A Classification-Based Approach to Policy Refinement

Udupi, Yathiraj B. ; Sahai, Akhil ; Singhal, Sharad

HP Development Company

关键词: classification;    policy refinement;    decomposition;    SLA;   

RP-ID  :  HPL-2007-6

学科分类：计算机科学（综合）

美国|英语

来源: HP Labs

【 摘 要 】

Systems are typically designed based on certain high- level goals, such as performance and availability. On the other hand, during operation, usually only low level metrics (e.g., CPU utilization) are measured, and system administrators or experts use domain knowledge to implicitly map bounds on these lower level metrics such that the high-level performance goals are met. The objective of this research is to create an automated and domain independent approach to derive policy bounds on the low level metrics such that the high level goals are met. These low-level policies may be also be used for monitoring the system for goal assessment purposes. The refinement is carried out using a combination of data classification and test & development approaches. A system is deployed within a test-and-development environment and a data set containing values of low-level metrics is collected by placing appropriate workloads on the system. The policy bounds are derived by applying classification techniques on this dataset. The classification rules are further refined using statistical distributions to arrive at certain low level rules that are useful for system monitoring and to check the system health when it is deployed and running. We show the validity of our approach for an e-commerce auctioning system (RubiS). Publication Info: A shorter version of this paper will be published in IM2007, 21-25 May 2007, Munich, Germany 7 Pages