240 likes | 342 Views
Economical and Robust Provisioning of N-Tier Cloud Workloads: A Multi-level Control Approach. Pengcheng Xiong 1 , Zhikui Wang 2 , Simon Malkowski 1 , Qingyang Wang 1 , Deepal Jayasinghe 1 , Calton Pu 1 1 Georgia Institute of Technology 2 HP Labs Email: xiong@gatech.edu. Overview.
E N D
Economical and Robust Provisioning of N-Tier Cloud Workloads: A Multi-level Control Approach Pengcheng Xiong1, Zhikui Wang2, Simon Malkowski1, Qingyang Wang1, Deepal Jayasinghe1, Calton Pu1 1Georgia Institute of Technology 2HP Labs Email: xiong@gatech.edu
Overview • Motivation • Background • Resource partition controller • Application controller • Conclusions
Overview • Motivation • Background • Resource partition controller • Application controller • Conclusions
Different feedback controller design for a single/multi-tiered application (1) Zhu et al, ACC 2006
Different feedback controller design for a single/multi-tiered application (2) TFF TUC TFB Wang et al, FeBID2007
Different controllability under different workload generator (1) Schroeder et al, NSDI 2006
Different controllability under different workload generator (2) Xionget al, NOMS 2010
Goals • Economical • We want to meet the performance requirement for the N-tier web application with the minimum total resources. • Robust • We want to be robust to different time-varying workload types, e.g., open, closed, semi-open.
Overview • Motivation • Background • Resource partition controller • Application controller • Conclusions
Test bed • Experiment Environment • Apache, Tomcat, Mysql • Xen hypervisor • Workload Generator • RUBiS “Browsing mix” workload that has 10 transaction types, e.g., Home, Browse, ViewItem. (just like eBay) • Workload types (open, closed, semi-open) • Workload intensity
Overview • Motivation • Background • Resource partition controller • Application controller • Conclusions
Optimal resource partition • Solution 1(Shares) • Solution 2(Util.) • Our solution(Opt.)
Overview • Motivation • Background • Resource partition controller • Application controller • Conclusions
Application controller design • System model between the RTT and S • System identification method based on ARMA model • Controller design • Root-locus method based on control theory
Controller design • ARX01 model • Proportional-integral (PI) controller • The closed model transfer function
Performance controller(setting=35ms) Util has MORE fluctuation than Opt.
Conclusions • We propose economical and robust provisioning for Cloud resources for N-tier web applications through a multi-level control approach. • Experimental results show that our solution outperforms other existing approaches • Almost the same performance but save up to 20% CPU resources. • Robust to deal with different workload styles.