Página Inicial

Missão

A missão do DEI é contribuir para o desenvolvimento da sociedade, promovendo um ensino superior de excelência na área da Engenharia Informática, nas vertentes de graduação e pós-graduação, levando para tal a cabo actividades de Investigação, Desenvolvimento e Inovação ao nível dos mais elevados padrões internacionais.


Anúncios

MathWorks seminar

8 Abril 2015, 16:38 - Maria Lucilia Gonçalves Abreu

MathWorks will give a seminar on MATLAB & Simulink at IST, on April 10, 8:30-11:30, room PA2 (Pavilhão de Pós-Graduação) - Campus Alameda.

The seminar, entitled “Tips and Tricks for MATLAB and Simulink,” includes talks about the latest developments of MATLAB & Simulink, advanced programming techniques, and interfacing MATLAB & Simulink with low-cost hardware platforms such as Arduino and Raspberry Pi. One of the talks will be given by Gareth Thomas, an IST alumnus and World Wide Academic Marketing Manager of MathWorks. The seminar is free and open to the IST community. Registration is required due to the limited number of seats: https://go2.mathworks.com/tips-and-tricks-for-matlab-and-simulink-sem-es-69982

This seminar is organized jointly by the Department of Electrical and Computer Engineering (DEEC) and by the Institute for Systems and Robotics (ISR).

For more information, please contact sec@deec.ist.utl.pt or refer to the poster for this event: https://fenix.tecnico.ulisboa.pt/downloadFile/1689468335554928/mathworks_ist_isr_seminar.pdf


Todos os Anúncios

Eventos

Prova de Doutoramento

24 Abril 2015, 15:17 - Ana Maria de Almeida Nogueira Marques

Título: Data Placement in Distributed Systems

Candidato: João Gonçalves Paiva Nº 53674/D

Presidente de Júri: Professor João Paulo Marques Silva

Orientador: Professor Luís Eduardo Teixeira Rodrigues

Vogais:

Professor Vivien René Claude Quéma 

Professor Rodrigo Seromenho Miragaia Rodrigues

Professor José Carlos Alves Pereira Monteiro

Professor António Luís Pinto Ferreira Sousa

Professor David Manuel Martins de Matos

Data e local: Dia 04 de Maio às 14h00 no Anfit. PA3 (Piso-1 do Pav. de Matemática) do IST – Alameda

Abstract: Data placement refers to the problem of deciding how to assign data items to nodes in a distributed system to optimize one or several of a number of performance criteria such as reducing network congestion, improving load balancing, among others.

This issue has a significant impact on several performance criteria of a distributed system. By placing data near the clients, one may reduce the number of remote accesses, significantly reduce the latency of operations, and avoid network congestion. By taking into account the capacity of nodes and the workload characterization, one may avoid the overload of a few nodes that could otherwise become a bottleneck in the entire system. By minding the probability of failure of individual nodes, one can place data in a way that maximizes its availability, while reducing the overhead caused by monitoring and multiple replica restores. Most of these criteria impose conflicting requirements and each application must prioritize how to optimize placement. This work addresses these criteria among others, in an independent as well as in a combined way.

However, the benefits achieved by a clever data placement must be weighted against the costs of data lookup. In fact, to support total flexibility in the data placement one needs to resort to some form of distributed directory, that stores the mapping between data items and nodes. Unfortunately, the costs of performing lookups to the directory and the overhead of maintaining the directory up-to-date can easily become the bottleneck. Due to this problem, many practical systems use simple data placement strategies, such as consistent hashing.

This thesis proposes techniques that provide different tradeoffs between plain consistent hashing schemes and full directory systems for different sizes of system scales. The main goal is to provide better options between having strong flexibility with limited scalability (typically employed in datacenter systems), and having good scalability with limited flexibility (the main choice for internet scale systems).


Todos os Anúncios