Automated Control of Multiple Virtualized Resources
Padala, Pradeep; Hou, Kai-Yuan; Shin, Kang G.; Zhu, Xiaoyun; Uysal, Mustafa; Wang, Zhikui; Singhal, Sharad; Merchant, Arif
Keyword(s): virtualization, consolidation, service level objective, control, service differentiation
Abstract: Virtualized data centers enable consolidation of multiple applications and sharing of multiple resources among these applications. However, current virtualization technologies are inadequate in achieving complex service level objectives (SLOs) for enterprise applications with time-varying demands for multiple resources. In this paper, we present AutoControl, a resource allocation system that automatically adapts to dynamic workload changes in a shared virtualized infrastructure to achieve application SLOs. AutoControl is a combination of an online model estimator and a novel multi-input, multi- output (MIMO) resource controller. The model estimator captures the complex relationship between application performance and resource allocations, while the MIMO controller allocates the right amount of resources to ensure application SLOs. Our experimental results using RUBiS and TPC-W benchmarks along with production-trace-driven workloads indicate that AutoControl can detect and adapt to CPU and disk I/O bottlenecks that occur over time and across multiple nodes and allocate multiple virtualized resources accordingly to achieve application SLOs. It can also provide service differentiation according to the priorities of individual applications during resource contention.
External Posting Date: November 21, 2008 [Fulltext]. Approved for External Publication
Internal Posting Date: November 21, 2008 [Fulltext]