Minimizing data center SLA violations and power consumption via hybrid resource provisioning

Abstract

This paper presents a novel approach to correctly allocate resources in data centers, such that SLA violations and energy consumption are minimized. Our approach first analyzes historical workload traces to identify long-term patterns that establish a "base" workload. It then employs two techniques to dynamically allocate capacity: predictive provisioning handles the estimated base workload at coarse time scales (e.g., hours or days) and reactive provisioning handles any excess workload at finer time scales (e.g., minutes). The combination of predictive and reactive provisioning achieves a significant improvement in meeting SLAs, conserving energy, and reducing provisioning costs. We implement and evaluate our approach using traces from four production systems. The results show that our approach can provide up to 35% savings in power consumption and reduce SLA violations by as much as 21% compared to existing techniques, while avoiding frequent power cycling of servers.

A control theoretical view of cloud elasticity: taxonomy, survey and challenges

“…The existing research works lack on providing an explanation on how a proposed method deals with such undesirable oscillatory behaviour. (vi) Resource usage analysis over-provisioning is used to avoid performance violation considering peak workload scenarios [110,111]. However, this results in the wastage of resources.…”

Service level agreements-driven management of distributed applications in cloud computing environments

“…Gandhi et al [60] propose a hybrid solution for predicting the data center resource demands using historical monitoring data. They are pro-actively predicting the load, for handling the periodic changes that characterize most cloud utilization patterns.…”

SOFTScale: Stealing Opportunistically for Transient Scaling

“…These works can be classified into reactive [7][8][9], predictive [10,11] and mixed [3,12,13] approaches. While these approaches can handle gradual changes in load, they cannot handle abrupt changes, especially load spikes that occur almost instantaneously.…”

“…The data tier is stateful, and is almost never turned off [4,5], even if there is a significant drop in load [6]. The application tier, on the other hand, is usually stateless and can be dynamically scaled using existing reactive [7][8][9], predictive [10,11] or mixed [3,12,13] approaches, provided that the load does not change too abruptly.…”