Marcus Hähnel

Measuring energy consumption for short code paths using RAPL

Measuring the energy consumption of software components is a major building block for generating models that allow for energy-aware scheduling, accounting and budgeting. Current measurement techniques focus on coarse-grained measurements of application or system events. However, fine grain adjustments in particular in the operating-system kernel and in application-level servers require power profiles at the level of a single software function. Until recently, this appeared to be impossible due to the lacking fine grain resolution and high costs of measurement equipment. In this paper we report on our experience in using the Running Average Power Limit (RAPL) energy sensors available in recent Intel CPUs for measuring energy consumption of short code paths. We investigate the granularity at which RAPL measurements can be performed and discuss practical obstacles that occur when performing these measurements on complex modern CPUs. Furthermore, we demonstrate how to use the RAPL infrastructure to characterize the energy costs for decoding video slices.

Has energy surpassed timeliness? Scheduling energy-constrained mixed-criticality systems

In the past, we have silently accepted that energy consumption in real-time and embedded systems is subordinate to time. That is, we have tried to reduce energy always under the constraint that all deadlines must be met. In mixed-criticality systems however, schedulers respect that some tasks are more important than others and guarantee their completion even at the expense of others. We believe in these systems the role of the energy budget has changed and it is time to ask whether energy has surpassed timeliness. Investigating energy as a further dimension of mixed-criticality systems, we show in a realistic scenario that a subordinate handling of energy can lead to violations of the mixed-criticality guarantees that can only be avoided if energy becomes an equally important resource as time.

eBond: energy saving in heterogeneous R.A.I.N

Network energy is a significant, although not the largest, cost factor in medium to large scale server installations. On the other hand, most server installations work with redundant link and infrastructure layouts to reduce the risk of network outages. Introducing eBond, an energy-aware bonding network device, we exploit possible heterogeneities in these redundant layouts to adapt network device energy consumption to dynamic server bandwidth demands. Replaying the trace of a realistic scenario in a simulation of eBond with fine grain energy profiles measured at two network cards we achieve energy savings up to 75% for the server-side network interconnect.

The case for practical multi-resource and multi-level scheduling based on Energy/Utility

Energy has become the dominating concern for resource management. We advocate an energy-centered design approach for resource-management systems. To this end, we structure systems in layers, where layers implement higher-level resources using lower-level ones. For each layer, we describe the relation of the performance delivered for the higher layer to its demands on the lower layer and refer to that relation as demand/performance function. The lowest layers are rooted in hardware and express demand in terms of energy, the highest layers provide performance in terms of user-specific utility, thus leading to an Energy/Utility characterization of a complete system. We describe the overall approach, some research challenges and few initial results on the representation of demand/performance functions.

Powernightmares: The Challenge of Efficiently Using Sleep States on Multi-core Systems

Sleep states are an important and well-understood feature of modern server and desktop CPUs that enable significant power savings during idle and partial load scenarios. Making proper decisions about how to use this feature remains a major challenge for operating systems since it requires a trade-off between potential energy-savings and performance penalties for long and short phases of inactivity, respectively. In this paper we analyze the default behavior of the Linux kernel in this regard and identify weaknesses of certain default assumptions. We derive pathological patterns that trigger these weaknesses and lead to ‘Powernightmares’ during which power-saving sleep states are used insufficiently. Our analysis of a workstation and a large supercomputer reveals that these scenarios are relevant on real-life systems in default configuration. We present a methodology to analyze these effects in detail despite their inherent nature of being hardly observable. Finally, we present a concept to mitigate these problems and reclaim lost power saving opportunities.

The Potential of Energy/Utility-Accrual Scheduling

The long term vision of energy/utility accrual scheduling is to use all system resources in a way that is most beneficial to the systems users. For that, a mapping of user requests all the way down to system resources is required and, vice versa, the energy requirements of resources must be attributed to the corresponding user requests. However, despite the attractiveness of this general approach, the complexities involved in these translations are scary. Sketching our approach to energy/utility accrual scheduling, we argue in this paper that many complexities of traditional power models can be avoided if we consider the potential of a resource to generate utility rather than the utility generating operation. Introducing modes for the resources CPU and network, we found that the energy required to keep these resources operational is a good approximation of their overall energy demand.

Modular Energy Modeling using Energy/Utility

The modeling of the relationship between power usage and performance for complex computing systems is challenging due to the vast amount of tunable parameters that influence both metrics. To simplify the energy management of information systems from individual embedded machines to whole data centers we use a modular, hierarchical concept called Energy/Utility to model individual parts of a system. We present first results that show the decomposition of an individual asymmetric multi-processing system into hardware and software models. We show that using the Energy/Utility approach these models can stay manageable reducing total benchmark running time and modeling overhead while providing sufficiently high precision for performance and energy usage prediction.

Energy-Utility Function-Based Resource Control for In-Memory Database Systems LIVE

The ever-increasing demand for scalable database systems is limited by their energy consumption, which is one of the major challenges in research today. While existing approaches mainly focused on transaction-oriented disk-based database systems, we are investigating and optimizing the energy consumption and performance of data-oriented scale-up in-memory database systems that make heavy use of the main power consumers, which are processors and main memory. In this demo, we present energy-utility functions as an approach for enabling the operating system to improve the energy efficiency of scalable in-memory database systems. Our highly interactive demo setup mainly allows attendees to switch between multiple DBMS workloads and watch in detail how the system responds by adapting the hardware configuration appropriately.

Energy-Efficiency of OWL Reasoners—Frequency Matters

While running times of ontology reasoners have been studied extensively, studies on energy-consumption of reasoning are scarce, and the energy-efficiency of ontology reasoning is not fully understood yet. Earlier empirical studies on the energy-consumption of ontology reasoners focused on reasoning on smart phones and used measurement methods prone to noise and side-effects. This paper presents an evaluation of the energy-efficiency of five state-of-the-art OWL reasoners on an ARM single-board computer that has built-in sensors to measure the energy consumption of CPUs and memory precisely. Using such a machine gives full control over installed and running software, active clusters and CPU frequencies, allowing for a more precise and detailed picture of the energy consumption of ontology reasoning. Besides evaluating the energy consumption of reasoning, our study further explores the relationship between computation power of the CPU, reasoning time, and energy consumption.

Demo abstract: An energy/utility demo - Energy-aware resource scheduling under utility considerations

Our works on energy/utility present an approach for energy-efficient resource management that considers the utility requirements of users and the energy restriction that are set by them. In this demo, we show an initial prototype of our approach, demonstrating energy/utility trade-offs in an example application and how modeling individual components and their interaction can simplify the scheduling of resources.