Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems
As recent heterogeneous systems comprise multi-core CPUs and multiple GPUs, efficient allocation of multiple data-parallel applications has become a primary goal to achieve both maximum total performance and efficiency. However, the efficient orchestration of multiple applications is highly challeng...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-11-01
|
Series: | Electronics |
Subjects: | |
Online Access: | https://www.mdpi.com/2079-9292/9/11/1825 |
id |
doaj-f5cdc95c70d2496bbae7e8282a05d731 |
---|---|
record_format |
Article |
spelling |
doaj-f5cdc95c70d2496bbae7e8282a05d7312020-11-25T04:04:21ZengMDPI AGElectronics2079-92922020-11-0191825182510.3390/electronics9111825Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous SystemsDonghyeon Kim0Seokwon Kang1Junsu Lim2Sunwook Jung3Woosung Kim4Yongjun Park5Department of Computer Science, Hanyang University, Seoul 04763, KoreaDepartment of Computer Science, Hanyang University, Seoul 04763, KoreaDepartment of Computer Science, Hanyang University, Seoul 04763, KoreaDepartment of Computer Science, Hanyang University, Seoul 04763, KoreaLG Electronics CTO Software Platform Laboratory, Seoul 06772, KoreaDepartment of Computer Science, Hanyang University, Seoul 04763, KoreaAs recent heterogeneous systems comprise multi-core CPUs and multiple GPUs, efficient allocation of multiple data-parallel applications has become a primary goal to achieve both maximum total performance and efficiency. However, the efficient orchestration of multiple applications is highly challenging because a detailed runtime status such as expected remaining time and available memory size of each computing device is hidden. To solve these problems, we propose a dynamic data-parallel application allocation framework called ADAMS. Evaluations show that our framework improves the average total execution device time by 1.85× over the round-robin policy in the non-shared-memory system with small data set.https://www.mdpi.com/2079-9292/9/11/1825device abstractiondynamic resource managementGPGPUsheterogeneous system architecturemultitaskingOpenCL |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Donghyeon Kim Seokwon Kang Junsu Lim Sunwook Jung Woosung Kim Yongjun Park |
spellingShingle |
Donghyeon Kim Seokwon Kang Junsu Lim Sunwook Jung Woosung Kim Yongjun Park Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems Electronics device abstraction dynamic resource management GPGPUs heterogeneous system architecture multitasking OpenCL |
author_facet |
Donghyeon Kim Seokwon Kang Junsu Lim Sunwook Jung Woosung Kim Yongjun Park |
author_sort |
Donghyeon Kim |
title |
Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems |
title_short |
Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems |
title_full |
Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems |
title_fullStr |
Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems |
title_full_unstemmed |
Resource-Aware Device Allocation of Data-Parallel Applications on Heterogeneous Systems |
title_sort |
resource-aware device allocation of data-parallel applications on heterogeneous systems |
publisher |
MDPI AG |
series |
Electronics |
issn |
2079-9292 |
publishDate |
2020-11-01 |
description |
As recent heterogeneous systems comprise multi-core CPUs and multiple GPUs, efficient allocation of multiple data-parallel applications has become a primary goal to achieve both maximum total performance and efficiency. However, the efficient orchestration of multiple applications is highly challenging because a detailed runtime status such as expected remaining time and available memory size of each computing device is hidden. To solve these problems, we propose a dynamic data-parallel application allocation framework called ADAMS. Evaluations show that our framework improves the average total execution device time by 1.85× over the round-robin policy in the non-shared-memory system with small data set. |
topic |
device abstraction dynamic resource management GPGPUs heterogeneous system architecture multitasking OpenCL |
url |
https://www.mdpi.com/2079-9292/9/11/1825 |
work_keys_str_mv |
AT donghyeonkim resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems AT seokwonkang resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems AT junsulim resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems AT sunwookjung resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems AT woosungkim resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems AT yongjunpark resourceawaredeviceallocationofdataparallelapplicationsonheterogeneoussystems |
_version_ |
1724437207673995264 |