01
background
02
What is Data Center Energy Efficiency
Data Center (DC for short) is a construction site that provides an operating environment for centralized electronic information equipment, including a host room, auxiliary area, support area, and administrative management area. It is an important base for computing power infrastructure. Among them, the electromechanical system of the data center consumes the most energy, which mainly provides stable and high-quality power supply and cooling for IT equipment.
The energy efficiency of the data center is currently measured by the core indicator PUE (Power Usage Effectiveness), which is calculated as the ratio of the total power consumption of the data center to the power consumption of IT equipment. The larger the proportion of electric energy used in IT equipment, the more energy-efficient the data center proves. The figure below shows the global PUE trend shown in the global data center research report of the authoritative organization Uptime Institute in 2021. By 2022, the general average PUE is around 1.55.
03
PUE optimization practice of OPPO's self-built and self-operated data center
Building A of OPPO Marina Bay Data Center is composed of two modules in the north and south, which are respectively connected to two 10MW power consumption capacity. According to estimates, before the optimization, by the end of 2022, the daily average PUE is about 1.9~2.0, and the overall data center storage rate is lower than 10%. The temperature of the day is 21-29℃, and the relative humidity is 60%-95%. According to calculations, the PUE is expected to be higher.
After optimization, as of August 2023, the weekly average PUE of the south module is optimized to about 1.4, the weekly average PUE of the north module is optimized to below 1.4, and the IT load rate is lower than 20%. The outdoor temperature is relatively high during this period, which is a hot time From the perspective of the whole year, it is expected that the average annual PUE will be lower.
To optimize a huge system, it is necessary to measure it in various ways to observe the operating status of the system, and at the same time it is beneficial to assist the verification results. Therefore, the first step in energy efficiency optimization is the production of energy efficiency reports.
The overall idea of HVAC professional cost reduction: from simple to deep, from single point to overall situation. Firstly, check and optimize a single independent device, and carry out simple single-point cost reduction. This can be achieved by evaluating and shutting down unnecessary devices, and adjusting the logic threshold of a single device under the premise of ensuring safety. The actual implementation measures of OPPO's own data center are as follows:
f) Under pre-cooling conditions, there is a short-circuit problem in the plate replacement. By adjusting the measuring point of the logic control and changing the switching threshold in the pre-cooling mode, this problem can be fixed. This measure is expected to save no less than XX million electricity per year.
g) The cooling water pump is automatically increased to full frequency under non-host cooling conditions. After changing the logic, the power of a single water pump is reduced from XXkW to an average of about XXkW. The two modules in the north and south save XXkW of electric power.
h) The operation strategy of using cold storage tanks for cold storage in the early stage of operation brings the following three benefits: first, it avoids the surge when the chiller is under low load; second, it avoids the long-term low-load and inefficient operation of the entire HVAC system; 3. According to the time-of-use electricity price, the unit price of electricity can be reduced to realize peak shaving and valley filling. During this period, this measure is expected to save XX million
i) In the case of low load, the cooling capacity of the water side is excessive, so the water valve of the terminal air conditioner is increased to reduce the power consumption of the fan on the wind side when the load on the water side and IT does not increase
– The air conditioner in the private room of the north module saves XXkW of power consumption, and it is estimated to save XX yuan in electricity costs every year.
– Set up the air conditioners in the access room of the north-south module operator, the weak motor room and the test machine room. There is no obvious change in temperature and humidity, and the electric power is saved by XXkW.
– Adjust the parameters of the precision air conditioner on the first floor and the second floor of the north-south module power distribution room, save XXKw of electric power, save XX million in the whole year, and save XX million in the rest of the year
–Optimize the operation strategy of the AB road power distribution room on the 3rd/4th floor of the north-south module , change from cold standby high speed to hot standby low speed, no obvious change in temperature and humidity, save electric power XX kW, and save XX yuan in electricity costs throughout the year. Save XX million yuan in electricity bills in the remaining time
▎ 3.3 Key technology: electrical
General idea of cost reduction for electric power majors:
a) Turn off the heating of the diesel generator, reduce the power consumption of the auxiliary system by XXkW, and save XX yuan in electricity costs every year
b) Complete the load balancing measures of AB road, save the basic electricity cost XX yuan per month, and it is estimated to save XX yuan in electricity costs every year
c) Optimization of UPS operation strategy, importing intelligent parallel program to UPS, reducing the number of UPS parallel machines, improving UPS load rate and efficiency, reducing loss of XXkW, and saving XX yuan in electricity costs throughout the year
d) Balance the power UPS load by switching ATS, increase the power UPS load rate, and reduce UPS loss
e) The eco mode of precision air-conditioning power UPS on the third and fourth floors reduces UPS loss by XXkW, and is expected to save XX yuan in electricity costs throughout the year and XX yuan in the rest of the year
f) All SVG active harmonic cabinets on the 34th floor are closed, the north module reduces the electric power by XX kW, and the annual cost reduction is XX yuan, and this year’s cost reduction is XX yuan; the south module reduces the electric power by XX kW, and the annual dimension is reduced The cost is XX million yuan, and the cost is reduced by XX million yuan this year
▎ 3.4 Key technologies: others
a) The lighting management strategy is implemented, the lighting power is reduced from XXkW to XXkW, and the annual electricity cost is saved by XX yuan
b) The rainwater recycling system was officially launched to reduce municipal water consumption
c)A栋大门封堵，机房内部桥架封堵，降低数据中心湿度和恒湿机用电量
04
总结
近些年经济形势紧张，各家公司对成本都异常关注，可以预见的未来将会结束跑马圈地的粗放式发展，进入精细化运营的时代，云类资源发展也不例外。 数据中心作为云的底座，能源成本占运营成本的大头。此次项目仅作为初步的尝试，后面会持续地对数据中心运行进行优化，以求达致最优的能效，这在降本增效大背景下非常有意义。
此类降本优化的项目从来不是一蹴而就的，需要相关工作人员在日常工作中发现可优化的点，每次发现一个点，优化一个点，日积月累才能实现质的突破，此类知识能力的积累也会变成公司难以被复制的技术竞争力，可以在未来的项目中加以复用和延伸。
对于新建数据中心而言，在选择技术架构时，可考虑模块化分布式的架构，IT上架时可按需开启，在一定程度上可以避免低IT负载情况下大马拉小车的情况。
但从另一个角度来看，系统的可用性与经济性从来都是呈现负相关的，在能效优化的过程中需要平衡可用性与经济性的关系，不能因为节省一些电费导致业务中断，因小失大。
主要从事IDC基础设施技术运营工作
本文分享自微信公众号 - 安第斯智能云（OPPO_tech）。
如有侵权，请联系 [email protected] 删除。
本文参与“OSC源创计划”，欢迎正在阅读的你也加入，一起分享。