A Scalable Runtime for the ECOSCALE Heterogeneous Exascale Hardware Platform

Abstract

Exascale computation is the next target of high performance computing. In the push to create exascale computing platforms, simply increasing the number of hardware devices is not an acceptable option given the limitations of power consumption, heat dissipation, and programming models which are designed for current hardware platforms. Instead, new hardware technologies, coupled with improved programming abstractions and more autonomous runtime systems, are required to achieve this goal.This position paper presents the design of a new runtime for a new heterogeneous hardware platform being developed to explore energy efficient, high performance computing. By combining a number of different technologies, this framework will both simplify the programming of current and future HPC applications, as well as automating the scheduling of data and computation across this new hardware platform. In particular, this work explores the use of FPGAs to achieve both the power and performance goals of exascale, as well as utilising the runtime to automatically effect dynamic configuration and reconfiguration of these platforms.

Next generation cloud computing: New trends and research directions

“…This is because it is not yet possible for a programmer to fully develop and execute code oblivious to the underlying hardware. There are a number of efforts in this direction, but the key challenge is achieving a high-level abstraction that can be employed across multiple architectures, such as GPUs, FPGAs and Phis [29,30,31,32]. Further applications that already execute on the cloud cannot be scheduled onto heterogeneous resources.…”

“…There is another dimension to the big data computing, which is the legal issues surrounding sensitive data generated by cloud applications. There are numerous legislative and regulatory constraints that aim to account for data access control, protection, privacy and security 32 . For example in the European Union (EU) the transfer of personal data to non-member countries is forbidden, unless an adequate level of protection is ensured 33 .…”

Energy-Aware Self-Adaptation for Application Execution on Heterogeneous Parallel Architectures

“…This paper focuses not only on workload placement but also upon the exact configuration of the application, which has multiple configurations available and various possibilities to use different hardware accelerators. The EcoScale project [16], [17] uses hardware performance monitors and models to project runtime and power consumption in heterogeneous environment using accelerators. The aim is to enable a runtime scheduler called the execution engine, to dynamically select and distribute software functions to either hardware acceleration or to software based execution.…”

Energy-aware Self-Adaptive Middleware for Heterogeneous Parallel Architectures

“…StarPU [14] is one such early work that uses abstractions to allow workloads to be placed upon various different accelerator based platforms, selecting between various different accelerators, to improve computational performance and not energy. EcoScale [15], [16] uses hardware performance monitors and models to project runtime and power consumption in heterogeneous environments, with the aim of dynamically selecting and distribute software functions to either hardware acceleration or to software based execution. This paper in comparison to existing literature brings together key aspects of performance, QoS and energy management, with application migration on heterogeneous architectures.…”