Energy consumption in hosting Internet services is becoming a pressing issue as these services scale up. Dynamic server provisioning techniques are effective in turning off unnecessary servers to save energy. Such techniques, mostly studied for request-response services, face challenges in the context of connection servers that host a large number of long-lived TCP connections. In this paper, we characterize unique properties, performance, and power models of connection servers, based on a real data trace collected from the deployed Windows Live Messenger. Using the models, we design server provisioning and load dispatching algorithms and study subtle interactions between them. We show that our algorithms can save a significant amount of energy without sacrificing user experiences. 1

Energy expenses are becoming an increasingly important fraction of data center operating costs. At the same time, the energy expense per unit of computation can vary significantly between two different locations. In this paper, we characterize the variation due to fluctuating electricity prices and argue that existing distributed systems should be able to exploit this variation for significant economic gains. Electricity prices exhibit both temporal and geographic variation, due to regional demand differences, transmission inefficiencies, and generation diversity. Starting with historical electricity prices, for twenty nine locations in the US, and network traffic data collected on Akamai’s CDN, we use simulation to quantify the possible economic gains for a realistic workload. Our results imply that existing systems may be able to save millions of dollars a year in electricity costs, by being cognizant of locational computation cost differences. Categories andSubject Descriptors

We present the design and evaluation of two forms of power management schemes that reduce the energy consumption of networks. The first is based on putting network components to sleep during idle times, reducing energy consumed in the absence of packets. The second is based on adapting the rate of network operation to the offered workload, reducing the energy consumed when actively processing packets. For real-world traffic workloads and topologies and using power constants drawn from existing network equipment, we show that even simple schemes for sleeping or rate-adaptation can offer substantial savings. For instance, our practical algorithms stand to halve energy consumption for lightly utilized networks (10-20%). We show that these savings approach the maximum achievable by any algorithms using the same power management primitives. Moreover this energy can be saved without noticeably increasing loss and with a small and controlled increase in latency (<10ms). Finally, we show that both sleeping and rate adaptation are valuable depending (primarily) on the power profile of network equipment and the utilization of the network itself.

This paper seeks to understand and design nextgeneration servers for emerging “warehouse-computing” environments. We make two key contributions. First, we put together a detailed evaluation infrastructure including a new benchmark suite for warehouse-computing workloads, and detailed performance, cost, and power models, to quantitatively characterize bottlenecks. Second, we study a new solution that incorporates volume nonserver-class components in novel packaging solutions, with memory sharing and flash-based disk caching. Our results show that this approach has promise, with a 2X improvement on average in performance-perdollar for our benchmark suite.

As our society becomes more information-driven, we have begun to amass data at an astounding and accelerating rate. At the same time, power concerns have made it difficult to bring the necessary processing power to bear on querying, processing, and understanding this data. We describe Gordon, a system architecture for data-centric applications that combines low-power processors, flash memory, and datacentric programming systems to improve performance for data-centric applications while reducing power consumption. The paper presents an exhaustive analysis of the design space of Gordon systems, focusing on the trade-offs between power, energy, and performance that Gordon must make. It analyzes the impact of flash-storage and the Gordon architecture on the performance and power efficiency of data-centric applications. It also describes a novel flash translation layer tailored to data-intensive workloads and large flash storage arrays. Our data show that, using technologies available in the near future, Gordon systems can out-perform disk-based clusters by 1.5 × and deliver up to 2.5 × more performance per watt.

As Internet services become ubiquitous, the selection and management of diverse server platforms now affects the bottom line of almost every firm in every industry. Ideally, such cross-platform management would yield high performance at low cost, but in practice, the performance consequences of such decisions are often hard to predict. In this paper, we present an approach to guide cross-platform management for real-world Internet services. Our approach is driven by a novel performance model that predicts application-level performance across changes in platform parameters, such as processor cache sizes, processor speeds, etc., and can be calibrated with data commonly available in today’s production environments. Our model is structured as a composition of several empirically observed, parsimonious sub-models. These sub-models have few free parameters and can be calibrated with lightweight passive observations on a current production platform. We demonstrate the usefulness of our cross-platform model in two management problems. First, our model provides accurate performance predictions when selecting the next generation of processors to enter a server farm. Second, our model can guide platform-aware load balancing across heterogeneous server farms. 1

Motivated by increased concern over energy consumption in modern data centers, we propose a new, distributed computing platform called Nano Data Centers (NaDa). NaDa uses ISP-controlled home gateways to provide computing and storage services and adopts a managed peer-to-peer model to form a distributed data center infrastructure. To evaluate the potential for energy savings in NaDa platform we pick Video-on-Demand (VoD) services. We develop an energy consumption model for VoD in traditional and in NaDa data centers and evaluate this model using a large set of empirical VoD access data. We find that even under the most pessimistic scenarios, NaDa saves at least 20 % to 30 % of the energy compared to traditional data centers. These savings stem from energypreserving properties inherent to NaDa such as the reuse of already committed baseline power on underutilized gateways, the avoidance of cooling costs, and the reduction of network energy consumption as a result of demand and service co-localization in NaDa.

Abstract-- We present a technique that controls the peak power consumption of a high-density server by implementing a feedback controller that uses precise, system-level power measurement to periodically select the highest performance state while keeping the system within a fixed power constraint. A control theoretic methodology is applied to systematically design this control loop with analytic assurances of system stability and controller performance, despite unpredictable workloads and running environments. In a real server we are able to control power over a 1 second period to within 1 W and over an 8 second period to within 0.1 W. Conventional servers respond to power supply constraint situations by using simple open-loop policies to set a safe performance level in order to limit peak power consumption. We show that closed-loop control can provide higher performance under these conditions and implement this technique on an IBM BladeCenter HS20 server. Experimental results demonstrate that closed-loop control provides up to 82 % higher application performance compared to open-loop control and up to 17 % higher performance compared to a widely used ad-hoc technique. 1.

This paper presents an energy management policy for reconfigurable clusters running a multi-tier application, exploiting DVS together with multiple sleep states. We develop a theoretical analysis of the corresponding power optimization problem and design an algorithm around the solution. Moreover, we rigorously investigate selection of the optimal number of spare servers for each power state, a problem that has only been approached in an ad-hoc manner in current policies. To validate our results and policies, we implement them on an actual multi-tier server cluster where nodes support all power management techniques considered. Experimental results using realistic dynamic workloads based on the TPC-W benchmark show that exploiting multiple sleep states results in significant additional cluster-wide energy savings up to 23 % with little or no performance degradation.

Abstract: Current capacity planning practices based on heavy over-provisioning of power infrastructure hurt (i) the operational costs of data centers as well as (ii) the computational work they can support. We explore a combination of statistical multiplexing techniques to improve the utilization of the power hierarchy within a data center. At the highest level of the power hierarchy, we employ controlled underprovisioning and over-booking of power needs of hosted workloads. At the lower levels, we introduce the novel notion of soft fuses to flexibly distribute provisioned power among hosted workloads based on their needs. Our techniques are built upon a measurement-driven profiling and prediction framework to characterize key statistical properties of the power needs of hosted workloads and their aggregates. We characterize the gains in terms of the amount of computational work (CPU cycles) per provisioned unit of power – Computation per Provisioned Watt (CPW). Our technique is able to double the CPW offered by a Power Distribution Unit (PDU) running the e-commerce benchmark TPC-W compared to conventional provisioning practices. Over-booking the PDU by 10 % based on tails of power profiles yields a further improvement of 20%. Reactive techniques implemented on our Xen VMM-based servers dynamically modulate CPU DVFS states to ensure power draw below the limits imposed by soft fuses. Finally, information captured in our profiles also provide ways of controlling application performance degradation despite overbooking. The 95 th percentile of TPC-W session response time only grew from 1.59 sec to 1.78 sec—a degradation of 12%.