Tutoriais e Palestras


Tutorial 1
Change and Configuration Management for Distributed Applications
Alexander Keller
IBM Research Division, Thomas J. Watson Research Center, USA
Quinta-feira 01/06/06, 08:00-12:00

Abstract: Distributed applications are a critical part of the foundation of a modern enterprise, supporting an ever increasing number of critical business functions. Effective management of distributed applications is a major objective of every IT organization. This tutorial presents state-of-the-art approaches and solutions covering the entire lifespan of web applications: planning, deployment, operations, and change management. In particular, we present and position recent standards for service and application deployment, installation, configuration and change management. Specific emphasis will be put on applying the IT Infrastructure Library (ITIL), a widely accepted process-based approach to IT service management, and OASIS Solution Deployment Descriptors, an emerging standard for the installation and configuration of complex, distributed application systems.

About the researcher: Alexander Keller is a Research Staff Member and manages the Service Delivery Technologies department at the IBM T.J. Watson Research Center. He received his M.Sc. and Ph.D. degrees in computer science from Technische Universität München, Germany, in 1994 and 1998, respectively, and has published more than 50 refereed papers in the area of distributed systems management. Alexander?s research interests revolve around change management for applications and services, information modeling for e-business systems, and SLAs. He serves on several technical program and organizing committees of related conferences and workshops and is a member of the USENIX Association, the IEEE, and the DMTF CIM Applications and Metric Extensions Working Groups. He was a main contributor to the IBM Web Service Level Agreement (WSLA) framework, which served as the basis for the upcoming GGF WS-Agreement standard.

Tutorial 2
Scheduling algorithms for input-queued IP routers
Marco Ajmone Marsan
Electronics Department, Istituto Politecnico di Torino, Italy
Quinta-feira 01/06/2006, 14:00-18:00

Abstract: This tutorial discusses the issues involved in the design of schedulers for inputqueued switching matrices, within high-performance IP routers. The goal of the tutorial presentation is to first illustrate the motivations for input-queued switching matrices controlled by simple and ef£cient scheduling algorithms within high performance routers, and then overview some of the most significant scheduling algorithms that were recently proposed in the technical literature. The performances of all overviewed schemes are discussed through numerical simulation results as well as theoretical statements about their stability properties.

Tutorial outline
  1. IP routers. The role of routers in the Internet and in corporate Intranets/Extranets. Backbone, enterprise and access routers.
  2. Router functions and architectures. Basic functions: routing and switching. Advanced functions: IP datagram classification and filtering. Key router elements: line cards, packet queues, packet processors, switching fabrics. Key design issues for different classes of routers. Non-blocking, bufferless packet switches. Buffering schemes: OQ, IQ and CIOQ, VOQ. Internal speedup.
  3. Scheduling in IQ switches. Definition as a matching problem.
  4. Optimal scheduling algorithms. Maximum weight matching algorithm. Performance and implementation.
  5. Unweighted scheduling algorithms. iSLIP, WFA, 2DRR, RC. Performance and implementation.
  6. Weighted scheduling algorithms. iLQF, iOCF, RPA. Performance and implementation.
  7. Learning scheduling algorithms. Use of randomization and parallel search. Observing arrivals. Performance and implementation.
  8. Variable-size packet support. Packet-scheduling. Performance and implementation.
  9. Multicast support. Throughput limitations. Per-flow queueing.
  10. QoS support. Performance limitation of tracking fluid policies. Proposed algorithms.
  11. Other structures. Two stage switches. Parallel architectures.
  12. CIOQ architectures. Exact emulation of OQ with speedup. Workconserving properties for speedup: LOOFA.
  13. Networks of IQ switches. Instability phenomena.
  14. Open problems. Frame-scheduling and delay control. Constrained switching systems. Power-sensitive scheduling. Multicast support: minimum speedup and optimal queueing.

About the researcher: Marco Ajmone Marsan is a Full Professor at the Electronics Department of Politecnico di Torino, in Italy. He holds degrees in Electronic Engineering from Politecnico di Torino and University of California, Los Angeles. Since November 1975 to October 1987 he was at the Electronics Department of Politecnico di Torino, first as a Researcher, then as an Associate Professor. Since November 1987 to October 1990 he was a Full Professor at the Computer Science Department of the University of Milan, in Italy. During the summers of 1980 and 1981 he was with the Research in Distributed Processing Group, Computer Science Department, UCLA. During the summer of 1998 he was an Erskine Fellow at the Computer Science Department of the University of Canterbury in New Zealand. He has coauthored over 300 journal and conference papers in the areas of Communications and Computer Science, as well as the two books "Performance Models of Multiprocessor Systems" published by the MIT Press, and "Modelling with Generalized Stochastic Petri Nets" published by John Wiley. He received the best paper award at the Third International Conference on Distributed Computing Systems in Miami, Fla., in 1982. His current interests are in the £elds of performance evaluation of communication networks and their protocols. He currently serves on the editorial boards of the IEEE/ACM Transactions on Networking; the Optical Networks Magazine; the Photonic Network Communications Journal; the Wireless Communications and Mobile Computing Journal. He was a keynote speaker at IPCCC 98 and ITC 17. M. Ajmone Marsan is a Fellow of IEEE.

Tutorial 3
Query processing in peer-to-peer systems
Jiannong Cao
Dept. of Computing, Hong-Kong Polytechnic University, Hong-Kong
Sexta-feira 02/06/2006, 14:00-18:00

Abstract: Peer-to-Peer (P2P) systems have become popular in the last decade mainly due to their desirable properties such as self-organization, flexibility, stability, fault resilience, and robustness. One of the most important applications of P2P systems is resource sharing. P2P systems have evolved from simple keyword search to supporting the sharing of structured and semantically rich data, which requires the research on data management issues. In this tutorial, we will focus on query processing in P2P systems which allows the system to discover the peers that contribute relevant data for a given query and efficiently execute the query. We will discuss challenging issues in providing robust query support for highly dynamic P2P systems. These issues include construction of P2P overlay network, query plan, query routing and range queries. Various approaches and techniques will be presented and analyzed.

About the researcher: Dr. Cao is a professor and the director of the Internet and Mobile Computing Lab in the Department of Computing, Hong Kong Polytechnic University. He received the B.S. degree from Nanjing University, China, and the M.S. degree and Ph.D. degree from Washington State University, USA. Before joining Hong Kong Polytechnic University, Dr. Cao has been on the faculty of computer science at James Cook University and University of Adelaide in Australia, and City University of Hong Kong. Dr. Cao's research interests include parallel and distributed systems, computer networks, mobile and wireless computing, fault tolerance, and programming methodology and environments. Dr. Cao is a vice chairman and member of the Technical Committee on Computer Architectures, an executive member of the System Software Committee, and a senior member of China Computer Federation. He is also a senior member of the IEEE (including the Computer Society and the IEEE Communication Society), and a member of ACM. Dr. Cao has served as an area editor and a member of editorial boards of several international journals, a reviewer for international journals / conference proceedings, and also as an organizing / program committee member for many international conferences.


Palestra 1
Systems Management Benchmarking or: How to make the Business Case for Systems Management
Alexander Keller
IBM Research Division, Thomas J. Watson Research Center, USA
Quarta-feira 31/05/2006, 14:00-16:00

Abstract: Over the last years, the Network and Systems Management community has witnessed a shift away from information models and protocols towards value-added management services that improve the configuration and fault management of a distributed system, or optimize its performance. However, whenever we try to articulate the value of Systems Management, it turns out that we have a hard time to assess where we as a discipline are on the maturity curve and how 'self-managing' the systems we build actually are. Metrics such as 'total cost of ownership' or 'number of servers per administrator' are often overly simplistic and essentially focus just on symptoms, not on the true factors that impact the value of management. It is therefore hard to quantify the value that the investment in management technology actually yields, and to find metrics that facilitate the comparison between management systems from different vendors. This keynote address will introduce the emerging discipline of Systems Management Benchmarking and identify an initial set of metrics that are helpful for measuring the value of Systems Management.

Palestra 2
From Packet-Level Models to Fluid Models of IP-Networks: a Step Towards Scalability
Marco Ajmone Marsan
Electronics Department, Istituto Politecnico di Torino, Italy
Quarta-feira 31/05/2006, 16:30-18:30

Abstract: Fluid models of IP networks are based on a set of ordinary differential equations, that provide an abstract {deterministic} description of the average network dynamics. When IP networks operate close to saturation, fluid models were proved to provide reliable performance estimates for networks of reasonable size. Instead, when the network load is well below saturation, standard fluid models lead to wrong performance predictions, since all buffers are forecasted to be always empty, so that the packet discard probability is predicted to be zero. These incorrect predictions are due to the fact that fluid models, being deterministic in nature, do not account for the random traffic variations that may induce temporary congestion of some network elements. In this talk we discuss three different approaches to describe random traffic variations in fluid models, considering randomness at both the flow and packet levels. With these approaches, fluid models allow reliable results to be obtained also in the case of IP networks that operate well below their saturation load. Numerical results are presented to prove the accuracy and the versatility of the proposed approaches, considering both stationary and non-stationary traffic regimes.

Palestra 3
Cooperative Caching in Mobile Peer-to-Peer Systems over MANET
Jiannong Cao
Dept. of Computing, Hong-Kong Polytechnic University, Hong-Kong
Quinta-feira 01/06/2006, 18:45-20:45

Abstract: Data dissemination and sharing among peer nodes in a mobile P2P network is an important problem and has many applications. However, developing high performance techniques for data transmission in a peer-to-peer system over a mobile ad hoc network (MANET) is a difficult task. Cooperative caching can be used as an effective means to improve the availability and scalability in data sharing and dissemination as well as to reduce the network traffic and query delay. In this talk, I will address the major issues of cooperative caching such as placement, discovery, and consistency. In particular, I will present our research on caching consistency, which has not been adequately addressed for MANETs. I will describe a framework which provides a basis for designing, analyzing and evaluating existing cache invalidation strategies, and then propose a relay peer based cache consistency maintenance strategy that is able to provide a generic and flexible method to carry out the cache invalidation.

Palestra 4 Autonomic Network Management & Operations
Marcelo Perazolo, IBM Watson Research, USA
Terça-feira 30/05, 08:30-09:30, Sala 3 (WGRS) (em inglês)

Abstract: Autonomic Computing is a new paradigm being promoted in the industry in the past few years. It is a collection of new concepts where a layered architecture is applied to facilitate autonomous management and operation of software and hardware elements, in an attempt to diminish human involvement in repetitive tasks. Autonomic Computing has been applied successfully in many areas, especially in management of distributed elements and applications in the enterprise. In this talk we will explain the fundamentals of the autonomic computing architecture, and show how it can be modeled and applied to the area of network management and operation. Specific problems related to distributed elements in a network will be addressed, and autonomic computing solutions will be proposed. In addition, a parallel will be drawn with the IT Infrastructure Library (ITIL), which is a de-facto standard for process-oriented IT service management in the industry today. Specific ITIL processes, such as Incident Management and Problem Management, will be depicted with emphasis on their Network Management and Operations aspects, and Autonomic Computing concepts will be applied to help streamline the integration of process and business goals with the management and operation of the network infrastructure. Standard artifacts such as the emerging OASIS Web Services Distributed Management (WSDM) Event Format (WEF), and novel concepts such as Autonomic Computing Symptoms will be presented and applied to the proposed autonomic network management architecture to help achieve event correlation, and autonomous analysis and reaction to network incidents and problems. Finally, future trends in the area of autonomic network management and operations will be discussed.

Marcelo Perazolo is a member of the IBM Autonomic Computing Architecture Group, where he serves as an architect for symptoms and other knowledge formats and defines taxonomies and ontologies related to autonomic computing. Marcelo is also responsible by the IBM autonomic incident & problem management and virtualization initiatives. He has worked for IBM since 1990, with various assignments in the area of network and systems management. Marcelo received an M.S. degree in Electrical Engineering in 1994 from State University of Campinas (Unicamp), Brazil. He is active in promoting autonomic computing to the industry and faculty. He has many technical papers published in the field, and is involved in many research projects with IBM Research as well as University applied research projects that he helps to fund. Marcelo is also involved with standardization activities in organization such as OASIS and DMTF. His interests include problem determination, prediction and avoidance, process optimization techniques, network & systems management, security management, correlation technologies, and knowledge representation.

Palestra 5
Networking and Distributed Systems Research at Caltech's HEP group
Xun Su, High Energy Physics Group, California Institute of Technology, USA
Terça-feira 30/05, 17:45 - 19:15, Sala 1+2

Abstract: In this talk I will give an overview of the networking and distributed systems development at High Energy Physics group at California Institute of Technology, a US CMS Tier2 center which also manages the trans-Atlantic LHCNet from CERN to US. I will cover our activities in infrastructure building, network protocol research, and grid monitoring system development. In particular, I will present the LHCNet operation, MonALISA/VINCI monitoring system, FAST TCP protocol development, and Ultralight testbed. I will conclude with our work with our Brazilian partners to help bridge the "digital divide".

Xun Su is a research engineer at California Institute of Technology working on various networking and grid systems projects funded by US NSF and DOE. He is the lead engineer of the CHEPREO project, a cyberinfrastructure and physics educational outreach collaboration between US and Brazilian networking and physics research communities. He graduated from The University of Texas of Austin with a Ph.D. in Electrical Engineering, and is currently interested in network measurement, online and approximation algorithms, and information theory.

Palestra 6
MyProxy: A Multi-Purpose Grid Authentication Service
Jim Basney, NCSA, University of Illinois at Urbana-Champaign, USA
Sexta-feira 02/06, 10:30-11:30, Sala 2 (WCGA) (em inglês)

Abstract: MyProxy is a Globus Toolkit component for managing certificates and private keys that that has been used by the grid community since 2000 in projects such as EGEE, EU DataGrid, Earth System Grid, FusionGrid, LHC Computing Grid, NASA Information Power Grid, NEESgrid, Open Science Grid, and TeraGrid. In the past year, we have added the ability for MyProxy to act as an online Certificate Authority (CA) in addition to a credential repository and have added support for additional authentication methods via PAM and the Pubcookie web single-sign on system. These new capabilities allow MyProxy to support new grid authentication scenarios. In this talk, I will describe these new MyProxy capabilities and how they are being used in grid projects.

Jim Basney is a Senior Research Scientist at the National Center for Supercomputing Applications at the University of Illinois, where he leads the MyProxy project. His research focuses on the security, reliability, efficiency, and usability of distributed systems. Dr. Basney obtained his Doctorate and Masters degrees in Computer Sciences from the University of Wisconsin-Madison.

Palestra 7
Research in Power Management: Past and Future
Daniel Mossé, University of Pittsburgh, USA
Sexta-feira 02/06, 14:30-15:30, Sala 4 (WTR) (em inglês)

Abstract: Following the footsteps of classical scheduling theory, power management in real-time systems started with simple results, unrealistic assumptions, attempts to simplify the problem and elegant solutions. As the field matured, assumptions started to become more realistic, and the the issues evolved from power management in single CPU scheduling to encompass many different incarnations. Among those are multiple-CPU scheduling, multi-device scheduling, to multi-computer scheduling, web servers, thermal management, wireless interfaces, and the like. Future challenges will also be discussed from the perspective of research and practice, because the practical aspects of such a system-oriented research field should not be ignored.

Daniel Mosse received a B.S. in Mathematics from the University of Brasilia in 1986, and M.S. and Ph.D. degrees in Computer Science from the University of Maryland in 1990 and 1993, respectively. He joined the faculty of The University of Pittsburgh in 1992, where he is currently a Professor. His research interest include fault-tolerant and real-time systems, as well as networking. The major thrust of his research in the new millennium is real-time systems, power management issues, and networks (wireless and security). Typically funded by NSF and DARPA, Dr. Mosse's projects combine theoretical results and actual implementations. Dr. Mosse has served on program committees for all major IEEE- and ACM-sponsored real-time related conferences and as the program and general chairs for RTSS, RTAS, Brazilian RT Workshp, and RT Education Workshop. Dr. Mosse has been an associate editor of IEEE Transactions on Computers and is currently on the editorial board of the Journal of Real-Time Systems. He is currently a member of IEEE Computer Society, the Association for Computing Machinery, and the American Society for Engineering Education.

Palestra 8
Tolerância a Faltas Bizantinas: Desafios e Resultados Recentes
Alysson Neves Bessani, UFSC
Segunda-feira 29/06, 15:15-16:00, Sala 2 (WTF) (em português)

Abstract: O crescimento do número de incidentes de segurança em sistemas computacionais renovou o interesse pelas técnicas de tolerância a faltas Bizantinas (TFB). A idéia é que estas técnicas possam ser exploradas na construção de sistemas que mantenham um serviço íntegro mesmo quando alguns de seus componentes são controlados por adversários. Esta palestra tem por objetivo discutir alguns dos resultados mais importantes da última década em TFB, incluindo novas abtrações, sistemas e algoritmos eficientes. Além disso, pretende-se apresentar alguns temas recentes que têm despertado interesse na comunidade acadêmica (ex. algoritmos muito rápidos, memória compartilhada por processos Bizantinos, novos esquemas de replicação) e alguns desafios a serem vencidos a área (ex. independência de falhas, escalabilidade, novas aplicações).
Alysson Neves Bessani é bacharel em Ciência da Computação pela Universidade Estadual de Maringá (2001) e mestre em Engenharia Elétrica pela Universidade Federal de Santa Catarina (2002), onde está em vias de concluir seu doutorado na área de tolerância a faltas e segurança. Ele trabalha há oito anos com sistemas distribuídos e há quatro anos se concentra no estudo de técnicas e algoritmos para tolerância a faltas Bizantinas nestes sistemas.

Palestra 9
Diversificando o uso de P2P: Gerenciamento de Redes e Serviços Baseado em P2P
Lisandro Zambenedetti Granville, UFRGS
Segunda-feira 29/05, 15:00-16:00, Sala 4 (WP2P) (em português)

Abstract: Tradicionalmente, do ponto de vista de gerenciamento de redes, sistemas peer-to-peer (P2P) representam um problema de administração devido ao alto consumo de banda introduzido. Nesta apresentação, entretanto, toma-se uma nova perspectiva sobre o relacionamento entre P2P e gerenciamento de redes: sistemas P2P são usados como ferramentas de gerenciamento de redes que potencialmente podem melhorar as soluções de gerenciamento tradicionais. Sistemas P2P podem ser usados como sistemas de gerenciamento flexíveis que perpassam vários domínios administrativos. Três exemplos de gerenciamento de redes baseado em tecnologias P2P são apresentados: como visões das redes gerenciadas podem ser compartilhadas no sistema P2P, como o roteamento em nível de aplicação dos sistemas P2P pode melhorar a comunicação entre entidades de gerenciamento, e como grupos de peers de gerenciamento podem ser utilizados para implementar o balanceamento de carga nas tarefas de gerenciamento. O principal objetivo é mostrar que as tecnologias P2P podem ser utilizadas como ferramentas em outros domínios da ciência da computação, neste caso, no gerenciamento de redes.

Professor do Instituto de Informática da Universidade Federal do Rio Grande do Sul (UFRGS), onde obteve seu mestrado e doutorado. Líder do grupo de Redes de Computadores, onde desenvolve projetos de pesquisa em gerência de redes e serviços. Também é Diretor-Executivo do Laboratório Nacional de Redes de Computadores (LARC) e Coordenador da Comissão em Redes e Sistemas Distribuídos da SBC. Poster Co-Chair IEEE/IFIP NOMS 2006, Poster Co-Chair IFIP/IEEE IM 2005, General Co-Chair LANOMS 2005, Program Chair WP2P 2005, Program Chair WFS 2005. Steering committee member SBRC 2005 and SBRC 2006. Membro do comitê de programa no DSOM 2006, GLOBECOM 2006, NOMS 2006, APNOMS 2006, HiPC 2006, SBRC 2006, LAACS 2006, JAI 2006, CTIC 2006, DSOM 2005, LANC 2005, LANOMS 2005, ICETE 2005, SBC CTD 2005, SBC CTIC 2005, WFS 2005, SBRC 2005, SBrT 2005, WGRS 2005, WP2P 2005, ERRC 2005, IM 2005.

Realização Promoção Agência de Turismo
Apoio Patrocínio
CGI.br Fundação Araucária Governo do Estado do Paraná
CCVB eSysTech
24º Simpósio Brasileiro de Redes de Computadores — SBRC 2006
Curitiba, Paraná - Brasil