The Community for Technology Leaders
2015 International Conference on Cloud Computing and Big Data (CCBD) (2015)
Shanghai, China
Nov. 4, 2015 to Nov. 6, 2015
ISBN: 978-1-4673-8350-9
pp: 3-10
It is desired but challenging to support users customizing their own SLO policy to manage the storage resource allocation among their VM groups. This relies on stable and accurate SLO enforcement characterized with fast convergence and precisely differentiating service levels. However, highly variable IO characteristics, IO interference among VMs and unstable IO capacity of storage devices can compromise the SLO enforcement by increasing performance error, degrading the SLO compliance and exacerbating the performance fluctuation. To address this challenge, we propose SASLO, an end-to-end VM-oriented control framework that supports stable and accurate SLO enforcement by means of proportional-integral IO control adopted from the classic control theory for each VM. Our extensive evaluation driven by representative benchmarks demonstrates that SASLO is able to achieve near-zero deviation on throughput sharing with lower variability than existing mainstream techniques. Based on the easy-to-use programmable interface provided by SASLO, user-customized SLO policy can be designed and executed in the form of a time-varying SLO series.
Throughput, Cloud computing, Servers, Convergence, Performance evaluation, Schedules, Companies

N. Li, H. Jiang, D. Feng and Z. Shi, "SASLO: Support User-Customized SLO Policy via Programmable End-to-End VM-Oriented IO Control," 2015 International Conference on Cloud Computing and Big Data (CCBD), Shanghai, China, 2015, pp. 3-10.
197 ms
(Ver 3.3 (11022016))