天井吹出し天井吸込み方式データセンターのための実大模型実験と空調条件の影響の把握

摘要

Demand for data centers has increased greatly with prevailing Internet and IT systems. Enormous heat generation of information handling requires reliable and high efficiency air-conditioning system. Recent high density servers are required to conserve more energy and have less running const than conventional ones.rnAccording to a previous study by Suwa et al., an air-conditioning system with both supplies and returns on the ceiling is the most effective air flow system; however, this study was done via computational fluid dynamics simulation and a shrink model experiment.rnIn addition, Hayama et al. have proposed some important indexes for determining cooling characteristics of air flow systems; however, the indexes are difficult to utilize to control an air-conditioning system.rnIn this study, the authors have performed actual model experiment for an air-conditioning system with supplies and returns on the ceiling. To determine cooling characteristics, the indexes were applied, and the authors compared the indexes of the actual model with those in the previous resrnearch. Furthermore, the authors applied RCI_(HI) to evaluate temperature distribution of the rack inlet in order to accumulate fundamental data for air-conditioning control of a data center.rnAs a result, the following cooling characteristics of the air -conditioning system with supplies and returns on the ceiling were observed.rn1. Increases in supply flow decrease the normalized rack inlet temperature, but increase the exhaust efficiency.rn2. The supply temperature does not affect the cooling characteristics nor the exhaust efficiency.rn3. A hanged wall decreases the normalized rack inlet temperature on the center racks, but increases it on the edge racks. Further, it increases the exhaust efficiency on all the racks because it facilitates separation of air between hot aisles and cold aisles.rnThe authors observed the following features by using RCI_(HI).rn1. All cases of all amounts of supply air under 20℃ without hanged wall result in over 0.93 RCI_(HI).rn2. If the supply air temperature is over 22℃ without hanged wall, RCI_(HI) increases with an increase in the amount of supply air.rn3. All cases of all amounts of supply air under 20℃ result in an increase in RCI_(HI) with an increase in the hanged wall ratio.rn4. Under the condition of 26℃ supply air and a hanged wall ratio of 0.8, the rack inlet temperature succeeds max allowable temperature. This is because the hanged wall prevents temperature increase on the center racks, but facilitates temperature increase on the edge racks.%コールドアイル?ホットアイル方式のサーバラック配置を有するデータセンターに、天井吹出し天井吸込み方式の空調気流方式を採用した場合を想定した実大模型実験を実施し、各種制御変数が冷却性能に与える影響を検討した。消費電力に影響を与える空調機の制御変数として空調給気温度および給気風量を、室内気流状態の制御変数としてサーバラック上部への垂壁設置の有無および設置した場合の垂壁長さを様々に変化させて、室内各部の温度を測定した。また、これらの制御変数がサーバの冷却特性に与える影響を把握し、省エネルギー制御のための基礎データとしてまとめた。さらに、実験結果をラック給気面の冷却状態を評価する無次元評価指標RCI_(Hi)にて評価し、制御変数が与える影響をまとめた。垂壁が長くなると、RCI_(Hi)が許容値の上限を超える可能性があることを示した。