We intend to show that our currentcy model can be used to formulate policies to address the above goals. To evaluate our policies, we use several applications (described in Table 1) to create typical workload scenarios for a battery-constrained laptop user. We envision situations in which the user may want to have multiple tasks running concurrently (e.g., doing background jpeg encoding of a set of stored images while viewing the already encoded jpegs in slide show mode or listening to an MP3 while running through the slides of a PowerPoint presentation). For each experiment we use different combinations of these applications to emphasize specific aspects of the policy space. Each application presents a different set of demands for CPU, network bandwidth, disk I/O, or interactive ``think time''.
Within ECOSystem, we monitor the currentcy available for allocation each epoch and the currentcy consumed by each application during each epoch. It is also possible to track consumption by device. We then present our results in terms of average power (mW) derived from the amount of currentcy consumed or allocated per epoch. We also present appropriate application-specific performance metrics.
In the following five sections (Section 4 through Section 8), we illustrate the construction of currentcy-based policies to address each of the reformulated energy goals. Our goal in this paper is not to provide an optimal policy, but to show that policies formulated within the unified currentcy model offer desirable properties compared to more traditional (per-device) policies.