Vanderlei Munhoz presented his work entitled Strategies for Fault-Tolerant Tightly-coupled HPC Workloads Running on Low-Budget Spot Cloud Infrastructures at the EEE International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD 2022), which was held in Bordeaux, France. This work was coauthored by Prof. Márcio Castro and Prof. Odorico Mendizabal.
In this work, Vanderlei evaluated the viability of budget-constrained cloud environments for tightly-coupled MPI applications, exploring both spot and traditional low-budget infrastructures from real public cloud platforms. Two different fault tolerance strategies tailored for unreliable spot cloud environments have been proposed: system-level rollback restart with Berkeley Labs Checkpoint/Restart (BLCR) and in-memory rollback restart with User-Level Failure Mitigation (ULFM). A provider-agnostic empirical method for testing and predicting MPI workloads execution times and cloud infrastructure costs was also proposed. The results showed that: (i) adequate cluster sizing plays an important role in the overall job execution performance and cost-effectiveness, regardless of the type of selected instances; (ii) fault tolerance strategies based on BLCR may have worse performance than ULFM, but still be cost-effective considering software migration costs; (iii) the use of spot infrastructure does not guarantee costs savings depending on the chosen machine flavors and discounts, as experiments with persistent low-budget options attained better cost-effectiveness in some conditions.
João Vicente Souto has earned his master’s degree in Computer Science from PPGCC/UFSC on 14/07/2022. His M.Sc. dissertation entitled “A Task-based Execution Engine for Distributed Operating Systems Tailored to Lightweight Manycores with Limited On-Chip Memory” was advised by professor Márcio Castro. The jury was composed by the following members: Rômulo de Oliveira (UFSC), Pedro Penna (Microsoft Research), Odorico Mendizabal (UFSC).
The defense is available at: https://www.youtube.com/watch?v=ptoyABxC6fA&t=252s.
The “XXIII Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD 2022)”, which is the brazilian conference on High Performance Computing (HPC), was organized by Prof. Márcio Castro (UFSC) and Prof. Odorico Mendizabal (UFSC) and was held at the Majestic Palace Hotel in Florianópolis/SC. Students from LaPeSD and PET/UFSC have also participated as local staff. In this year, more than 140 people came to Florianópolis/SC to participate in the event.
João Vicente Souto is the winner of the M.Sc. Dissertation Contest in Computer Architecture and High Performance Computing (WSCAD-CTD 2022). In his work, entitled “A Task-based Execution Engine for Distributed Operating Systems Tailored to Lightweight Manycores with Limited On-Chip Memory“, João proposed a task-based execution engine that allows the implementation of Distributed Operating System services with low memory consumption and low latency. He was advised by Prof. Márcio Castro (UFSC).
The paper entitled “Virtualização e Migração de Processos em um Sistema Operacional Distribuído para Lightweight Manycores”, co-authored by Nicolas Vanz, João Souto and Márcio Castro (UFSC), received the Best Paper Award at the XXII Escola Regional de Alto Desempenho da Região Sul (ERAD/RS 2022), which took place in Curitiba/PR, Brazil from April 18th to 20th, 2022. Nicolas is an undergraduate student supervised by Prof. Márcio working on process migration in a Operating System that targets lightweight manycore processors.
Students from LaPeSD participated in the Marathon of Parallel Programming, which was held in conjunction with International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD) and Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD) (online events). The teams were composed of the following students:
- LaPeSD/UFSC: Emmanuel Podestá Jr., João Vicente Souto and Luiz Gustavo Xavier
- LaPeSD Equipe 2: Sérgio Pfleger, Vanderlei Munhoz and José Luiz de Sousa
Both teams were coached by Prof. Márcio Castro and they had about 28 hours to paralelize and optimize up to 5 problems (higher the performance improvements, the higher the score achieved by the teams). The final score board is shown below.
This year, we are very proud to announce that “LaPeSD/UFSC” and “LaPeSD Equipe 2” teams were ranked first (score 184.470338) and fourth (score 67.346131) places in the competition, respectively! Congratulations guys!
João Fellipe Uller has defended his final project on 13/05/2021. His research work entitled “A Simple MPI Library for Lightweight Manycore Processors” was advised by prof. Márcio Castro and Pedro Henrique Penna. The jury was composed by prof. Odorico Mendizabal (UFSC) and prof. Frank Siqueira (UFSC).
Vinicius Freitas received the Best MSc Student Award of 2020 from the Graduate Program in Computer Science (PPGCC) of UFSC. In his work, entitled “Diffusive Load Balancing for High Performance Computing Applications: Workload Discretization and Graph Processing“, Vinicius proposed distributed load balancers for highly parallel platforms. He was advised by Prof. Márcio Castro (UFSC) and coadvised by Prof. Laércio Pilla (CNRS – France).
You can check out his MSc defense on Youtube.
João Fellipe Uller presented his work entitled Enhancing Programmability in NoC-Based Lightweight Manycore Processors with a Portable MPI Library at the Simpósio em Sistemas Computacionais de Alto Desempenho (WSCAD), which was held virtually this year due to the COVID-19 pandemic. This work was advised by Prof. Márcio Castro.
In this work, João proposed a portable and lightweight MPI library (LWMPI) designed from scratch to cope with restrictions and intricacies of lightweight manycores. This library was integrated into a distributed OS that targets these processors and evaluated it on the Kalray MPPA-256 processor. Results obtained with three applications from a representative benchmark suite unveiled that LWMPI achieves similar performance scalability in comparison with the low-level vendor-specific API narrowed for MPPA-256, while exposing a richer programming interface.
The paper presentation is available at Youtube.
João Vicente Souto and João Fellipe Uller presented their papers at the Escola Regional de Alto Desempenho da Região Sul (ERAD/RS). This year, ERAD/RS was a virtual conference due to the COVID-19 pandemic. Their works were advised by Prof. Márcio Castro.
Papers presented at ERAD/RS 2020:
- Mecanismos de Comunicação entre Clusters para Lightweight Manycores no Nanvix OS, authored by João Vicente Souto (UFSC), Pedro Henrique Penna (PUC Minas), Márcio Castro (UFSC) and Henrique Cota Freitas (PUC Minas). Click here to see his presentation at Youtube.
- Proposta de Suporte ao Padrão MPI sobre Infraestrutura de Comunicação de Baixo Nível no Nanvix OS, authored by João Fellipe Uller (UFSC), João Vicente Souto (UFSC), Pedro Henrique Penna (PUC Minas) and Márcio Castro (UFSC). Click here to see his presentation at Youtube.