With increasing costs of energy consumption and cooling, power management in server clusters has become an increasingly important design issue. Current clusters for real-time applications are designed to handle peak loads, where all servers are fully utilized. In practice, peak load conditions rarely happen and clusters are most of the time underutilized. This creates the opportunity for using slower frequencies, and thus smaller energy consumption, with little or no impact on the Quality of Service (QoS), for example, performance and timeliness. In this work we present a cluster-wide QoS-aware technique that dynamically reconfigures the cluster to reduce energy consumption during periods of reduced load. Moreover, we also investigate the effects of local QoS-aware power management using Dynamic Voltage Scaling (DVS). Since most real-world clusters consist of machines of different kind (in terms of both performance and energy consumption) we focus on heterogeneous clusters. For validation, we describe and evaluate an implementation of the proposed scheme using the Apache Webserver in a small realistic cluster. Our experimental results show that using our scheme it is possible to save up to 45 % of the total energy consumed by the servers, maintaining average response times within the specified deadlines and number of dropped requests within the required amount. 1

Abstract—Power consumption is an important issue in the design of real-time embedded systems. As many embedded systems are powered by batteries, the goal is to extend the autonomy of the system as much as possible. To reduce power consumption, modern processors can change their voltage and frequency at runtime. A power-aware scheduling algorithm can exploit this capability to reduce power consumption while preserving the timing constraints of real-time tasks. In this paper, we present GRUB-PA, a novel power-aware scheduling algorithm based on a resource reservation technique. In addition to providing temporal isolation and time guarantees and, unlike most of the power-aware algorithms proposed in the literature, GRUB-PA can efficiently handle systems consisting of both hard and soft, aperiodic, sporadic, and periodic tasks. We compared our algorithm with existing power-aware scheduling algorithms on an extensive set of simulation experiments on synthetic task sets. The results show that the performance of our algorithm is in line with the state-of-the-art power-aware algorithms. We also present the implementation of our algorithm in the Linux operating system and discuss practical implementation issues like switching overhead and power models. Finally, we show the results of experiments performed on a real testbed application. Index Terms—DVS, real-time, resource-reservation, scheduling, power-aware. Ç

Reducing energy consumption is one of the main concerns in the design and the implementation of embedded real-time systems. For this reason, the current generation of processors allows to vary voltage and operating frequency to balance computational speed versus energy consumption. This technique is called Dynamic Voltage Scaling (DVS). When applying DVS to hard real-time systems, it is important to provide the worst-case computational requirement, otherwise a task may miss some timing constraint. However, the probability of a task executing for its worst-case execution time is very low. In this paper, we show how to exploit probabilistic information about the execution time of a task in order to reduce the energy consumed by the processor. Optimal speed assignments and transition points are found using a very general model for the processor. The model accounts for the processor idle power and for both the time and the energy overheads due to frequency transitions. We also show how these results can be applied to some significant cases.