JouleSort: A Balanced Energy-Efficiency BenchmarkSuzanne RivoireStanford UniversityMehul A. ShahHP LabsParthasarathyRanganathanHP LabsChristosKozyrakisStanford UniversityABSTRACTThe energy efficiency of computer systems is an importantconcern in a variety of contexts. In data centers, reducingenergy use improves op erating cost, scalability, reliability,and other factors. For mobile devices, energy consumptiondirectly affects functionality and usability. We propose andmotivate JouleSort, an external sort b enchmark, for evaluat-ing the energy efficiency of a wide range of computer systemsfrom clusters to handhelds. We list the criteria, challenges,and pitfalls from our experience in creating a fair energy-efficiency benchmark. Using a commercial sort, we demon-strate a JouleSort system that is over 3.5x as energy-efficientas last year’s estimated winner. This system is quite differ-ent from those currently used in data centers. It consists ofa commodity mobile CPU and 13 laptop drives, connectedby server-style I/O interfaces.Categories and Subject DescriptorsH.2.4 [Information Systems]: Database Management—SystemsGeneral TermsDesign, Experimentation, Measurement, PerformanceKeywordsBenchmark, Energy-Efficiency, Power, Servers, Sort1. INTRODUCTIONIn contexts ranging from large-scale data centers to mobiledevices, energy use in computer systems is an importantconcern.In data center environments, energy efficiency affects anumb er of factors. First, power and cooling costs are signifi-cant components of operational and up-front costs. Today, atypical data center with 1000 racks, consuming 10MW totalpower, costs $7M to power and $4-$8M to cool per year, withPermission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provided that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. To copy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission and/or a fee.SIGMOD’07,June12–14,2007,Beijing,China.Copyright 2007 ACM 978-1-59593-686-8/07/0006 ...$5.00.$2-$4M of up-front costs for cooling equipment [28]. Thesecosts vary depending upon the installation, but they aregrowing rapidly and have the potential eventually to outstripthe cost of hardware [2]. Second, energy use has implicationsfor density, reliability, and scalability. As data centers housemore servers and consume more energy, removing heat fromthe data center becomes increasingly difficult [27]. Sincethe reliability of servers and disks decreases with increasedtemperature, the power consumption of servers and othercomponents limits the achievable density, which in turn lim-its scalability. Third, energy use in data centers is startingto prompt environmental concerns of pollution and excessiveload placed on local utilities [28]. Energy-related concernsare severe enough that companies like Google are starting tobuild data centers close to electric plants in cold-weather cli-mates [24]. All these concerns have led to improvements incooling infrastructure and in server power consumption [28].For mobile devices, battery capacity and energy use di-rectly affect usability. Battery capacity determines how longdevices last, constrains form factors, and limits functional-ity. Since battery capacity is limited and improving slowly,device architects have concentrated on extracting greaterenergy efficiency from the underlying components, such asthe processor, the display, and the wireless subsystems inisolation [20, 29, 31].To drive energy-efficiency improvements, we need bench-marks to assess their effectiveness. Unfortunately, there hasbeen no focus on a complete benchmark, including a work-load, metric, and guidelines, to gauge the efficacy of energyoptimizations from a whole-system perspective. Some effortsare under way to establish benchmarks for energy efficiencyin data centers [33, 35] but are incomplete. Other work hasemphasized metrics such as the energy-delay pro duct or per-formance per Watt to capture energy efficiency for proces-sors [13, 21, 27] and servers [34] without fixing a workload.Moreover, while past emphasis on processor energy efficiencyhas led to improvements in overall power consumption, therehas been little focus on the I/O subsystem, which plays asignificant role in total system power for many importantworkloads and systems.In this paper, we propose JouleSort as a holistic bench-mark to drive the design of energy-efficient systems. Joule-Sort uses the same workload as the other external sort bench-marks [1, 17, 25], but its metric incorp orates total energy,which is a combination of power consumption and perfor-mance. The benchmark can be summarized as follows:• Sort a fixed number of randomly permuted 100-byterecords with 10-byte keys.365• The sort must start with input in a file on non-volatilestore and finish with output in a file on non-volatilestore.• There are three scale categories for JouleSort: 108(∼10GB), 109(∼ 100GB), and 1010(∼ 1TB) records• The winner in each category is the system with theminimum total energy use.We choose sort as the workload for the same basic rea-son that the Terabyte Sort, MinuteSort, PennySort, andPerformance-price Sort benchmarks do [16, 17, 25]: it issimple to state and balances system component use. Sortstresses all core components of a system: memory, CPU,and I/O. Sort also exercises the OS and filesystem. Sort isa portable workload; it is applicable to a variety of systemsfrom mobile devices to large server configurations. Anothernatural reason for choosing sort is that it represents sequen-tial I/O tasks in data management workloads.JouleSort is an I/O-centric benchmark that measures theenergy efficiency of systems at peak use. Like previous sortbenchmarks, one of its goals is to gauge the end-to-end ef-fectiveness of improvements in system components. To doso, JouleSort allows us to compare the energy efficienciesof a variety of disparate system configurations. Because ofthe simplicity and portability of sort, previous sort bench-marks have been technology trend bellwethers, for example,foreshadowing the transition from supercomputers to clus-ters. Similarly, an important purpose of JouleSort is to chartpast trends and gain insight into future trends in energy ef-ficiency.Beyond the benchmark definition, our main contributionsare twofold. First, we