Abstract
The spectacular growth in the number of cores in current supercomputers poses design challenges for the development of performance analysis and tuning tools. To be effective, such analysis and tuning tools must be scalable and be able to manage the dynamic behaviour of parallel applications. In this work, we present ELASTIC, an environment for dynamic tuning of large-scale parallel applications. To be scalable, the architecture of ELASTIC takes the form of a hierarchical tuning network of nodes that perform a distributed analysis and tuning process. Moreover, the tuning network topology can be configured to adapt itself to the size of the parallel application. To guide the dynamic tuning process, ELASTIC supports a plugin architecture. These plugins, called ELASTIC packages, allow the integration of different tuning strategies into ELASTIC. We also present experimental tests conducted using ELASTIC, showing its effectiveness to improve the performance of large-scale parallel applications.
Highlights
Supercomputers are a widely used resource in many areas of modern research
Dynamic analysis and tuning of the application during its execution, such as that performed by MATE [10], Active Harmony [14], and other tools, is necessary. Most of these tuning tools do not scale well, mainly due to a centralised analysis process. Taking into consideration these facts, this paper addresses the lack of large-scale dynamic tuning in the current performance analysis area
None of them, except for latest efforts in Periscope under the AutoTune Project [8], consider application tuning. Having these facts in mind, in this work we address the lack of large-scale dynamic tuning in the current performance analysis area
Summary
Supercomputers are a widely used resource in many areas of modern research. they are a costly resource, and access is often limited to an allocation of execution hours.Normally, parallel applications running on supercomputers do not make efficient use of resources. Supercomputers are a widely used resource in many areas of modern research. They are a costly resource, and access is often limited to an allocation of execution hours. Parallel applications running on supercomputers do not make efficient use of resources. This provokes longer than expected running times, which “waste” computation hours and reduce the available time for further executions. In this context, analysis and tuning tools that identify, understand and fix performance problems are more valuable than ever. To apply performance analysis and tuning to parallel applications executed on supercomputers, it is paramount that these tools have been designed following a scalable and modular architecture that enables the control and analysis of an extremely large number of tasks
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
