Publicado em: 16/11/2011

Proposta de Tese em Arquitetura de Sistemas Computacionais

UNIVERSIDADE FEDERAL DO RIO GRANDE DO SUL

INSTITUTO DE INFORMÁTICA

PROGRAMA DE PÓS-GRADUAÇÃO EM COMPUTAÇÃO
——————————————————————————————–
DEFESA DE PROPOSTA DE TESE

Aluno: Monica Magalhães Pereira

Orientador: Luigi Carro

Título:A Reliability Analysis Approach to Assist the Design of Aggressively Scaled Reconfigurable Architectures

Linha de Pesquisa: Arquitetura e Projeto de Sistemas Computacionais

Data: 18/11/2011

Horário: 9h

Local: 43412(65) – sala ANFV (Anfiteatro Vermelho)- Instituto de Informática

Banca Examinadora:

Prof. Dr. Antônio Otávio Fernandes (UFMG)

Prof. Dr. Flávio Rech Wagner (UFRGS)

Profa. Dra. Fernanda Gusmão de Lima Kastensmidt (UFRGS)

Presidente da Banca*: Luigi Carro

Abstract:

As computer systems are built with aggressively scaled and unreliable technologies, some implementations rely on function specialization with reconfigurable computing to increase performance by exploiting parallelism, with possible energy gains. However, the use of reconfigurable devices in general purpose computing also brings extra reliability challenges at the system level. Solutions to cope with that are generally accompanied with the addition of excessive area, performance and power overheads to the overall system. These overheads could be reduced if a more extensive analysis was performed to evaluate the best fault tolerance strategy to balance the tradeoff between reliability and the mentioned aspects. In this context, we propose a comprehensive analysis of architectural design that includes the use of eliability modelling and takes into consideration aspects such as area, performance, and power. The analysis aims to assist the design of reliable-aware reconfigurable architectures by giving some indications about what kind of investments and where they should be made to design a reliable reconfigurable architecture, avoiding unnecessary overheads. In the proposed analysis, we show that interconnects are the critical elements to the overall reliability of reconfigurable rchitectures. Therefore, the interconnection model is the one where most of the investments should be made. The preliminary results with our case study demonstrate that one can improve the reliability by two orders of magnitude when adding 34% of redundancy to the interconnection model, with only 3% of performance degradation. On the other hand, even adding 50% of redundant functional units, there are no significant reliability gains when compared to the non-redundant architecture reliability.

Keywords:
Reconfigurable architecture, Fault tolerance, Reliability analysis, Scaling.