Electricity bills of large-scale Internet services are a major component in overall operating costs . As a result, it is increasingly important to identify cost-reduction approaches, either through reducing energy consumption or by strategically avoiding periods or data centers with the highest electricity costs. The first group of approaches based on reduction of energy consumption has been explored for many years. However, approaches that exploit various types of electricity price variability have been recently proposed. There is a need for a holistic optimization framework to explore various strategies to mitigate exposure to price variation. While prior work has considered independently aspects of spatial price variability (through an intelligent request distribution for multi-site Internet services) and temporal price variability (via energy storage), ours is the first to do both. In this paper, we propose battery-oriented techniques in combination with geographically-distributed request routing, in order to fully optimize the electricity costs of multi-site Internet services exposed to both types of electricity price variations. Furthermore, we present an optimization based framework designed to analyze the cost-saving leverage afforded by the combination of these techniques. By exploring a range of design decisions, such as battery capacity and service level agreement (SLA) requirements, and operating conditions such as common power pricing policies and price variability, we identify which strategies are most effective and under what conditions. For example, across a range of SLAs, our results show that batteries offer roughly 7% additional savings in multi-site environments already implementing intelligent request distribution.