Focus Session
The Value of Data: Using Constraints to Drive Efficiencies in HPC and Cloud Computing
Event Type
Focus Session
TimeMonday, June 25th4:40pm - 5pm
LocationPanorama 3
DescriptionHistorically, high-performance computing has focused on solving tightly coupled computations with high efficiency. This task-oriented perspective is being supplemented more recently by a data-oriented perspective. The cloud computing paradigm leverages the ability to provision computational resources elastically to address data-oriented use cases in machine learning, optimization, and other similar areas. Often, these use cases involve streaming data from sources such as distributed sensor networks or Internet of Things devices that require increasingly sophisticated techniques for ensuring that data is processed in a timely manner. To that end, data has objective value and often may have time-value, and these notions of value represent constraints that must be addressed. As the field of high-performance computing receives more influence from data-oriented use cases, resource management techniques used in supercomputing must adapt to managing resources in relation to new constraints such as the time-value of data while remaining aligned with traditional constraints related to managing resources such as nodes, processors, and I/O. This talk will discuss techniques for effectively managing these constraints in high-performance computing and cloud computing environments.