No abstract

Abstract. This work presents a performance evaluation of single node and subdomain communication schemes available in EdgeCFD, an implicit edgebased coupled fluid flow and transport code for solving large scale problems in modern clusters. A natural convection flow problem is considered to assess performance metrics. Tests, focused in single node multi-core performance, show that past Intel Xeon processors dramatically suffer when large workloads are imposed to a single node. However, the problem seems to be mitigated in the newest Intel Xeon processor. We also observe that MPI non-blocking pointto-point interface sub-domain communications, although more difficult to implement, are more effective than collective interface sub-domain communications.

Abstract. High Performance Computing (HPC) is becoming much more popular nowadays. Currently, the biggest supercomputers in the world have hundreds of thousands of processors and consequently may have more software and hardware failures. HPC centers managers also have to deal with multiple clusters from different vendors with their particular architectures. However, since there are not enough HPC experts to manage all the new supercomputers, it is expected that non-experts will be managing those large clusters. In this paper we study the new challenges to manage HPC environments containing different clusters with different sizes and architectures. We review available tools and present LEMMing [1], an easy-to-use open source tool developed to support high performance computing centers. LEMMing integrates machine resources and the available management and monitoring tools on a single point of management.

The High-Performance Computing Center (NACAD) of COPPE / UFRJ is a laboratory specialized in applying high-performance computing to engineering and science problems in general. NACAD also has extensive experience in the administration, management, and tools development to support the Supercomputing Center and to develop and implement innovations in the machine management environment. The minicourse proposes sharing some of the best practices adopted by NACAD-COPPE/UFRJ for deploying an HPC cluster using a public cloud. ResumoO Núcleo Avançado de Computação de Alto Desempenho (NACAD) da COPPE/UFRJ é um laboratório especializado na aplicação de computação de alto desempenho a problemas de engenharia e ciências em geral. O NACAD também possui grande experiência na administração, gerencia e ferramentas de apoio ao Centro de Supercomputação e de desenvolver e implementar inovações no ambiente de administração e gerencia da máquina. O presente minicurso propõe compartilhar algumas dessas melhores práticas adotadas pelo NACAD-COPPE/UFRJ para a implantação de um cluster de Alto Desempenho usando uma nuvem pública.