JP2012193877A - Cooperative control method of air conditioner with data processing load distribution - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a cooperative control method of an air conditioner with a data processing load distribution which is suitable for a data processing system for carrying out the centralized and distributed control of data processing by integrating a plurality of information communication machine rooms (data center).SOLUTION: When receiving a processing load increase ΔL(t+1) command in the next time zone (t+1) from a host server, the estimation and the comparison of an air conditioner output and a partial load factor corresponding to the processing load increase in each site are carried out. Next, it is determined that the air conditioning output estimated value R3(t+1) in the time zone t+1 about the air conditioner does not exceed the air conditioner rating output Rmax. When R3(t+1)≤Rmax, the output increase in the air condition of a site preference is carried out preceding the allocation of an ICT load. Next, the increment is allocated t, the site preference. When R3(t+1)>Rmax, the air condition increment ΔR'(t+1) corresponding to the processing load ΔL'(t+1) of capacity excess portion is allocated to an air conditioner having conditioning efficiency ranked in second place.

Description

  The present invention relates to a linkage control method for air conditioner output and data processing distribution, and is particularly suitable for a data processing system that integrates a plurality of information communication machine rooms (data centers) and performs centralized / distributed control of data processing. The present invention relates to a linkage control method for data processing and data processing distribution.

Conventionally, power capping and virtualization techniques have been adopted as energy saving measures for ICT devices / devices (hereinafter sometimes collectively referred to as servers) in data centers. In power capping, the power consumption of the server can be optimized by controlling the frequency in accordance with the usage status of the CPU. In addition, with virtualization technology, control beyond the physical resources of a plurality of servers is possible. For example, when a large number of servers perform distributed processing at a low operating rate, data processing is consolidated on some servers, or conversely, when some servers perform centralized processing at a high operating rate, It is possible to perform distributed processing on other servers.
However, these control technologies are generally operated independently of the indoor air conditioning control. If the processing load of the server is concentrated in a place where it is difficult to supply cold air, hot spots are generated and excessive air conditioning Power is required, which is against the demand for energy saving.

  In order to deal with the above problem, the applicant of the present application grasps or predicts the heat generation state of the devices based on the data processing load information of the plurality of ICT devices arranged in the cooling target zone shared by each air conditioner, and By controlling the cooling capacity and air volume, we have proposed a linkage control method between air conditioners and data processing distribution that does not generate local high-temperature areas even if the server's operating load is unevenly distributed (patent) Reference 1).

JP 2009-293551 A

  In a data center, there are cases where operations are performed by distributing to a plurality of locations with small modules in response to requests for risk distribution and sequential expansion (for example, a container data center). A technology that applies power capping and virtualization technology has been put into practical use for such an operation mode, and a link control method for air conditioners and data processing distribution that can be applied to such a situation is required.

The present invention is a technique for solving the above-described problems, and has the following contents. That is, the linkage control method of air conditioner operation and data processing load distribution according to the present invention is as follows.
(1) A method for assigning a data processing load to a plurality of data centers each including a plurality of ICT devices and one or more air conditioners for cooling the ICT devices,
When there is a change in the data processing load, or when there is a change in the air conditioner operating conditions or outside air environmental conditions at any of the data centers, The data processing load is allocated to the lower data center sequentially from the data center that is predicted to have the highest operating efficiency of the machine.
In the present invention, the “ICT device” refers to an information communication device / device such as a server, storage, or router.
Further, “change in data processing load” means an increase or decrease in data processing load.
The “data center” is a concept including, in addition to data centers having different locations, another room in the same location, a group of containers installed together, and the like.

(2) The above invention is characterized in that the upper limit value of the cooling capacity of the air conditioner of the selected data center is sequentially assigned to lower data centers.
In the present invention, the “cooling capacity upper limit value” refers to a cooling capacity rating value of an air conditioner.

(3) In each of the above inventions, the outside air environmental condition is any one of an outside air temperature, a wet bulb temperature, an equivalent outside air temperature, or a combination of two or more thereof.

(4) In each of the above inventions, the air conditioner output change is a first derivative of a partial load efficiency curve of the air conditioner.

(5) In each of the above inventions, the operating state is presence or absence of occurrence of an abnormal operation of the air conditioner.

(6) In the invention of (5), the allocation reduction rate of the data processing load is determined by the number of air conditioner abnormality occurrences in the corresponding data center.

(7) In each of the above inventions, when the data processing load increases, an operation for increasing the air conditioner output according to the increase in the processing load is performed in advance.

According to the present invention, since the data processing load is assigned in consideration of the air conditioning efficiency, the air conditioning power can be greatly reduced, and the total power consumption can be reduced.
In addition, in the invention that assigns the data processing load in consideration of the upper limit of the cooling capacity of the air conditioner or the occurrence of abnormal operation of the air conditioner, it is possible to avoid risks such as processing stoppage due to a high temperature failure of the ICT device. effective.
In addition, since the optimal air conditioner operation and data processing load are assigned in consideration of the different outdoor air environmental conditions at each location, it is possible to make the most of the advantages of locations with large outdoor air cooling such as cold districts. There is an effect that power consumption can be reduced.
In the invention that uses the first derivative of air conditioner efficiency as an index of air conditioner output change, by assigning the optimal air conditioning function, operation with the optimum air conditioning efficiency is possible even after the air conditioner function changes. It is possible to maintain the power consumption, and there is an effect that the maximum power consumption can be reduced in total in a year.

It is a figure which shows the structure of the linkage control system 1 which concerns on 1st embodiment. It is a figure which shows notionally the partial load factor-air-conditioning efficiency (COP) relationship table (TB1) of each air conditioner. It is a figure which shows notionally the relationship table (TB2) of the air-conditioner air conditioning load corresponding to a data processing load. It is a figure which shows the linkage control flow of an air conditioning machine output and data processing distribution in 1st embodiment. It is a figure which shows a priority site selection method notionally. It is a figure which shows notionally the priority site selection method in consideration of the air conditioning machine rated capacity. It is a figure which shows the structure of the linkage control system 20 which concerns on 2nd embodiment. It is a figure which shows notionally the partial load factor-air-conditioning efficiency relationship table (TB3) of each air conditioner which used outside temperature as a parameter. It is a figure which shows the linkage control flow of an air conditioning machine output and data processing distribution in 2nd embodiment. It is a figure which shows notionally the priority site selection method of 3rd embodiment. It is a figure which shows the linkage control flow of an air conditioning machine output and data processing distribution in 3rd embodiment. It is a figure which shows the structure of the linkage control system 30 which concerns on 4th embodiment. It is a figure which shows the linkage control flow of an air conditioning machine output and data processing distribution in 4th embodiment. It is a figure which shows the structure of the linkage control system 40 which concerns on 5th embodiment. It is a figure which shows typically the relationship table of the data processing speed (server power consumption upper limit setting mode) combination of each site, and an air-conditioner power consumption total value. It is a figure which shows the preparation method of the relationship table of FIG. It is a figure which shows the data processing distribution in 5th embodiment, and the linkage control flow of an air-conditioner output.

  Hereinafter, an embodiment of the linkage control method according to the present invention will be described in more detail with reference to FIGS. In order to avoid redundant description, the same components are denoted by the same reference numerals in the respective drawings. It is needless to say that the scope of the present invention is described in the claims and is not limited to the following embodiments.

<First embodiment>
A first embodiment of the present invention will be described with reference to FIGS. The present embodiment relates to a mode in which an increase is allocated to a data center that is predicted to have the highest air conditioning efficiency when the data processing load increases, and the air conditioning efficiency is improved as a total.

Referring to FIG. 1, the air conditioning system of linkage control system 1 includes base air conditioners A1 to A4 that are respectively disposed in a plurality of data centers (hereinafter referred to as sites) S1-S4. Each air conditioner cools the ICT device 5a in the server rack 5 group accommodated in the site.
The control system of the linkage control system 1 includes, for each site, an ICT device management server (hereinafter referred to as ICT server) 2, an air conditioning control server 3, and an integrated control server 4 that controls these in an integrated manner. It is prepared for. The ICT server 2 distributes the data processing commanded from the host server (not shown) to each ICT device 5a based on the command from the integrated control server 4. Note that the ICT server 2 and the integrated control server 4 do not necessarily need to be physically independent servers, and may be programs in the ICT devices 5a that perform the above functions, for example.
The integrated control server 4 includes a partial load factor-air conditioning efficiency (COP) relationship table TB1 (see FIG. 2) for each air conditioner, and a relationship table TB2 between the data processing load (L) and the air conditioner required cooling output (R). (See FIG. 3) is stored, and as described later, when there is a processing load increase command from the host server, the required air conditioner output can be calculated. Here, the partial load factor r is r = R / Rmax (where Rmax is the rated output of the air conditioner).
For convenience of explanation, in this embodiment, the air conditioners A1 to A4 are the same model, and all have the same efficiency curve.

所定の時間帯（ｔ）において各サイトの空調機Ａｉ（ｉ＝１−４）がそれぞれのデータ処理負荷Ｌｉ（ｔ）（ｉ＝１−４）に対応する冷房出力（Ｒｉ(t)）により運転されている状態（図２参照）を想定する（Ｓ１０１）。
この状態において、ホストサーバ（図示せず）から次の時間帯(ｔ+１)の処理負荷増加ΔＬ(ｔ+１)指令を受けた場合（Ｓ１０２）、各サイトについて処理負荷増加に対応する空調機出力、部分負荷率の予測及び比較を行う（Ｓ１０３）。具体的には関係テーブルＴ１に基づき、各空調機の部分負荷率ｒｉに対する空調効率ηｉを求めて比較する（図５参照）。さらに、空調効率最大の空調機が設置されているサイトを優先サイトとして選定する（Ｓ１０４）。同図の例では空調機Ａ３がこれに該当し、サイトＳ３が優先サイト候補となる。 The linkage control system 1 is configured as described above. Next, the linkage control flow of air conditioner output and data processing distribution executed by commands to the integrated control server 4 will be described with reference to FIG. Note that the following control is performed at predetermined time intervals.
In the following description, the data processing load amount (L), air conditioner required cooling output (R), and air conditioner partial load factor (r) at each site (Si (i = 1-4)) are expressed as shown in Table 1. .

In a predetermined time zone (t), the air conditioner Ai (i = 1-4) at each site uses a cooling output (Ri (t)) corresponding to each data processing load Li (t) (i = 1-4). A driving state (see FIG. 2) is assumed (S101).
In this state, when a processing load increase ΔL (t + 1) command for the next time period (t + 1) is received from the host server (not shown) (S102), air conditioning corresponding to the increase in processing load for each site. Machine output and partial load factor are predicted and compared (S103). Specifically, based on the relation table T1, the air conditioning efficiency ηi with respect to the partial load factor ri of each air conditioner is obtained and compared (see FIG. 5). Furthermore, the site where the air conditioner with the maximum air conditioning efficiency is installed is selected as the priority site (S104). In the example of the figure, the air conditioner A3 corresponds to this, and the site S3 is a priority site candidate.

Next, for the air conditioner A3, the predicted cooling output in the time zone (t + 1) R3 (t + 1) = R3 (t) + ΔR (t + 1)
Is not determined to exceed the rated output of the air conditioner (Rmax) (S105).
When R3 (t + 1) ≦ Rmax (Y in S105), the output of the air conditioner (A3) at the priority site is increased prior to the allocation of the ICT load (S107). Next, an increased processing load is assigned to the priority site (S108).
Further, when N in S105, that is, when R3 (t + 1)> Rmax, as shown in FIG. 6, the cooling output (ΔR ′) corresponding to the processing load (ΔL ′ (t + 1)) for the excess capacity as shown in FIG. (t + 1)) is assigned to an air conditioner (air conditioner A4 in this embodiment) having the next highest air conditioning efficiency (S106).

In addition, although the example which arrange | positions one air conditioner in each site was shown in this embodiment, it can also be set as the aspect which each arranges several air conditioners. In this case, the processing load can be allocated in accordance with the above description by selecting the priority site based on the total of the air conditioning functional capabilities for each site.
In the present embodiment, the example in which the ICT server 2 distributes the load to each ICT device 5a based on the command of the integrated control server 4 has been described. However, the integrated control server 4 directly distributes the load to each ICT device 5a. Also good.

<Second Embodiment>
Next, another embodiment of the present invention will be described with reference to FIGS. The present embodiment relates to a form in which an increase is allocated based on the outside air temperature at each site when the data processing load increases.
Referring to FIG. 7, the configuration of the present embodiment is different from that of the first embodiment in that temperature sensors 21 that measure the outside air temperature are arranged at each site, and the measurement data is always stored in the integrated control server 4. Is to be sent.
Further, the DB 4a of the integrated control server 4 is provided with a partial load factor-air conditioning efficiency relationship table TB3 (see FIG. 8) of each air conditioner using the outside air temperature as a parameter. Since other configurations are the same as those of the first embodiment, a duplicate description is omitted.

Hereinafter, with reference to FIG. 9, the linkage control flow of the air conditioner and data processing distribution in this embodiment will be described.
Assume that the air conditioner Ai (i = 1-4) at each site is operated at the output Ri (t) corresponding to the data processing load Li (t) in the predetermined time zone (t). (S201). During the control, the temperature sensor 21 at each site measures the outside air temperature T (i) and constantly transmits it to the integrated control server 4 (S202).

  In this state, when a processing load increase (ΔL (t + 1)) command is received for the next time period (t + 1) (S203), each site is in the process of increasing the processing load under the current outside air temperature environment. The efficiency of each air conditioner is predicted and compared (S204). Specifically, referring to FIG. 8, the air conditioning efficiency ηi with respect to the partial load factor ri (t + 1) of each air conditioner under the outside air temperature condition at each site is obtained. These are compared, and the air conditioner installation site having the maximum value is selected as the priority site (S205). In the example of FIG. 2 (and Table 2), the air conditioner A2 corresponds to this, and the site 2 is selected as a priority site candidate.

その後のＳ３０１以降のフローは上述の実施形態のＳ１０５乃至Ｓ１０８と同様であるので、重複説明を省略する。

The subsequent flow after S301 is the same as S105 to S108 of the above-described embodiment, and thus the duplicated description is omitted.

In the present embodiment, an example in which the priority site is selected in consideration of the outside air temperature has been described. However, the priority site may be selected based on the wet bulb temperature or the equivalent outside air temperature.
Furthermore, more simply, the priority site can be selected based only on the outside temperature without considering the air conditioner partial load factor and the air conditioning efficiency.

<Third embodiment>
Next, another embodiment of the present invention will be described with reference to FIGS. The difference between the present embodiment and the first embodiment is not based on the air conditioning efficiency (COP) of the air conditioner when selecting the priority site, but based on the numerical value of the differential coefficient (first derivative) of the air conditioning efficiency curve. The present invention relates to the mode of determination (see FIG. 10). Since the configuration of the present embodiment is the same as that of the first embodiment, redundant description is omitted.

Hereinafter, with reference to FIG. 11, the linkage control flow of the air conditioner and data processing distribution in the present embodiment will be described.
It is assumed that the air conditioner Ai (i = 1-4) at each site is operated corresponding to each data processing load Li (t) (i = 1-4) in a predetermined time zone (t). (S301).
In this state, when a processing load change ΔL (t + 1) command is received in the next time zone (t + 1) (S302), the partial load factor and COP of each air conditioner at the current time point are grasped for each site. (S303). Further, the differential coefficient (dη / dri) of the air conditioning efficiency curve is obtained for each air conditioner (S304).

Next, when the processing load increases (ΔL (t + 1)> 0), the increased processing load is assigned to the air conditioner at the site with the largest differential coefficient (S305). Next, prior to the allocation of the ICT load, the capacity of the air conditioner at the priority site is increased (S306). Next, an increased processing load is allocated to the priority site (S307).
On the other hand, when the processing load decreases (ΔL (t + 1) <0), the processing load amount corresponding to the site having the smallest differential coefficient is reduced (S308).

<Fourth embodiment>
Next, another embodiment of the present invention will be described with reference to FIGS. The present embodiment relates to a mode for facilitating the smooth operation of data processing by transferring a data processing load to another site as necessary when an air conditioner abnormality occurs at any site.
Referring to FIG. 12, the linkage control system 30 according to this embodiment is different from the above-described embodiments in that the configuration of the air conditioning system, that is, a plurality of air conditioners are disposed at one site. . Taking site S31 as an example, base air conditioners A11 and A12 and task air conditioners B11 and B12 are arranged in the rack row. The sites S32 and S33 are similarly configured. Other configurations are the same as those in the first embodiment including the control system, and thus the duplicate description is omitted.

Hereinafter, the linkage control flow in the present embodiment will be described with reference to FIG.
During normal operation, the air conditioner is operated in accordance with the processing load assigned to each site (S401).
From that state, when an air conditioner abnormality signal is received at one of the sites (Y in S402), the processing load reduction amount corresponding to the decrease in the air conditioning functional capability at the site is grasped (S403).
Next, it is determined whether or not the processing load amount to be transferred can be complemented by increasing the output of air conditioners other than abnormal machines in the site (S404).
When complementation is possible (Y in S404), the load is redistributed according to the priority order of the on-site air conditioners in accordance with the flow of S105 to S108 of the first embodiment (S405).
If the air conditioner in the site cannot be supplemented (N in S404), the necessary supplement is distributed to the air conditioners in other sites for the excess (S406).

<Fifth embodiment>
Further embodiments of the present invention will be described with reference to FIGS. In this embodiment, the review of the processing load sharing is not performed when there is a change in the data processing load, but by periodically reviewing the environmental conditions (here, the outside temperature) at each site, the server and air conditioning are continuously The present invention relates to a mode for reducing power consumption of a machine.

Referring to FIG. 14, the linkage control system 40 according to the present embodiment uses the integrated control server 4 to process the processing loads of data centers (sites) Sa and Sb arranged in two different regions (for example, the southern hemisphere and the northern hemisphere). A mode of virtualization control is assumed. In the control, it is assumed that the minimum requirement of processing capacity of the entire control (here, 5 PFLOP) is determined. Where PFLOPS (Peta Floating
Point Operation Per Second) is a unit that represents the processing performance of a computer, and is the ability to perform a floating point operation 1000 trillion times per second.

The DB 4a of the integrated control server 4 stores an air conditioner power consumption total table for performing this control. The contents of the table will be described below.
As shown in FIG.
(1) Outside air temperature average values (av (Ta), av (Tb)) over the past week for the sites Sa and Sb
(2) It is configured as a relationship table of air conditioner total power consumption (ΣE = Ea + Eb) for various combinations of server power consumption upper limit setting modes (P0, P1, P2, P3) of each site.
Here, the power consumption upper limit setting mode of the server is a control for suppressing the operation so as not to exceed the set power consumption upper limit by controlling the CPU performance (core frequency) of the server and the number of operating servers. . The maximum data processing speed is determined by selecting the server power consumption upper limit setting mode (P0, P1, P2, P3) (hereinafter abbreviated as “setting mode” as appropriate).
In the figure, the values of ΣE according to the combination of the setting modes of the respective servers in the case of av (Ta) = Ti and av (Tb) = Tj conditions for the sites Sa and Sb are schematically shown. Such a relationship table is prepared for combinations of various outside air temperature average values (av (Ta), av (Tb)). In the figure, the shaded portion represents a combination range of setting modes that satisfies the condition that the required minimum processing capacity is 5 PFLOP or more.

Next, ΣE can be obtained as follows. Referring to FIG. 16, the relationship between the setting mode and the amount of data processing (PFLOPS) is obtained as shown in the diagram (a) of FIG. As shown in FIG.
The value of PFLOPS for a specific setting mode combination is obtained from FIG. 5A (4PFLOPS at P0, 3PFLOPS at P1,...). In addition, since the setting mode and the server heat generation amount are proportional, and the server heat generation amount is proportional to the air conditioning load, ΣE = Ea + Eb is obtained for a specific combination of setting modes according to FIGS. Can do. For example, in the setting condition = P0, av (Ta) = 10 ° C. for the site Sa, PFLOPS = 4 and Ea = E (P0, 10). Further, for the site Sb, under the setting mode = P3, av (Tb) = 30 ° C., PFLOPS = 1, Eb = E (P3, 30), and ΣE = E (P0, 10) + E (P3, 30) is obtained. . In this way, the relationship table of FIG. 15 can be created.

Hereinafter, the linkage control flow in the present embodiment will be described with reference to FIG.
During normal operation, load processing is performed at a given (previous update) setting (processing speed) condition for each site, and the air conditioner is in operation (S501). When the update timing comes from that state (Y in S502), the average outside air temperature values (av (Ta), av (Tb)) in the past one week at the sites Sa and Sb are calculated (S403). Further, the air conditioner total power consumption table to be applied is updated based on the calculation result (S504), and the combination of the minimum power consumption satisfying the required processing speed (5 PFLOPS or more) is selected using this table (S505). .
The setting mode is changed at each site, and data processing is performed according to the changed setting mode (S506). In conjunction with this, the capacity of the air conditioner is controlled.

In the present embodiment, an example is shown in which the total power consumption of the air conditioner is set as a control target of power consumption. However, the present invention is not limited to this. You can also.
Moreover, although the example of linkage control of the air conditioner and ICT apparatus regarding two sites was shown, it can also be set as the aspect of linkage control of three or more sites.
Moreover, although the example which regularly reviews processing load sharing based on the external temperature of each site was shown, it can also be set as the aspect based on other environmental conditions.

  The present invention can be widely applied to a linkage control system in a machine room that houses an ICT device regardless of a heat source of an air conditioner, a refrigerant, an air conditioning system, a building structure, a data processing type, and an ICT device type.

1, 20, 30, 40... Linkage control system 2... ICT device management server (ICT server)
3 ... Air conditioning control server 4 ... Integrated control server 5 ... Server rack 21 ... Temperature sensors A1-A4, A11-A32 ... Base air conditioners B11-B32 ... Task air conditioning Machines S1-S4, S31-S33, Sa, Sb ... Site (data center)

Claims (7)

A data processing load assignment method for a plurality of data centers each including a plurality of ICT devices and one or more air conditioners for cooling the ICT devices,
When there is a change in the data processing load, or when there is a change in air conditioner operating conditions or outside air environmental conditions at any data center,
Air conditioner operation characterized by allocating data processing loads to lower data centers sequentially from the data center that is predicted to have the highest operating efficiency of the air conditioner in response to changes in the air conditioner output at the changed data processing load And data processing load distribution linkage control method.   2. The linkage control method for air conditioner operation and data processing load distribution according to claim 1, wherein the upper limit value of the cooling capacity of the air conditioner of the selected data center is assigned as a limit.   The linkage between the air conditioner operation and the data processing load distribution according to claim 1 or 2, wherein the outside air environmental condition is any one of outside air temperature, wet bulb temperature, equivalent outside air temperature, or a combination of two or more thereof. Control method.   The linkage control method for air conditioner operation and data processing load distribution according to any one of claims 1 to 3, wherein the air conditioner output change is a first derivative of a partial load efficiency curve of the air conditioner.   5. The linkage control method for air conditioner operation and data processing load distribution according to any one of claims 1 to 4, wherein the operation state is presence or absence of occurrence of an air conditioner operation abnormality.   6. The linkage control method for air conditioner operation and data processing load distribution according to claim 5, wherein the allocation reduction ratio of the data processing load is determined by the number of air conditioner abnormality occurrences in the corresponding data center.   The data processing load distribution according to any one of claims 1 to 6, wherein when the data processing load increases, an operation of increasing an air conditioner output in accordance with the increase of the processing load is performed in advance. Linkage control method.

Images