Tutoriais

Tutorial:
"Adaptive and Reflective Middleware: A Perspective from Lancaster"
Geoff Coulson - Lancaster University, UK.
Abstract e Biography ( )
Abstract
In this tutorial, I survey the development of the field of adaptive and reflective middleware (ARM) and sum up the state of the art in this field. The perspective will rely substantially on the point of view of the middleware research group at Lancaster University, but will also encompass work from elsewhere in the ARM research community. The tutorial will take a retrospective stance, presenting experiences and lessons learned from ARM research, and highlighting areas of future research. In particular, I will start by discussing fundamental principles and models of structure, reflection and adaptation. I will then consider the extension of basic ARM prototypes along the two orthogonal dimensions of 'depth' and 'breadth'. The 'depth' extension applies ARM principles to systems that lie beneath the traditional middleware domain: e.g. operating systems and networks. The 'breadth' extension then applies the principles in a broader range of application areas than those traditionally considered. These include Real-time Control Systems, Grid Computing environments, and Wireless Sensor Network applications. I also briefly consider future work in applying ARM to the development of self-managing distributed systems.
Biography
Geoff Coulson is a Professor of Distributed Computing at Lancaster University, UK. His research interests include distributed systems, adaptive sensor networks, embedded systems, middleware technologies, grids, and component-based software development. Since receiving his PhD in 1992, Geoff has led many successful projects in the distributed systems/ middleware area. He also serves on numerous PCs in his research areas, has been programme co-chair for the ACM/IFIP Middleware Conference series, and has published over 40 journal and 100 conference papers. Further information is available at: http://www.comp.lancs.ac.uk/computing/users/geoff/ . 
Tutorial:
"Mobile Peer-to-Peer Networks for Vehicular Applications"
Mario Gerla – UCLA – California - USA.
Abstract e Biography ( )
Abstract
Peer-to-peer (P2P) systems have met enormous success in the Internet because of the many advantages they offer with respect to traditional client/server systems, namely: scalability, dependability and robustness. Peer to Peer systems are now emerging also in wireless, mobile networks. For example: vehicle and people based content distribution, file sharing and network games; urban environment sensing with vehicles or personal cellular phones; collaborative group-work in people networks, etc. Mobility and radio propagation pose formidable challenges to mobile P2P architectures, requiring a fundamental redesign with respect to traditional P2P strategies.
In this tutorial we address the vehicular scenario, a scenario that has attracted growing interest for the range of applications it enables, from safe driving to content distribution, advertising, commerce and games. We first introduce basic P2P models for fixed and mobile networks. Then, we review popular Vehicle to Vehicle (V2V) applications supported by the P2P dissemination paradigm (Car/CodeTorrent, Mobeyes and Last Encounter Routing). Finally, we discuss the design challenges including information harvesting and privacy
Outline:
Part I : Basic P2P models
- Internet Overlays: Multicast
- DHT Indexing: Chord, Pastry
- A Mobile DHT: VRR
Part II: Vehicular Applications
- Mobility models
- Content Distribution: CarTorrent; CodeTorrent; Network Coding
- Dissemination: Urban Sensing, Mobeyes
- Last Encounter routing – a routing scheme supported by P2P dissemination
- Secure dissemination: Situation Aware Trust (SAT)
Biography
Dr. Mario Gerla, Professor, UCLA, Computer Science Dept. Dr. Gerla received his Engineering degree from the Politecnico di Milano, Italy, in 1966 and the M.S. and Ph.D. degrees from UCLA in 1970 and 1973. He became IEEE Fellow in 2002. At UCLA, he was part of a small team that developed the early ARPANET protocols under the guidance of Prof. Leonard Kleinrock. He worked at Network Analysis Corporation, New York, from 1973 to 1976, transferring the ARPANET technology to several Government and Commercial Networks. He joined the Faculty of the Computer Science Department at UCLA in 1976, where he is now Professor. At UCLA he has designed and implemented some of the most popular and cited network protocols for ad hoc wireless networks including distributed clustering, multicast (ODMRP and CODECast) and transport (TCP Westwood) under DARPA and NSF grants. He has lead the $12M, 6 year ONR MINUTEMAN project, designing the next generation scalable airborne Internet for tactical and homeland defense scenarios. He is now leading two advanced wireless network projects under ARMY and IBM funding. In the commercial network scenario, with NSF and Industry sponsorship, he has led the development of vehicular communications for safe navigation, urban sensing and location awareness. A parallel research activity covers personal P2P communications including cooperative, networked medical monitoring (see www.cs.ucla.edu/NRL for recent publications).

Palestras
Palestra:
“The ‘OpenCom’ Approach to Developing Systems Software in Resource-Constrained Environments"
Geoff Coulson - Lancaster University, UK.
Abstract e Biography ( )
Abstract In this talk I will give an overview of the Lancaster approach to the development of configurable and reconfigurable middleware. This approach is based on the foundation of a systems-oriented component model called OpenCom, together with associated notions of component frameworks, reflection services and 'platform extensions'. The approach is programming language and deployment environment independent and has been implemented in a number of languages and applied in a range of OS and bare hardware environments. After introducing our basic approach I will discuss a particular component framework - Gridkit - which we currently use in a range of projects. In particular, I will talk about how we are using Gridkit to build a flood monitoring system that integrates a sensor network with a backend computational grid. I will focus on how we use Gridkit i) to perform 'local grid computation' within the sensor network itself; and ii) to dynamically optimise the sensor network (at both the physical network level and the overlay level) to different operating conditions (e.g. dry summer or extreme flooding).
Biography
Geoff Coulson is a Professor of Distributed Computing at Lancaster University, UK. His research interests include distributed systems, adaptive sensor networks, embedded systems, middleware technologies, grids, and component-based software development. Since receiving his PhD in 1992, Geoff has led many successful projects in the distributed systems/ middleware area. He also serves on numerous PCs in his research areas, has been programme co-chair for the ACM/IFIP Middleware Conference series, and has published over 40 journal and 100 conference papers. Further information is available at: http://www.comp.lancs.ac.uk/computing/users/geoff/ .

Palestra:
"Vehicular Urban Sensing: Dissemination and Retrieval"
Mario Gerla – UCLA – California - USA.
Abstract e Biography ( )
Abstract There has been growing interest in vehicle to vehicle communications for a broad range of applications ranging from safe driving to content distribution, advertising, commerce and games. One emerging application is urban sensing. Vehicles monitor the environment, classify the events, e.g., license plates, pollution readings, etc. and exchange metadata with neighbors in a peer-to-peer fashion, creating a distributed index from which mobile users can extract different views. For instance, the Department of Transportation captures traffic statistics; the Department of Health monitors pollutants, and; Law Enforcement Agents investigate crimes. Mobile, vehicular sensing differs significantly from conventional wireless sensing. Vehicles have no strict limits on battery life, processing power and storage capabilities. Moreover they can generate enormous volumes of data, making conventional sensor data collection inadequate. In this talk we first review popular V2V applications and then introduce MobEyes, a middleware solution that diffuses data summaries to create a distributed index of the sensed data. We discuss the challenges of designing and maintain such a system, from information dissemination to harvesting, routing and privacy.
Biography
Dr. Mario Gerla, Professor, UCLA, Computer Science Dept. Dr. Gerla received his Engineering degree from the Politecnico di Milano, Italy, in 1966 and the M.S. and Ph.D. degrees from UCLA in 1970 and 1973. He became IEEE Fellow in 2002. At UCLA, he was part of a small team that developed the early ARPANET protocols under the guidance of Prof. Leonard Kleinrock. He worked at Network Analysis Corporation, New York, from 1973 to 1976, transferring the ARPANET technology to several Government and Commercial Networks. He joined the Faculty of the Computer Science Department at UCLA in 1976, where he is now Professor. At UCLA he has designed and implemented some of the most popular and cited network protocols for ad hoc wireless networks including distributed clustering, multicast (ODMRP and CODECast) and transport (TCP Westwood) under DARPA and NSF grants. He has lead the $12M, 6 year ONR MINUTEMAN project, designing the next generation scalable airborne Internet for tactical and homeland defense scenarios. He is now leading two advanced wireless network projects under ARMY and IBM funding. In the commercial network scenario, with NSF and Industry sponsorship, he has led the development of vehicular communications for safe navigation, urban sensing and location awareness. A parallel research activity covers personal P2P communications including cooperative, networked medical monitoring (see www.cs.ucla.edu/NRL for recent publications).

Palestra:
“ONELAB. Can we build a testbed to explore the future Internet?"
Serge Fdida – University Pierre et Marie Curie – Paris – France.
Abstract e Biography ( )
Abstract
Current networking technology is built around the Internet Protocol (IP), a communications standard that is thirty years old. While IP has been very successful, due in large part to its minimalist design, its dominance has made it difficult to experiment with new approaches. In addition, the current Internet is a complex system that evolves rapidly over time. As it will include more diversity and functionality in the future, there is a risk that we miss an efficient and robust design. The main challenge for network research is to develop and anticipate the methodologies and tools to assess the design and assumptions for these systems. We argue that it is of utmost importance to develop a Network Science as well as prepare a framework to conduct experiments for proof-of-concept.
Network Science is by essence multidisciplinary, including but not limited to distributed systems, information theory, queueing systems, signal theory, graph, social sciences to name a few. Moreover, it is when you put all components in interactions that the system will reveal its real nature and structure. These correlations, invariants and laws are hard to isolate and model properly. Theory is necessary for the analysis of components in isolation. Eventually, it can be extended to capture correlation effects as in complex systems. However, the very specific nature of some components or the objective to analyze the system at scale or with heterogeneous components will justify the utilization of a proof-of-concept testbed. The design of the test-bed itself, its deployment and operation is a challenge because of the resources required in developing it, the management cost, the technical challenges to build its components or the sustainability issue. Research problems are related to the tools that have to be embedded in the test-bed, such as virtualization to isolate experiments, monitoring to capture data, benchmarking to control the experiments and associated results, federation to ring some other facilities in context … Other challenges are related to the various technologies that have to be made available to provide diversity and scale. This is the case with wireless components where specific concerns arose such as virtualization of the communication channel. In addition, the test-bed should be made easily accessible for experimentation and therefore, the utility function of the facility has to be high enough whilst the entry cost should be remained low.
The European IST OneLab2 project is addressing some of these ambitious goals in tight association with a similar international effort. It operates PlanetLab Europe, extending PlanetLab service across Europe, and federating with other PlanetLab infrastructures worldwide. It will integrate new features and technologies into the system. In particular, Onelab2 will enhance the testbed-native network monitoring service that supports experiments. Finally, it will cooperate with its potential customers by directly involving pilot projects testing novel ideas under synthetic or real-world situations. OneLab2 will build PLE gateways to unusual, cutting-edge networking environments. And OneLab2 will push forward a federation model such that PLE can serve as a basis for a future highly heterogeneous communications environment.
This talk is presenting the major assumptions, progress and issues for building such a facility.
Biography
Serge FDIDA is a full professor at the University Pierre et Marie Curie (Paris) since 1991. From 1989 to 1995, he was a Full Professor to the University Rene Descartes (Paris). Professor Fdida was a Visiting Scientist at IBM Research during the 1990/91 academic year. His research interests are in the area of content networking, pervasive communication, resource management and performance analysis. He was the chairman (or co-chair) of the following events: IFIP Modelling Techniques and Tools'87, IFIP High Performance Networking'94 (HPN'94), Performance of Data Communication'95 (PCN'95) and European Conference on Multimedia Applications, Services and Techniques'97 (ECMAST'97), Networked Group Communication (NGC'99), IFIP Networking'2000 and Conext’2005. He is or served on the editorial boards of many journals : Computer Communication, SIGCOMM CCR, Computer Networks Journals, Journal of Wireless and Optical Communications, and Annales des Télécommunications. He was on the steering committee of the European FP6 network of Excellence ENEXT. He is the Chair of the Conext Steering Committee.
Serge Fdida has also developed a long experience working with various organizations, including government or research agencies. From 2000 to 2005, Serge Fdida was working part-time as a scientific adviser with CNRS-STIC (French National Scientific Research Center / Information Science & Technology). He was the Vice-President of the french national research network on telecommunications involving academia and industry (RNRT) from 2005 to 2007. Serge Fdida is the coordinator of the IST Onelab project (An Open Networking Laboratory Supporting Communication Network Research Across Heterogeneous Environments).


Palestra do Patrocinador:
“Greening Google: Challenges and Opportunities?"
Hugo Santana – Google.
Abstract e Biography ( )
Abstract: Google continues to grow, and the infrastructure that underpins that growth depends on electricity. We will discuss a number of challenges and opportunities in making our infrastructure greener. How can we make our servers and data centers more energy efficient? How can we reduce the environmental impact of the energy that we do consume?
Biography
Hugo Santana is a Software Engineer at Google Inc. He received his B.Sc. and M.Sc. degree from Universidade Federal de Pernambuco, with emphasis on Machine Learning and Distributed Artificial Intelligence. He joined Google in early 2007, having worked in the area of Search Quality since then.
|