In high performance computing, platforms (the supercomputers) are composed of computational resources divided into racks. Each rack have one (sometimes several) I/O nodes to access the parallel file system (PFS). Following the growth and the optimization of the machines, the amount of data involved in the calculation is increasing. As different parts of these racks are allocated to independent applications, and as the storage is shared between these, too many concurrent accesses can lead to contention and performance loss.This thesis consist in studying strategies to map applications on the platform and scheduling the different accesses to the storage system in order to avoid contention. We discuss the relevance of our solution both from th...
The field of High Performance Computing (HPC) is characterized by the continuous evolution of comput...
With the advent of multicore and manycore processors as buildingblocks of HPC supercomputers, many a...
Computing systems are becoming increasingly data-intensive because of the explosion of data and the ...
In high performance computing, platforms (the supercomputers) are composed of computational resource...
In HPC platforms, concurrent applications are sharing the same file system. This can lead to conflic...
International audienceScheduling in High-Performance Computing (HPC) has been traditionally centered...
Scheduling in High-Performance Computing (HPC) has been traditionally centered around computing reso...
Distributed applications, especially the ones being I/O intensive, often access the storage subsyste...
International audienceObservations show that some HPC applications periodically alternate between (i...
With the advent of multicore and manycore processors as building blocks of HPC supercomputers, many ...
This thesis mainly deals with the mapping and the scheduling of applications on large heteroge- neou...
This work revisits I/O bandwidth-sharing strategies for HPC applications. When several applications ...
International audienceUnmatched computation and storage performance in new HPC systems have led to a...
This paper investigates co-scheduling algorithms for processing a set of parallel applications. Inst...
High performance computers (HPCs) is the go-to solution for running computationally demanding applic...
The field of High Performance Computing (HPC) is characterized by the continuous evolution of comput...
With the advent of multicore and manycore processors as buildingblocks of HPC supercomputers, many a...
Computing systems are becoming increasingly data-intensive because of the explosion of data and the ...
In high performance computing, platforms (the supercomputers) are composed of computational resource...
In HPC platforms, concurrent applications are sharing the same file system. This can lead to conflic...
International audienceScheduling in High-Performance Computing (HPC) has been traditionally centered...
Scheduling in High-Performance Computing (HPC) has been traditionally centered around computing reso...
Distributed applications, especially the ones being I/O intensive, often access the storage subsyste...
International audienceObservations show that some HPC applications periodically alternate between (i...
With the advent of multicore and manycore processors as building blocks of HPC supercomputers, many ...
This thesis mainly deals with the mapping and the scheduling of applications on large heteroge- neou...
This work revisits I/O bandwidth-sharing strategies for HPC applications. When several applications ...
International audienceUnmatched computation and storage performance in new HPC systems have led to a...
This paper investigates co-scheduling algorithms for processing a set of parallel applications. Inst...
High performance computers (HPCs) is the go-to solution for running computationally demanding applic...
The field of High Performance Computing (HPC) is characterized by the continuous evolution of comput...
With the advent of multicore and manycore processors as buildingblocks of HPC supercomputers, many a...
Computing systems are becoming increasingly data-intensive because of the explosion of data and the ...