TY - GEN
T1 - EdgeCloudBenchmark
T2 - 6th IEEE International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2022
AU - Wen, Shilin
AU - Deng, Hongjie
AU - Qiu, Ke
AU - Han, Rui
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - With the rapid development of 5G and IoT technology, edge computing, as an extension of the cloud computing paradigm, has been widely used to handle some latency-sensitive tasks. Due to insufficient and limited resource of edge devices, when the edge handles some complex tasks, it is often necessary to cooperate with the cloud, which forms the cloud-edge collaboration scenarios. In real cloud-edge collaboration cluster, different scheduling algorithms will greatly affect the resource allocation and workload completion time. Therefore, how to measure the quality of a scheduling algorithm has become critical. However, there is no existing benchmark test sets for such scenarios at present. Based on this problem, this paper proposes EdgeCloudBenchmark, which is a benchmark generation system driven by real Alibaba cluster trace. In this system, we can generate two different benchmark test sets for CPU cluster and GPU cluster, respectively. The experimental results show that these workloads generated from the proposed system can maintain the consistency with the characteristics of the real cluster workloads, and are highly available. Therefore, our proposed system has high concurrency, availability and fault tolerance.
AB - With the rapid development of 5G and IoT technology, edge computing, as an extension of the cloud computing paradigm, has been widely used to handle some latency-sensitive tasks. Due to insufficient and limited resource of edge devices, when the edge handles some complex tasks, it is often necessary to cooperate with the cloud, which forms the cloud-edge collaboration scenarios. In real cloud-edge collaboration cluster, different scheduling algorithms will greatly affect the resource allocation and workload completion time. Therefore, how to measure the quality of a scheduling algorithm has become critical. However, there is no existing benchmark test sets for such scenarios at present. Based on this problem, this paper proposes EdgeCloudBenchmark, which is a benchmark generation system driven by real Alibaba cluster trace. In this system, we can generate two different benchmark test sets for CPU cluster and GPU cluster, respectively. The experimental results show that these workloads generated from the proposed system can maintain the consistency with the characteristics of the real cluster workloads, and are highly available. Therefore, our proposed system has high concurrency, availability and fault tolerance.
KW - benchmark test sets
KW - cloud-edge collaboration
KW - edge computing
UR - http://www.scopus.com/inward/record.url?scp=85141536647&partnerID=8YFLogxK
U2 - 10.1109/SDPC55702.2022.9915888
DO - 10.1109/SDPC55702.2022.9915888
M3 - Conference contribution
AN - SCOPUS:85141536647
T3 - Proceedings of 2022 IEEE International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2022
SP - 377
EP - 382
BT - Proceedings of 2022 IEEE International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2022
A2 - Yu, Qibing
A2 - Cabrera, Diego
A2 - Luo, Jiufei
A2 - Pu, Zhiqiang
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 5 August 2022 through 7 August 2022
ER -