395. Scheduling, key to the efficiency of parallel computing
Invited abstract in session WB-1: Oliver Sinnen, stream Keynotes.
Wednesday, 10:30-12:00Room: Great Hall
Authors (first author is the speaker)
| 1. | Oliver Sinnen
|
| University of Auckland |
Abstract
Computers, in all their forms from smartwatches to supercomputers, are a major enabling technology for remarkable advances in so many domains. Today, most of those computing systems have more than one processor (or core). But what to do with all those processors and how to use them efficiently?
Scheduling of the computing tasks and workloads is key to answer that question. Here scheduling refers to the allocation of tasks and workloads to processors and their ordering for execution, while optimising for performance. In the large majority of the cases, such scheduling is a hard combinatorial optimisation problem. In this talk we look at scheduling problems of parallel computing and the different solution techniques that are employed to solve them. This does not only address the meaningful scheduling of computations, but also the consideration of communication (data transfer) between the processors.
New challenges emerge, for example from the breathtaking advances in machine learning and its seemingly insatiable hunger for computing power. To tame the corresponding huge energy consumption, multi-objective scheduling approaches are necessary that not only minimise execution time, but also look at the optimisation of power and energy consumption.
Keywords
- Scheduling
- Machine Learning
- Multi-Objective Decision Making
Status: accepted
Back to the list of papers