阅读说明：本技术 对数据中心中的冷却进行控制的方法 (Method of controlling cooling in a data center ) 是由 温斯顿·加西亚-加滨 卡捷琳娜·米申科 于 2019-08-13 设计创作，主要内容包括：对数据中心中的服务器的冷却进行控制的方法,数据中心包括多个服务器机架,每个服务器机架包括多个服务器,冷却由过道冷却单元经由冷过道、以及由服务器风扇提供,其中方法包括：从数据中心中的温度传感器获得(Si)温度测量结果；基于温度测量结果、利用对冷过道温度和对过道冷却单元的气流水平的约束,执行过道冷却单元和服务器的总电力消耗的优化(S2)；以及基于优化来控制(S3)过道冷却单元。(A method of controlling cooling of servers in a data center, the data center including a plurality of server racks, each server rack including a plurality of servers, the cooling being provided by aisle cooling units via cold aisles, and by server fans, wherein the method comprises: obtaining (Si) temperature measurements from temperature sensors in the data center; performing an optimization of the total power consumption of the aisle cooling units and the servers based on the temperature measurements with constraints on cold aisle temperatures and on air flow levels of the aisle cooling units (S2); and controlling (S3) the aisle cooling unit based on the optimization.)

1. A method of controlling cooling of servers (6) in a data center (1) comprising a plurality of server racks (7a-7d), each server rack comprising a plurality of servers (6), the cooling being provided by aisle cooling units (9) via cold aisles, and by server fans (11), wherein the method comprises:

obtaining (S1) temperature measurements from a temperature sensor (15) in the data center (1);

performing an optimization (S2) of a total power consumption of the aisle cooling unit (9) and the servers (6) based on the temperature measurements with constraints on cold aisle temperatures and on air flow levels of the aisle cooling unit (9); and

controlling (S3) the aisle cooling unit (9) based on the optimization.

2. The method of claim 1, wherein the optimization is a multi-objective optimization including only two objectives, a first objective of the two objectives being power consumption of the aisle cooling unit and a second objective of the two objectives being power consumption of the server.

3. The method of claim 2, comprising: prior to the step of optimizing (S3), an initial single-objective optimization is performed on each of the first and second objectives separately to obtain a minimum value for each of the first and second objectives.

4. The method of claim 2 or 3, wherein said optimizing results in a set of pareto optimal solutions, wherein said controlling (S3) is based on one of said pareto optimal solutions.

5. The method of claim 4, comprising: selecting a pareto optimal solution from the set of pareto optimal solutions based on a predetermined selection rule.

6. The method according to claim 5, wherein the selection rule is to use the pareto optimal solution that provides the smallest power consumption of the aisle cooling unit (9).

7. The method of claim 5, wherein the selection rule is to use the pareto optimal solution that provides the minimum power consumption of the server.

8. The method of any of the preceding claims, wherein the constraint on the cold aisle temperature is a range of cold aisle temperatures between a minimum allowable contract temperature and a maximum allowable contract temperature.

9. The method as claimed in any one of the preceding claims, wherein the constraint on the air flow level is a range of fan speeds of an aisle cooling unit fan, the aisle cooling unit (9) being designed to operate in the range of fan speeds.

10. The method according to any of the preceding claims, wherein the total power consumption of the server (6) is the sum of the power consumption of the server fans (11) and the power consumption due to IT loads.

11. The method according to any of the preceding claims, wherein the aisle cooling unit (9) is a Computer Room Air Handler (CRAH) unit.

12. A computer program comprising computer code which, when executed by processing circuitry (13b) of a data centre control system (13), causes the data centre control system (13) to perform the steps of the method according to any one of claims 1 to 11.

13. A computer program product comprising a storage medium (13a) comprising a computer program according to claim 12.

14. A data center control system (13) for controlling cooling of servers (6) in a data center (1) comprising a plurality of server racks (7a-7d), each server rack comprising a plurality of servers (6), the cooling being provided by aisle cooling units (9) via cold aisles, and by server fans, wherein the data center control system (13) comprises:

a storage medium (13a) comprising computer code; and

processing circuitry (13b) which, when executing the computer code, causes the data centre control system (13) to perform the steps of the method according to any one of claims 1 to 11.

Technical Field

The present disclosure relates generally to cooling control in a data center.

Background

A data center is a facility for housing servers. The data center has multiple rows of racks, where each rack includes multiple servers. The server generates a large amount of energy that causes the room in which the server is located to heat up. Since the electronics of the servers are sensitive to high temperatures, the data center is provided with a cooling system for cooling the servers. The cooling system includes a cooling unit that generates cold air or cold liquid and a distribution network for distributing the cold air/liquid to racks that include servers. The cool air is typically supplied via openings in the floor in an aisle that houses a row of server racks. This aisle is called the cold aisle. Cold air flows from the cold aisle through the server racks to the next aisle on the other side of the server racks of the cold aisle. This next aisle is called a hot aisle.

Cooling of large data centers requires a significant amount of electrical power. Large data centers operating at the industrial level may use as much electricity as small towns. Almost half of the total power is consumed by the cooling system, which is indispensable for the safe operation of the servers.

The temperature of the cold aisle in a typical data center should be maintained within a given range defined by the contract of the operator of the data center. During operation, the desired temperature in the cold aisle is usually set manually by the operator after observing the temperature measured in the cold aisle by the sensor network. IT is therefore the objective of the operator to avoid hot spots and excessive cooling in the data centre by manually varying the cooling power supplied by the cooling equipment to the data centre room in accordance with dynamic changes in the IT load.

The power consumption in a data center depends on the power consumption of multiple types of cooling equipment and servers due to IT loads. For example, aisle cooling units such as Computer Room Air Handlers (CRAH) will consume more power when they have to produce higher cooling power. This means that cooling units operating at low temperature settings and high airflow settings will consume more power. In addition, servers typically have small server fans for cooling heat sources internal to the server, such as a Central Processing Unit (CPU) or memory. Server fans are typically very inefficient. The efficiency is typically around 25%. A midrange data center has hundreds of server racks, with tens of servers per server rack and multiple server fans per server. Thus, power losses in the server fans constitute a significant portion of the power consumption of the data center.

US patent application US2017/187592a1 discloses a system for determining a cooling set point that meets dynamic requirements. To reduce overall cooling costs, this document proposes that it is important to ensure coordinated optimization across multiple cooling hardware. Approaches that minimize the energy used for a single cooling device may result in an overall increase in energy costs.

Disclosure of Invention

However, the approach disclosed in US2017/187592a1 does not provide an optimal solution for the trade-off between performing efficient cooling and minimizing power consumption. In particular, according to US2017/187592a1, "optimization" is performed by: the cooling set point is first determined, then used to predict the CPU temperature and compared to a critical threshold. If the predicted temperature is above the critical temperature, the cooling set point is adjusted and the process is repeated until the predicted temperature is below the critical temperature. In this case, the accuracy of the predicted CPU temperature is determined, and the "optimal" set point is ultimately applied to the cooling hardware.

Based on the above, it is an object of the present disclosure to provide a method of controlling cooling in a data center, which solves or at least alleviates the problems of the prior art.

Thus, according to a first method of the present invention, there is provided a method of controlling cooling of servers in a data center, the data center comprising a plurality of server racks, each server rack comprising a plurality of servers, the cooling being provided by aisle cooling units via cold aisles, and by server fans, wherein the method comprises: obtaining temperature measurements from temperature sensors in the data center; performing an optimization of the total power consumption of the aisle cooling unit and the servers based on the temperature measurements with constraints on the cold aisle temperature and on the air flow level of the aisle cooling unit; and controlling the aisle cooling unit based on the optimization.

The overall power consumption of the data center can thus be reduced while sufficient cooling of the servers can be maintained. The reduction of the total power consumption is advantageous from the viewpoint of sustainability. It also reduces the cooling costs. Additionally, by avoiding server overheating, the useful life of data center equipment may be extended, and potentially increase maintenance intervals.

The constraint on the airflow level of the aisle cooling unit is the interval of airflow in which the aisle cooling unit is designed to operate.

According to one embodiment, the optimization is a multi-objective optimization comprising only two objectives, a first objective of the two objectives being the power consumption of the aisle cooling units and a second objective being the power consumption of the servers.

Thus, a balance may be achieved between the two opposing characteristics (i.e. optimal power consumption of the aisle cooling unit and optimal power consumption of the servers), thereby minimizing the overall cooling power consumption.

One embodiment includes: prior to the step of optimizing, performing an initial single-objective optimization on each of the first and second objectives separately to obtain a minimum value for each of the first and second objectives.

According to one embodiment, the optimization results in a set of pareto optimal solutions, wherein control is based on one of the pareto optimal solutions.

One embodiment includes: a pareto optimal solution is selected from a set of pareto optimal solutions based on a predetermined selection rule.

According to one embodiment, the selection rule is to use the pareto optimal solution that provides the least power consumption of the aisle cooling units.

According to one embodiment, the selection rule is to use the pareto optimal solution that provides the minimum power consumption of the server.

According to one embodiment, the constraint on the cold aisle temperature is a range of cold aisle temperatures between a minimum allowable contract temperature and a maximum allowable contract temperature.

According to one embodiment, the constraint on the airflow level is a range of fan speeds of the aisle cooling unit fans in which the aisle cooling unit is designed to operate.

According to one embodiment, the total power consumption of the server is the sum of the power consumption of the server fans and the power consumption caused by the IT loads. The power consumption due to IT load is the power consumed by the operating server.

Preferably, the power consumption due to IT loads is only used as a disturbance input parameter when performing the optimization and is not a control variable.

According to one embodiment, the aisle cooling unit is a computer room air handler CRAH unit.

According to a second aspect of the present disclosure, there is provided a computer program comprising computer code which, when executed by processing circuitry of a data centre control system, causes the data centre control system to perform the steps of the method according to the first aspect.

According to a third aspect of the present disclosure, there is provided a computer program product comprising a storage medium comprising a computer program according to the second aspect.

According to a fourth aspect of the present disclosure, there is provided a data center control system for controlling cooling of servers in a data center, the data center comprising a plurality of server racks, each server rack comprising a plurality of servers, the cooling being provided by aisle cooling units via cold aisles, and by server fans, wherein the data center control system comprises: a storage medium comprising computer code; and processing circuitry which, when executing the computer code, causes the data centre control system to perform the steps of the method according to the first aspect.

In general, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field to which they pertain, unless explicitly defined otherwise herein. All references to "a/an/the element, device, component, means, etc" are to be interpreted openly as referring to "at least one instance of the element, device, component, means, etc., unless explicitly stated otherwise.

Drawings

Specific embodiments of the inventive concept will now be described, by way of example, with reference to the accompanying drawings, in which:

FIG. 1 schematically illustrates a top view of an example of a data center including multiple rows of server racks and cooling equipment; and

FIG. 2 illustrates a flow chart of a method of controlling cooling of servers in a data center.

Detailed Description

The concepts of the present invention will now be described more fully hereinafter with reference to the accompanying drawings, in which exemplary embodiments are shown. The inventive concept may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided by way of example so that this disclosure will be thorough and complete, and will fully convey the scope of the inventive concept to those skilled in the art. Like numbers refer to like elements throughout the specification.

Fig. 1 shows an example of a data center 1. The data center 1 has a data center room 3. The data centre room 3 may comprise a raised floor provided with holes. The data center room 3 includes a plurality of rows 5a-5d of server racks 7a-7 d. Each server rack 7a-7d includes a plurality of servers 6. The server racks 7a-7d are arranged on an elevated floor.

The data center 1 further includes a cooling device configured to cool the server 6. In the present example, the cooling device comprises an aisle cooling unit 9 and a server fan 11. The aisle cooling unit 9 may be, for example, a CRAH. The CRAH has two controllable parameters, namely the temperature of the cooling coils, and the speed of the CRAH fan, or the speed of the aisle cooling unit fan that blows air over the cooling coils. Each server fan 11 is arranged to cool a specific server 6. In particular, the server fan 11 may be configured to cool one or more heat generating electronic components (e.g., a CPU), one or more memories, or a power supply of the server. A plurality of server fans 11 may be provided to cool each server 6.

The aisle cooling unit 9 has a cooling outlet arranged to provide cold air under the raised floor by means of an aisle cooling unit fan. Below the elevated floor, there may be an airflow distribution system configured to distribute a flow of cold air from the aisle cooling unit fans to the cold aisles, where the cold air flows into the cold aisles of the data center room 3 through the perforated elevated floor to cool the server racks 5a-5 d. Each of the other aisles in which a row of server racks 7a-7d is erected is typically a cold aisle, such that cold aisles and hot aisles are arranged alternately. In the example of fig. 1, the aisle 8a to the left of the row 5a is a cold aisle and the aisle 8b to the right of the row 5a is a hot aisle.

The data center control system 13 is configured to control the aisle cooling units 9. The data center control system 13 includes a storage medium 13a including computer code and processing circuitry 13 b. The data center control system 13 may also include a plurality of controllers. The processing circuitry 13b may be configured to provide set point values to the controller controlling the aisle cooling unit 9, thereby controlling cooling of the servers in the server racks 7a-7 d. The data center control system 13 may be particularly configured to provide set points to control the temperature of the aisle cooling unit 9 and to control the airflow level of the aisle cooling unit fans.

The storage medium 13a may be implemented, for example, as a memory such as a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM) or an electrically erasable programmable read-only memory (EEPROM), and more particularly as a non-volatile storage medium of the device in an external memory such as a USB (universal serial bus) memory or a flash memory such as a compact flash.

The processing circuitry 13b may use any combination of one or more of the following operations capable of performing any of the operations disclosed herein with respect to optimization and cooling control: suitable Central Processing Units (CPUs), image processing units (GPUs), multiprocessors, microcontrollers, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), and the like.

The data center room 3 may include a plurality of temperature sensors 15. The temperature sensors 15 may be configured to measure the temperature at a number of different points of the machine room. According to some example embodiments, the temperature sensor may be configured to measure a temperature in the cold aisle, i.e. the cold aisle temperature. The temperature measurements of the cold aisle temperatures from all the sensors 15 in the aisle may form a temperature measurement vector. The temperature sensor 15 may be configured to communicate with the data center control system 13. In particular, the sensor 15 may be configured to send the temperature measurements to the data center control system 13.

The temperature measurements may be used to control a cooling device, as will be described in more detail later.

The data center control system 13 may also be configured to obtain the IT load for each server 6.

An example of a method of controlling the cooling of the cooling apparatus by means of the data centre control system 13 will now be described with reference to fig. 2.

In step S1, a temperature measurement is obtained by the data center control system 13 from the temperature sensor 15.

In step S2, optimization of the total power consumption of the aisle cooling unit 9 and the servers 6 is performed.

According to one example of optimization, there may be a constraint on cold aisle temperature. The constraints may relate to a range of allowable cold aisle temperatures, i.e., minimum and maximum allowable contract temperatures, according to contracts with data center operators.

There may also be constraints on the level of airflow provided by the aisle cooling unit 9. The restriction on the air flow level may for example be based on physical limitations of the aisle cooling unit 9, i.e. the minimum and maximum air flows that may be delivered by the aisle cooling unit 9, in particular by the aisle cooling unit fans.

The optimization in step S2 may be a multi-objective optimization, preferably a dual-objective optimization. In particular, only two targets are optimized. The first of the two targets is the power consumption of the aisle cooling unit 9, and the second of the two targets is the power consumption of the servers 6. An example of a two-target optimization formula is shown below:

min{F1(T,A),F2(T,ITload)} (1)

so that T_min≤T≤T_max (2)

A_lower≤A≤A_upper (3)

Where F1 is the total power consumption of all aisle cooling units 9, and F1 ═ Σ F1i, where i ═ 1, …, n, and n are the number of aisle cooling units 9, and F1i is the power consumption of the ith aisle cooling unit 9. F2 is the total power consumption of all servers 6, and F2 ═ Σ F2j, where j ═ 1, …, m, and m are the number of server fans 11. T is a cold aisle temperature vector, which in this example has a length n, and T_minAnd T_maxIs the contractual limit for cold aisle temperature. A is an airflow horizontal vector of airflow levels of the aisle cooling units 9, the airflow horizontal vector having n elements. ITload is a fixed vector of IT load for each server. The vector has a length k, where k ≦ m and equal to the number of servers. IT is an input parameter in optimizations (1) - (3) to describe the current IT load distribution among the servers. According to some variations of the optimization, there may be additional constraints.

In one example, the goal may be to minimize the total power consumption of the aisle cooling unit 9 and the servers 6 in a manner that provides a single solution. This may be the special case of multi-objective optimization in one example, where the sum of two objective functions F1 and F2 is optimized.

The power consumption of the aisle cooling unit 9 may be, for example, a digital model or a Computational Fluid Dynamics (CFD) model based on first principles of linearity. Alternatively, the model may be based on actual operational data. The optimization or decision variables are the cold aisle temperature T and the airflow level a of the aisle cooling unit 9. The additional input parameters are a vector of the IT load for each server.

The power consumption of the server 6 is related to the speed (i.e. the angular speed of the server fan 11) and the IT load. The digital model can for example be used to obtain the power consumption of the server 6. Alternatively, actual measurements from the history may be used to determine power consumption based on measured IT loads. The history may be contained, for example, in a look-up table, wherein the power consumption of each server 6 can be obtained based on a given cold aisle temperature. The IT load is a fixed input parameter and the sum of the power consumed by each server at a particular IT load is the output. It may be noted that since the function F2 depends on the cold aisle temperature, which depends in part on the operation of the server fans 11, the function F2 implicitly depends on the speed of the server fans.

The dual objective optimization problem can be solved numerically, for example, by means of an epsilon-constraint method, a weighting method or a lexicographical method.

In step S3, the data center control system 13 controls the aisle cooling units 9 based on the optimization result in step S2.

Step S3 may involve providing a set point to a controller controlling the aisle cooling unit 9. The set point is used to set the temperature in the aisle cooling unit 9 and/or the speed of the aisle cooling unit fan, i.e. to set the air flow level.

According to an example, before step S1, settings for optimization may be defined. For example, it may be set whether a single optimization and/or a double optimization is to be performed. In the case of dual objective optimization, the number of pareto optimal solutions to be calculated may be determined, a method of multi-objective optimization may be selected, and corresponding parameters may be selected according to the method. In addition, constraints for optimization may be defined.

Additionally, according to one example, two initial single-objective optimizations may be performed prior to step S2, one for each of the first and second objectives to obtain a minimum value for each of the first and second objectives.

If step S2 is a dual target optimization, then the optimization results in a set of pareto optimal solutions. If two initial single-objective optimizations have been performed, the two solutions thus obtained form a pareto optima point on the pareto front. Therefore, they may be included in the set of pareto optimal solutions.

If during setup it has been defined that a plurality of pareto optimal solutions may be allowed, one of the pareto optimal solutions will be selected for controlling the aisle cooling units 9.

The pareto optimal solution may be selected, for example, by means of a predetermined selection rule. The selection rule may be, for example, to use the pareto optimal solution that provides the least power consumption of the aisle cooling units or servers, i.e. one of the two solutions obtained in the two initial single-objective optimizations. In general, the selected pareto optimal solution may be that solution of a single objective optimization that provides a minimum power consumption for the aisle cooling units 9 or close to the minimum power consumption of the aisle cooling units 9, the objective of the single objective optimization being to minimize the power consumption of the aisle cooling units 9. Thus, preferably a pareto optimal solution is selected, wherein the first objective (i.e. the power consumption of the aisle cooling units 9) is prioritized in a sense, e.g. by means of a weight.

According to a variant, the operator or user may manually select the preferred pareto optimal solution to be used for step S3. The pareto optimal solution may for example be visualized as a pareto frontier, and the user may then easily compare different alternative pareto optimal solutions and select the most suitable one based on e.g. experience.

The inventive concept has been described above in general terms with reference to a few examples. However, as is readily appreciated by a person skilled in the art, other embodiments than the ones described above are equally possible within the scope of the inventive concept, as defined by the appended claims.