Design and Implementation of a General-Purpose API of Progress and Performance Indicators.
Iván Rodero, Francesc Guim, Julita Corbalán, Jesús Labarta · 2007
In High Performance Computing centers, queuing systems are used by the users to access and manage the HPC resources through a set of interfaces. After job submission, users lose control of the job and they only have a very restricted set of interfaces for accessing data concerning the performance and progress of job events. In this paper we present a general-purpose API to implement progress and performance indicators of individual applications. The API is generic and it is designed to be used at different levels, from the operating system to a grid portal. We also present two additional components built on top of the API and their use in the HPC-Europa portal. Furthermore, we discuss how to use the proposed API and tools in the eNANOS project to implement scheduling policies based on dynamic load balancing techniques and self tuning in run time. 1