in 2005. She received her M.A.Sc. degree in Systems and Computer Engineering from Carleton University in 2007. She is currently a Ph.D. candidate at the department of Systems and Computer Engineering at Carleton University. Her research area involves DEVS and CELL-DEVS simulations as well as Parallel CD++ Simulators.


Seminar: Modeling and Simulation of Generalized Stochastic Discrete Event Systems

  • Speaker
Rodrigo Castro, Universidad Nacional de Rosario, Argentina
  • When
4:00 pm, May 13, 2008
  • Where
3222 VS (Campus Map here).
  • Abstract
The DEVS formalism is a general system theoretic based formalism that can represent all the systems whose input/output behavior can be described by sequences of events. Nevertheless, DEVS has only been defined formally for deterministic systems, which limits the stochastic treatment of the systems under study.
In this seminar, we will introduce an extension of the classic DEVS formalism that includes stochastic features. Based on the use of Probability Spaces, the STochastic DEVS specification (STDEVS) provides a formal framework for modeling and simulation of generalized non deterministic discrete event systems. We will go through the main theoretical properties of STDEVS and illustrate its use in a stochastic oriented simulation example with the main purpose of performance analysis in computer systems and data networks.
  • Short Biography
Rodrigo Castro is a Ph.D. candidate at the Universidad Nacional de Rosario, Argentina. He is also an adjunct Lecturer at the Computer Science Department, Faculty of Sciences (FCEN), Universidad de Buenos Aires, an invited lecturer for the post-graduate course "Information Integration Technologies in the Factory" at Universidad de Buenos Aires, Argentina. Since 2000 he worked in several projects for Siemens, Cisco, Hewlett Packard, in the area of networking optimization, performance analysis, information systems infrastructure design, and software development. In 2007 he was awarded a Fundación Repsol YPF - José A. Estenssoro Ph.D. Fellowship (Argentina) for his Ph.D. studies. Also in 2007, he was awarded an INGENIO(CSIC-UPV)/OEI Grant (Spain) for studies on "Best Practices in the Cooperation University-Enterprise for Development". He is an IEEE Member of Computer Society and Communications Society.


OLDER SEMINARS

You can find information about older seminars organized by our lab here).