| With the increasing demand of information processing and exchange business, the size of data centers and the single cabinet cooling power has increased rapidly. The conventional small data centers are faced with expansion and reform projects. The optimization to the air distribution system and water cooling system has been put forward for the energy-saving reconstruction for small data centers, providing a reference for small data center reconstruction and water cooling data center design.The data of air distribution in a small data center has been tested. Four kinds of simplified cabinet model has been put forward to do the three dimensional numerical simulations of the thermal environment of the data center. The error between the simulation results of simplified model 4 and the testing results is within 0.95% ~ 8.23% while these two groups of numerical value present almost the same change trend. The simulation model and the set of simulation results could be considered reliable. On this basis, three side air supply way and original roof air supply way has been simulated. According to the simulation results, the airflow distribution of side air supply form is more reasonable than the roof air supply way. The supply air temperature can be raised to reduce energy consumption. Roof return-wind outlets would get better airflow distribution than sidewall outlets while the return air path shorter. Setting air supply outlets on both sides is better than one side.Problems needed attention in water piping system design applicable to the racks using water-cooling back piece has been discussed. The difference of hydraulic balance between direct return system(DRS) and reverse return system(RRS) in the common cabinet load changing cases has been discussed.When some of cabinets in one user branch were shut off, only this user branch would have a change of hydraulic balance. With more racks been shut off, this branch would have bigger hydraulic imbalance index and too much cooling capacity. With the same load circumstance, the hydraulic imbalance index in a RRS would be smaller than it in a DRS. When all of cabinets in one user branch were shut off, all the user branches in data center would have a change of hydraulic balance. In a DRS, when the shut off one is the worst loop, there would be smaller increasing of volumes in other branches. The hydraulic imbalance index would be about 110%, meeting the design requirements. When the shut off one is not the worst loop, the hydraulic imbalance index of this branch would be about 80% which may cause equipment overheating. The hydraulic imbalance index of other branches would be around 110%. In a RRS, it has litter impact where the shut off branch is. The hydraulic imbalance index of each branch would be around 110%, meeting the operation requirements. |
