Semantic-Aware Online Workload Characterization and Consolidation
2018 IEEE 11th International Conference on Cloud Computing (CLOUD)(2018)
摘要
Analyzing behavioral patterns of workloads is critical to understanding Cloud computing environments. How-ever, until recently, allocation of resources to Virtual Machines running the workloads were static, based on user specifications. Cloud providers performed resource consolidation mostly by packing low priority, best effort workloads with regular workloads with strict QoS requirements. This paper is building on recent efforts towards dynamic, on-line resource consolidation based on workload recognition and resource usage prediction. We introduce a new methodology for online VM consolidation that is based on a combination of resource usage data and program features for accurate resource prediction of the running workloads. We show a 15% improvement in prediction accuracy versus a baseline method using resource usage alone and an average 30% saving in resources after online consolidation with around 25% less resource capacity violations using our method.
更多查看译文
关键词
Workload Characterization, Neural Network, Program Snapshots, Sampling
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络