In practice, econo-mizer implementation has been hindered by IT equipment reli-ability concerns associated with exposing IT equipment to largevolumes of outside air [6]. Our earlier research showed that econ-omizer use can increase the indoor particle concentration in a datacenter by an order of magnitude [16], but that improved filtrationcan eliminate this increase in particle concentration with minimaladditional energy use [17]. However, even with particle filtration,reliability concerns could remain since indoor levels of otherpollutants (e.g., gaseous contaminants) that originate outdoorsmight still be higher during economizer use than for the baselinedesign. The Economizer results reported here provide an estimate ofthe energy saving benefits associated with accepting or addressingthis potential reliability risk.Permitting a wider range of temperature and RH operatingconditions in data centers with economizers could expand poten-tial energy savings, since economizers could then be operated formore hours of the year. However, expanding the operatingtemperature and RH limits in data centers might adversely impactIT equipment reliability. At present, the extent of potentialdisruption to typical data center operations is not well understood[1]. The Economizer Plus simulations were undertaken to betterunderstand if energy savings from operating conditions mightwarrant exploring strategies that would maintain appropriate ITequipment reliability while allowing the equipment to be exposedto a broader range of indoor temperatures and humidities. Incalculating the Economizer Plus results, models were executed fora humidity range of 1e100% (that is, no restriction) and for supply/return temperatures of 17.8 C (64 F)/28.9 C (84 F), whichrepresent setpoints recommended to improve data center energyefficiency [18].Size-specific building energy models were developed to accountfor the different IDC-defined space types summarized inTable 1. Eachmodel was run using annual hourly climate data for each represen-tative city. The sizes and efficiencies of mechanical system compo-nents assumed for each data center space type were based ona combination of manufacturer design guidelines, fundamentalHVAC sizing equations, and observations gained through profes-sional experience that have been documented and applied inprevious analyses [12,13].
The modeling parameters are describedbelow. Details concerning the size and efficiency assumptions for themodeledmechanical equipment are available in the SupplementaryMaterial section.2.1. Server closetsServer closets are assumed to be impractical candidates foreconomizer use. Such spaces are defined as only housing 1e2servers and they are thermally conditioned solely through thecentral HVAC system of a building without any dedicated coolingequipment [9]. Furthermore, server closets may be located withinthe interior portion of a building, making the ductwork needed fora dedicated economizer to the closet prohibitively costly for therelatively small internal heat load. While server closets can beconsiderably different than other data center types, in the absenceof measured server closet PUE data, the current average PUE of 2.0for all data centers (based on industry consensus rather than anydetailed analysis) was used to characterize current server closetperformance. This PUE of 2.0 was used to represent the non-ITequipment performance in server closets for the Baseline, Econo-mizer, and Economizer Plus scenarios.2.2. Server roomsA server room is modeled as a 23 m2(250 ft2) space with aninternal IT load of 0.43 kW/m2(40W/ft2). Underfloor air distribu-tion, common in larger data center facilities, is assumed not to bepresent owing to the small size of the room. Also, unlike the otherspace types modeled, no humidity restrictions are imposed for anysimulation on this small data center space, since typical practice inthis type of infrastructure design is to control only temperature [1].In the Baseline scenario, the room is cooled with a single dedicatedair-cooled direct expansion (DX) computer room air-conditioner(CRAC) unit placed within the room. This space type is assumed tohave no redundancy requirements, meaning that no additionalCRAC units are used. The CRAC unit contains a constant speed fanthat draws air through the unit directly from the roomwithout theuse of any ductwork. For the Economizer scenario, this coolingsystem is converted to a dedicated outdoor package DX air-condi-tioning unit with an on-board air-side economizer. A constantspeed fan in the outdoor unit supplies air to the room througha ceiling duct system and a second fan exhausts air from the roomduring economizer activity. The ductwork increases the pressuredrop experienced by the fans relative to the CRAC system. Sepa-rating the hot and cold airstreams by alternating aisles of supply airand return air is common practice to minimize spatial temperaturevariations in large data centers. Server rooms are assumed toexperience greater temperature variations since the limited floorspace in server rooms prevents this orientation of IT equipment.The use of only one CRAC unit in the server room also wouldcontribute to temporal temperature variability as the compressor inthe DX unit cycles on and off, leading to possible periods when thesupply air temperature reaching the IT equipment is greater thanintended. Elevated temperature setpoints were therefore assumedimpractical and the Economizer Plus scenario was not modeled forthis space type.2.3. Localized data centersThe localized data center space type is modeled as a 47 m2(500 ft2) room with an internal IT load of 0.65 kW/m2(60W/ft2).The Baseline scenario uses three air-cooled DX CRAC units withconstant speed fans placed in the data center that supply airthrough an underfloor plenum. Redundancy of Nþ 1 is assumed,signifying that the system is designed with one more CRAC unitthan is actually needed tomeet the cooling load. The CRAC units areequipped with active humidity control and include electrichumidifiers. In the Economizer and Economizer Plus scenarios, theCRAC units are replaced with two outdoor DX package air-condi-tioning units with on-board air-side economizers. Air is ductedfrom the outdoor air-handlers to an underfloor plenum usingconstant speed fans. The package units are equipped with activehumidity control, but use more efficient adiabatic humidifiers toaddress the wider range of humidity that the system must condi-tion owing to the high rate of outside air entering the data center.Adiabatic humidifiers use the heat from incoming air to vaporizewater, while conventional electric humidifiers require additionalenergy to vaporize water through electric-resistance heating. Anupper dewpoint limit for economizer operation is incorporated inthe design to reduce the amount of latent cooling required duringeconomizer hours.2.4. Mid-tier data centersA 232 m2(2500 ft2) roomwith an IT load of 0.86 kW/m2(80W/ft2) is used to representmid-tier data centers. The cooling systeminthe Baseline scenario consists of seven air-handling units, referred to as computer room air-handlers (CRAH). Situated on the datacenter floor, the CRAH units receive chilled water from two air-cooled chillers located outside of the data center. A redundancy ofNþ 1 is assumed for the CRAH units in this space type. Constantspeed fans in the CRAH units supply conditioned air through anunderfloor plenum. A typical minimum ventilation requirement of2.7 m3/h per m2of floor space [15] and humidity controls areprovided through a separatemakeup air-handler that supplies onlyoutside air and uses an electric humidifier. The separate air-handleralso receives chilled water from the air-cooled chilled water plant.The makeup air-handler is balanced with an exhaust fan of equalsize. In the Economizer and Economizer Plus scenarios, four air-handling units with on-board air-side economizers are placedoutdoors and receive chilledwater fromtwo air-cooled chillers. Theoutdoor air-handlers use variable speed fans to supply air to thedata center via ducts through an underfloor plenum and areequipped with active humidity controls that use adiabatic humid-ifiers. An exhaust fan in each of the outdoor air-handlers is acti-vated during economizer periods to balance the influx of outdoorair.2.5. Enterprise data centersThe enterprise data center space type is represented as a 465 m2(5000 ft2) data centerwith a
- 上一篇:氟伐他汀抑制昆虫保幼激素合成英文文献和中文翻译
- 下一篇:注塑模具的设计与热分析英文文献和中文翻译
-
-
-
-
-
-
-
NFC协议物理层的软件实现+文献综述
江苏省某高中学生体质现状的调查研究
上市公司股权结构对经营绩效的影响研究
巴金《激流三部曲》高觉新的悲剧命运
g-C3N4光催化剂的制备和光催化性能研究
C++最短路径算法研究和程序设计
高警觉工作人群的元情绪...
中国传统元素在游戏角色...
现代简约美式风格在室内家装中的运用
浅析中国古代宗法制度