Rights statement: ©2020 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Accepted author manuscript, 3.51 MB, PDF document
Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License
Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - Performance-aware Speculative Resource Oversubscription for Large-scale Clusters
AU - Yang, Renyu
AU - Sun, Xiaoyang
AU - Hu, Chunming
AU - Garraghan, Peter
AU - Wo, Tianyu
AU - Wen, Zhenyu
AU - Peng, Hao
AU - Xu, Jie
AU - Li, Chao
N1 - ©2020 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
PY - 2020/1/28
Y1 - 2020/1/28
N2 - It is a long-standing challenge to achieve a high degree of resource utilization in cluster scheduling. Resource oversubscription has become a common practice in improving resource utilization and cost reduction. However, current centralizedapproaches to oversubscription suffer from the issue with resource mismatch and fail to take into account other performance requirements, e.g., tail latency. In this paper we present ROSE, a new resource management platform capable of conducting performance-aware resource oversubscription. ROSE allows latency-sensitive long-running applications (LRAs) to co-exist with computation-intensive batch jobs. Instead of waiting for resource allocation to be confirmed by the centralized scheduler, job managers in ROSE can independently request to launch speculative tasks within specific machines according to their suitability for oversubscription. Node agents of those machines can however avoid any excessive resource oversubscription by means of a mechanism for admission control using multi-resource threshold control and performance-aware resource throttle. Experiments show that in case of mixed co-location of batch jobs and latency-sensitive LRAs, the CPU utilization and the disk utilization can reach56.34% and 43.49%, respectively, but the 95th percentile of read latency in YCSB workloads only increases by 5.4% against the case of executing the LRAs alone.
AB - It is a long-standing challenge to achieve a high degree of resource utilization in cluster scheduling. Resource oversubscription has become a common practice in improving resource utilization and cost reduction. However, current centralizedapproaches to oversubscription suffer from the issue with resource mismatch and fail to take into account other performance requirements, e.g., tail latency. In this paper we present ROSE, a new resource management platform capable of conducting performance-aware resource oversubscription. ROSE allows latency-sensitive long-running applications (LRAs) to co-exist with computation-intensive batch jobs. Instead of waiting for resource allocation to be confirmed by the centralized scheduler, job managers in ROSE can independently request to launch speculative tasks within specific machines according to their suitability for oversubscription. Node agents of those machines can however avoid any excessive resource oversubscription by means of a mechanism for admission control using multi-resource threshold control and performance-aware resource throttle. Experiments show that in case of mixed co-location of batch jobs and latency-sensitive LRAs, the CPU utilization and the disk utilization can reach56.34% and 43.49%, respectively, but the 95th percentile of read latency in YCSB workloads only increases by 5.4% against the case of executing the LRAs alone.
KW - Resource scheduling
KW - Oversubscription
KW - Cluster utilization
KW - Resource throttling
KW - QoS
U2 - 10.1109/TPDS.2020.2970013
DO - 10.1109/TPDS.2020.2970013
M3 - Journal article
VL - 31
SP - 1499
EP - 1517
JO - IEEE Transactions on Parallel and Distributed Systems
JF - IEEE Transactions on Parallel and Distributed Systems
SN - 1045-9219
IS - 7
ER -