WPerformance: Solenidade de Abertura do WPerformance 2009
Horário: 08:45 - 09:15
Local: Auditório 9

WPerformance: Palestra 1: "É divertido modelar e analisar sistemas de computação: o que podemos aprender dos modelos?"
Edmundo de Souza e Silva (UFRJ)
Horário: 10:30 - 12:00
Local: Auditório 9

Short Bio:
Edmundo de Souza e Silva é Engenheiro Eletrônico e Mestre em Engenharia Elétrica pela PUC/RJ e Ph.D. em Ciência da Computação pela UCLA. É atualmente Professor Titular da COPPE/UFRJ, Programa de Engenharia de Sistemas e Computação.
Edmundo foi professor visitante do centro de pesquisas T.J. Watson da IBM, Departamento de Ciência da Computação da UCLA e da USC, Politecnico de Torico, IBM Tokyo Research Laboratory e Chinese University of Hong Kong. Participou em vários comitês técnicos de conferências nacionais e internacionais, e foi co-presidente do comitê de programa técnico das conferências: the IFIP/IEEE Data Comm. & Performance (1987), SBRC'98; IEEE/GLOBECOM'99, ITC'2001 (vice-co-pres.), ACM/SIGMETRICS'2002, e IEEE/INFOCOM'2009. Tem feito parte também do corpo editorial do JBCS. Foi membro do Board of Directors da ACM/SIGMETRICS, e atualmente é presidente do IFIP W.G. 7.3 (Performance) e do board da Computer Performance Foundation.
Edmundo é pesquisador I-A do CNPq, e membro da Academia Brasileira de Ciências. Suas principais linhas de pesquisa versam sobre modelagem e análise de sistemas de computação (performance e confiabilidade) e rede de computadores.
 
Resumo:
A modelagem é um ingrediente natural e imprescindível do processo de projeto e desenvolvimento da maioria dos sistemas e, em particular, sistemas de computação. Modelos são necessários para entender o funcionamento de um sistema ou para prever o comportamento futuro deste quando submetido a determinadas condições que são especificadas pelo projetista/analista. Citando apenas alguns poucos exemplos, existem modelos para prever o tempo, modelos econômicos, modelos biológicos, e modelos de sistemas de computação que é o foco desta palestra. Nosso objetivo é descrever o processo de modelagem fazendo uso de várias aplicações em sistemas de computação. Não nos preocuparemos na solução dos modelos, e sim nas aplicações. Em alguns dos exemplos, ilustraremos o processo de modelagem usando cadeias de Markov como uma das ferramentas largamente usadas na construção e solução de um modelo
Modelos de desempenho e confiabilidade, modelos de mecanismos e protocolos de redes de computadores, modelos de sistemas P2P, "ranqueamento" de páginas web a-lá Google, e ainda exemplos encontrados em áreas distintas da computação serão empregados durante a palestra.

WPerformance: Palestra 2: "Block Switching: Towards a Robust Protocol Stack for Diverse Wireless Networks"
Arun Venkataramani (University of Massachussets Amherst, EUA)
Horário: 16:15 - 18:15
Local: Auditório 9

Short Bio:
Arun Venkataramani has been an Assistant Professor in the Department of Computer Science at the University of Massachusetts Amherst since Spring 2005. He completed his PhD in Computer at the University of Texas at Austin in 2004 and was a Visiting Faculty at University of Washington in Fall 2004. He received his bachelor's degree in Computer Science and Engineering at IIT Bombay in 1999. His research interests are in networking and distributed systems and his current research focuses on wireless and mobile networks, peer-to-peer systems, network architecture, virtualization, and security. His dissertation work developed TCP Nice, a protocol for low priority background transport without network support. He is a recipient of an NSF CAREER award and his work has received several best paper awards including USENIX NSDI 2007 and 2008. He has served on the program committees of ACM Sigcomm, USENIX NSDI, IEEE Infocom and several other conferences in networking and systems.

Abstract:
The Internet's TCP/IP stack was not designed with the vagaries of the wireless channel and mobility in mind and is showing signs of age today. For example, TCP has well-known problems over wireless networks as it conflates congestion and loss, performs poorly over lossy or fluctuating links, and is fragile in the presence of route changes and disconnections induced by mobility.
This talk is motivated by the following question: how can we design a protocol stack that ensures robust performance across diverse wireless networks such as access point networks, multi-hop mesh networks, mobile ad hoc networks, and disruption-tolerant networks? To address this question, I will present two ideas and proof-of-concept systems. The first is to use "reliable per-hop block transfer" as a building block for wireless transport. I will show the benefits of the block transfer approach through the design and implementation of Hop, a wireless transport protocol that achieves dramatically better throughput, fairness, delay, and robustness to partitions. The second is the use of "block replication" as opposed to simple packet forwarding for routing under uncertainty through the design and implementation of RAPID, a disruption-tolerant network routing protocol. Finally, I will outline research challenges and opportunities enabled by block transfer in order to realize the vision of a robust protocol stack for diverse wireless networks.

"Desempenho de Algoritmos de Proteção por Caminho Compartilhado Sensíveis às Limitações da Camada Física em Redes WDM"
Sávio Rodrigo Antunes dos Santos Rosa (UNICAMP), André Drummond (UNICAMP), Nelson Fonseca (UNICAMP)

WPerformance: Palestra 3: "Queues Are Everywhere (in Performance Modeling and Analysis)"
Richard Gail (IBM T.J. Watson Research Center, USA)
Horário: 10:45 - 11:45
Local: Auditório 9

Short Bio:
Richard Gail received the BA in mathematics from the University of California, Riverside (UCR), the MA in mathematics from the University of California,Los Angeles (UCLA) and the PhD in Engineering from UCLA under the direction of Leonard Kleinrock. He then joined the IBM T.J. Watson Research Center in Yorktown Heights, NY. At IBM he worked on a software tool for the design and analysis of computer networks, and he also worked on theoretical problems about priority queues, Neuts-type queues (M/G/1 and GI/M/1 systems), and multidimensional Markov chains. Other areas of work included buffering algorithms, performability theory and a software tool for call centers to diagnose hardware problems.
He received an IBM Outstanding Innovation Award, an IBM Research Division Award, and was also awarded four US patents. He was a visiting researcher at Universidade Federal do Rio de Janeiro (UFRJ Brasil), Chinese University of Hong Kong (China) and Tsukuba University (Japan). Currently he has a visiting professor appointment at UFRJ. His major research interests include queueing theory and performability modeling and analysis.
 
Abstract:
Queueing models constitute an important building block in the specification of performance models of computer systems and networks. Several different queueing models are reviewed and a discussion of their application and solution is presented.
Examples include polling system models, switch and multiplexor models, and models of server farms.
In addition the use of queueing systems as a part of the development of certain availability/performability models is described.
An introduction to several of the techniques utilized in the analysis of queueing systems with special structure is also briefly discussed.

WPerformance: Painel: "Ensino e pesquisa em análise de desempenho"
Horário: 16:15 - 18:15
Local: Auditório 9