IANOS: An Intelligent Application Oriented Scheduling Framework For An HPCN Grid

Abstract

We present the architecture and implementation of IANOS middleware that is a result of integration of the Intelligent Scheduling Service into the VIOLA meta-scheduling environment. The goal of the new, integrated environment is to provide a general middleware infrastructure allowing optimal positioning and scheduling of HPC Grid applications. The scheduling algorithms used to calculate best-suited resources are based on an objective cost function that exploits information on the parameterization of applications and resources. The overall middleware architecture comprises four general services and two modules. The implementation is based on state-of-the-art Grid and Web services technology as well as existing and emerging standards (WSRF, WS-Agreement, JSDL, and GLUE). IANOS in general is agnostic to a Grid middleware. Currently, the IANOS integration with the Unicore Grid system is under development.

Scheduling Medical Application Workloads on Virtualized Computing Systems

“…In speedup. This coincides with a similar observation made in [36] when the authors ran BLAST [37] jobs, in which they suggested that VM double caching caused the virtualized execution to be faster. …”

“…The effort described in [36] allows for cross-platform performance prediction of parallel applications. The prediction is achieved by combining the application's performance in a reference system and the relative performance between the two systems derived from a partial execution on the target platform.…”

Models and internals of the IANOS resource broker

“…The IANOS resource broker implements all the models and features developed in this paper. More detailed on the implementation can be found in [8].…”

Performance prediction of weather forecasting software on multicore systems

“…Gruber et al [15] present a similar approach in trying to model the execution time of an application on a particular set of resources for use in meta-scheduling decisions in a grid environment as part of the Ianos project. Their model includes several application-specific parameters and characteristics, which can be done with our model, but is not necessarily a prerequisite for good results.…”

“…Other works focus on system specific approaches [15]. In [16], the authors describe the use of the Dimemas tool to perform prediction of the execution behavior of message passing applications.…”