CN114901057A - Multi-point energy consumption detection and dynamic regulation system in data center machine room - Google Patents

Multi-point energy consumption detection and dynamic regulation system in data center machine room Download PDF

Info

Publication number
CN114901057A
CN114901057A CN202210814549.9A CN202210814549A CN114901057A CN 114901057 A CN114901057 A CN 114901057A CN 202210814549 A CN202210814549 A CN 202210814549A CN 114901057 A CN114901057 A CN 114901057A
Authority
CN
China
Prior art keywords
heat
scheduling
heat dissipation
value
dissipation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202210814549.9A
Other languages
Chinese (zh)
Other versions
CN114901057B (en
Inventor
高锡超
程伟
余伟雄
毛彦堃
李堉鑫
魏蕤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Unicom Guangdong Industrial Internet Co Ltd
Original Assignee
China Unicom Guangdong Industrial Internet Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Unicom Guangdong Industrial Internet Co Ltd filed Critical China Unicom Guangdong Industrial Internet Co Ltd
Priority to CN202210814549.9A priority Critical patent/CN114901057B/en
Publication of CN114901057A publication Critical patent/CN114901057A/en
Application granted granted Critical
Publication of CN114901057B publication Critical patent/CN114901057B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
    • H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
    • H05K7/00—Constructional details common to different types of electric apparatus
    • H05K7/20—Modifications to facilitate cooling, ventilating, or heating
    • H05K7/20709—Modifications to facilitate cooling, ventilating, or heating for server racks or cabinets; for data centers, e.g. 19-inch computer racks
    • H05K7/20836—Thermal management, e.g. server temperature control
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • H—ELECTRICITY
    • H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
    • H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
    • H05K7/00—Constructional details common to different types of electric apparatus
    • H05K7/20—Modifications to facilitate cooling, ventilating, or heating
    • H05K7/20709—Modifications to facilitate cooling, ventilating, or heating for server racks or cabinets; for data centers, e.g. 19-inch computer racks
    • H05K7/20718—Forced ventilation of a gaseous coolant
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The invention provides a multipoint energy consumption detection and dynamic regulation system in a data center machine room.A environment monitoring module comprises a monitoring unit, a data processing unit and a data processing unit, wherein the monitoring unit is used for establishing an actually-measured environment data model and generating a heat production model of the machine room; the environment configuration module is configured with a heat dissipation strategy database in which heat dissipation strategies are stored and a scheduling path scheduling database in which scheduling paths are stored; the heat dissipation limitation judging unit screens heat dissipation strategies as a first strategy group according to the power limitation heat dissipation value; the heat dissipation fitting unit is used for fitting and calculating the heat exchange efficiency value through the heat dissipation runner index in the first strategy group and the heat production model of the machine room to obtain a scheduling model; the scheduling fitting unit generates a scheduling strategy to obtain a scheduling efficiency value; the strategy selection module is configured with a scheduling selection algorithm and selects a heat dissipation strategy with the minimum scheduling total energy consumption value as a target heat dissipation strategy; and the execution module is used for executing the target heat dissipation strategy and the scheduling strategy. The system determines the optimal heat dissipation and scheduling strategy to regulate the temperature of the machine room of the data center, and reduces the energy consumption of equipment to the maximum extent.

Description

Multi-point energy consumption detection and dynamic regulation system in data center machine room
Technical Field
The invention relates to the field of machine room environment monitoring and adjusting, in particular to a multipoint energy consumption detection and dynamic adjusting system in a data center machine room.
Background
At present, a data center is a very common design in modern network technology, and the use of the data center enables a large amount of data operation and data storage to be supported at present, but what is urgently needed to be solved in China is the energy consumption problem of the data center at present, the data center serves as a large energy consumption user, a refrigeration system becomes an important part of the energy consumption of the data center, at present, PUE (power utility efficiency, power utilization efficiency) becomes an internationally popular data center use efficiency measurement index, a PUE value is equal to the ratio of the total energy consumption of the data center to the energy consumption of IT equipment, wherein the total energy consumption of the data center comprises the energy consumption of the IT equipment and the energy consumption of systems such as refrigeration and power distribution, and the value is greater than 1, and the closer to 1 indicates that the energy consumption of non-IT equipment is less, and the level of energy efficiency is better.
According to statistics, the PUE value of the foreign advanced machine room can reach 1.21, while the average value of the PUE of our country is more than 2.5, which means that every time the IT equipment consumes 1 degree of electricity, the electricity of 1.5 degrees is consumed by the machine room facilities, so that the PUE value reduction becomes a very key problem in the energy consumption and energy saving technology of the data center.
Disclosure of Invention
The invention aims to overcome at least one defect of the prior art, and provides a multipoint energy consumption detection and dynamic regulation system in a data center machine room, which is used for solving the problems that the conventional refrigeration system of the data center has high energy consumption and cannot effectively reduce the PUE value.
The technical scheme adopted by the invention comprises the following steps:
in a first aspect, the invention provides a multipoint energy consumption detection and dynamic regulation system in a data center machine room, which comprises an environment monitoring module, an environment configuration module, a heat dissipation fitting module and an execution module; the environment monitoring module comprises a plurality of monitoring units arranged at different positions of a data center machine room; the monitoring unit is used for monitoring the change of environmental information and feeding back environmental data; the environment monitoring module establishes an actual measurement environment data model according to environment data fed back by monitoring units at different positions and generates a heat generation model of the machine room according to the actual measurement environment data model; the environment configuration module is configured with a heat dissipation strategy database and a scheduling database, wherein the heat dissipation strategy database stores heat dissipation strategies, and each heat dissipation strategy corresponds to a heat dissipation flow channel index and a power limiting heat dissipation value; the scheduling database is configured with a plurality of scheduling paths, and each scheduling path corresponds to a scheduling heat exchange value; the heat dissipation fitting module comprises a heat dissipation limitation judging unit, a heat dissipation fitting unit, a scheduling fitting unit and a strategy selecting unit; the heat dissipation limitation is used for calculating to obtain an expected heat dissipation value by the judgment unit according to the heat production model of the machine room, and taking a heat dissipation strategy that the corresponding power limitation heat dissipation value is higher than the expected heat dissipation value as a first strategy group; the heat dissipation fitting unit is used for fitting the heat dissipation flow channel indexes of the heat dissipation strategies in the first strategy group with the heat production model of the machine room so as to calculate the heat exchange efficiency value of each heat dissipation strategy and obtain a corresponding scheduling model; the scheduling fitting unit is used for obtaining a corresponding scheduling path according to the analysis of each scheduling model, generating a corresponding scheduling strategy, and summing a scheduling heat exchange value corresponding to each scheduling path to obtain a scheduling efficiency value of each heat dissipation strategy; the strategy selection module is configured with a scheduling selection algorithm, the scheduling selection algorithm is used for calculating a scheduling total energy consumption value of each heat dissipation strategy according to the heat exchange efficiency value and the scheduling efficiency value of each heat dissipation strategy, and the heat dissipation strategy with the minimum scheduling total energy consumption value is selected as a target heat dissipation strategy; the execution module is used for executing the target heat dissipation strategy and the scheduling strategy corresponding to the same heat dissipation strategy.
The system provided by the invention is mainly divided into three modules, namely an environment monitoring module for monitoring environment data in a data center machine room, an environment configuration module for providing heat dissipation scheduling strategy data, and a heat dissipation fitting module for finally selecting a proper heat dissipation and scheduling strategy through fitting, calculation and judgment, and finally executing the corresponding strategy through an execution module so as to achieve the effect of effectively reducing the energy consumption of equipment in the machine room. The environment data model is established through the environment data generated by the monitoring units in the environment monitoring module, the heat generation model of the machine room is generated, the heat generation condition of the machine room can be predicted, the actual heat generation point of the machine room is known, and the selection of a more targeted heat dissipation strategy is facilitated. Secondly, based on the heat dissipation strategy database and the scheduling database in the environment configuration module, the heat dissipation fitting module considers the balance between the scheduling heat exchange value and the scheduling total energy consumption value in the intelligent operation process, and the optimal heat dissipation strategy is obtained through calculation, so that the effect of saving the energy consumption of the refrigeration equipment is achieved to the greatest extent.
Further, the environment monitoring module comprises a model conversion unit configured with a model conversion strategy; the model conversion strategy is used for generating a machine room heat generation model according to the actually measured environment data model; the model conversion strategy comprises: a heat generation prediction step, which comprises generating a heat generation prediction function according to the historical power information of each server; the historical load database stores historical power information of each server; the heat dissipation iteration step comprises the steps of calculating the theoretical heat value of each dissipation position through a thermal dissipation algorithm, substituting the theoretical heat value into the thermal iteration algorithm to calculate the iteration heat value at the next moment, and acquiring the iteration heat value of a corresponding server in a preset period to establish a heat value data set; a model generating step including generating the machine room heat production model based on marking a location for each thermal value data set.
A model conversion strategy in the environment monitoring module generates a machine room heat generation model according to actual environment data, wherein the heat value generated by the server in a specific period is determined based on historical power information of each server, the machine room heat generation model is generated based on heat value data, and the heat generation condition of the machine room can be effectively predicted, so that the heat generation point of the machine room is determined, and the heat is radiated in a targeted manner.
Further, the thermal runaway algorithm is:
Figure 236444DEST_PATH_IMAGE001
Figure 506068DEST_PATH_IMAGE002
(ii) a Wherein the content of the first and second substances,
Figure 987865DEST_PATH_IMAGE003
the theoretical thermal value of the corresponding dissipation location at time t,
Figure 135950DEST_PATH_IMAGE004
the weight parameter for the nth server associated with the escape location is
Figure 754013DEST_PATH_IMAGE005
，
Figure 962140DEST_PATH_IMAGE006
A is a preset first heat exchange parameter, b is a preset second heat exchange parameter, c is a preset third heat exchange parameter,
Figure 665654DEST_PATH_IMAGE007
the temperature values for the escape locations in the environmental data model,
Figure 86271DEST_PATH_IMAGE008
is the indoor ambient temperature value in the ambient data model,
Figure 824420DEST_PATH_IMAGE009
the outdoor environment temperature value in the environment data model; the hot iteration algorithm is
Figure 937869DEST_PATH_IMAGE010
(ii) a Wherein the content of the first and second substances,
Figure 128679DEST_PATH_IMAGE011
the corresponding heat generation prediction function for the nth server associated with the escape location is at t 1 Predicted heat production at time, t 1 And t, a preset iteration time interval is formed, d is a preset fourth heat exchange parameter, e is a preset fifth heat exchange parameter,
Figure 352987DEST_PATH_IMAGE012
an escape weight for the escape location, is
Figure 945643DEST_PATH_IMAGE013
And m is the total number of the dissipation positions of the system, and the heat production predicted value generated by the heat iteration algorithm is used as the iteration heat value.
In the practical application process, the temperature of the data center is also influenced by the environmental conditions at the moment, and the temperature of each server is also influenced by the position of the server. And the heat dissipation iteration step is configured with a heat dissipation algorithm and a heat iteration algorithm, the theoretical heat value of each dissipation position is calculated through the heat dissipation algorithm, the theoretical heat value is brought into the heat iteration algorithm to calculate the iteration heat value at the next moment, and the iteration heat value corresponding to the server in the preset period is obtained to establish a heat value data set. The purpose of the thermal dissipation algorithm is to calculate the heat corresponding to the dissipation position, and since the dissipation heat at the dissipation position is determined by the heat production of each server, the measured temperature of the current dissipation position and the overall heat dissipation efficiency, the theoretical heat value at time t can be obtained by calculating the function, wherein the weight corresponding to each server needs to be considered
Figure 495573DEST_PATH_IMAGE014
The longer the distance and the larger the height deviation are, the smaller the thermal influence on the dissipation position is, so that the corresponding weight is configured according to the position of each server according to the distance and height relation, the a/b/c are preset heat exchange parameters and are related to the environmental humidity, the larger the humidity is, the smaller the heat exchange parameters are, the corresponding temperature value is converted into a heat calculation unit to enable the numerical magnitude to be equal, the heat of the dissipation position is subtracted from the heat generated by the actual server, the first heat difference can be obtained, the second heat difference can be obtained through the heat diffused to other spaces by the servers, and the second heat difference is obtained through the heat generated by the serversOpening ventilation is needed to be considered, heat diffusion under a refrigeration condition is not considered, so that the heat effect of each server on the dissipation position can be obtained, the heat of the dissipation position is estimated, then heat generation of the server is iterated through a heat iteration algorithm, as the heat generation function of an original server only considers the heat generated by processing data of the server, the dissipation heat is also considered at the moment, more accurate prediction can be generated on the server by combining environmental factors, the dissipation position is an important position in a heat dissipation step and is also a key position of a model, as a heat dissipation flow channel mainly flows through the dissipation position, namely heat change of the dissipation position directly influences heat exchange efficiency, and the original function directly monitors the server by neglecting the dissipation position, heat superposition of the server is calculated through the actual influence of the dissipation position through the heat generation iteration algorithm, and is more reliable. The heat iteration algorithm is to calculate the heat value of the heat generated by the server in the primitive function, and the value is a predicted value, so the heat value can be directly obtained in the primitive function, then the heat dissipation at the previous moment is calculated, and meanwhile, the inverse influence of the heat at the dissipation position at the previous moment is subtracted, namely, the theoretically required equivalent cold quantity of the heat taken away from the dissipation position is obtained, the higher the heat at the dissipation position is, the closer the dissipation position is to the heat exhaust port, the lower the height of the dissipation position is, the easier the heat is taken away from the dissipation position, and the heat at the dissipation position can be calculated, and the weight is obtained
Figure 173679DEST_PATH_IMAGE015
According to the distribution of the height and the distance of the heat exhaust opening, d/e is also a preset heat exchange parameter and is related to humidity, so that the heat influence on the server from the heat change of the dissipation position can be calculated, the heat of the server is equivalent to the heat of the dissipation position, and then a heat dissipation strategy, such as the setting of a heat dissipation flow channel and the setting of power, only needs to consider the influence on the dissipation position, so that the heat influence value of the server under the effect of no cold is obtained, and the heat dissipation strategy can be guided more accurately.
Further, the heat production predicting step further comprises: the cluster analysis sub-strategy is used for acquiring the average power and the average CPU temperature of the servers in the characteristic time periods from the historical power information according to the characteristic time periods, performing cluster analysis on the average power and the average CPU temperature through a cluster analysis algorithm to establish a grouped index for the servers in each characteristic time period, taking the servers with the completely same grouped index as the same server type group, calculating the average heat generation waveform of each server type group according to the historical power information, fitting the average heat generation waveform according to the reference heat generation waveform to obtain a fitted heat generation waveform, and connecting the heat generation fitted waveforms in different characteristic time periods to obtain a heat generation prediction function.
The cluster analysis sub-strategy firstly obtains a discrete power value of the server and a CPU temperature value according to a characteristic time period, then carries out cluster analysis on discrete points in the same time period, and carries out type division on a plurality of servers. The server power value and the CPU temperature value data are counted to avoid the possibility that multiple types of servers exist in one computer room, it is possible to process the same amount of information, but with different heat generation, and therefore it is necessary to acquire both features, so that servers of different types are not divided into different feature groups in the cluster analysis, when the grouping is obtained, the average heat generation waveform is obtained through all the historical data in the group to obtain the characteristic sample, the effects of increasing the number of samples and reducing abnormal historical data are achieved, a plurality of servers can share the sample to reduce the operation amount when processing the data, and meanwhile, the data layer difference caused by different types of the servers can be avoided, and judging the heat generation change of each server in the future time according to the samples, recording the waveform as a heat generation function, and reflecting the change of the heat generated by the servers along with the time.
Further, the environmental monitoring module includes: and the model correction unit is used for judging whether the difference value between the iterative heat value and the heat prediction value at the same moment is greater than a first deviation threshold value or not, or whether the difference value is greater than a second deviation threshold value or not, if so, correcting the corresponding reference heat generation waveform to reduce the difference value between the iterative heat value and the heat prediction value, and if so, correcting the neighborhood radius parameter in the corresponding cluster analysis algorithm to generate a new server type group.
In order to continuously optimize a processing strategy according to a newly input sample, generally, an iteration threshold value and a thermal prediction value do not deviate too much, and if the deviation is too great, a model needs to be corrected.
Further, the environment configuration module includes: the heat dissipation configuration unit is used for acquiring control information of each refrigeration device in the machine room, acquiring heat dissipation permission directions from the control information, associating the heat dissipation permission directions of different refrigeration devices to acquire heat dissipation paths, screening the heat dissipation paths through preset path screening conditions to acquire heat dissipation channels, generating a corresponding heat dissipation strategy according to the acquired heat dissipation channels, and generating power limit heat dissipation values according to rated powers of the corresponding refrigeration devices in the heat dissipation strategies; the path screening conditions comprise cold quantity superposition constraint, airflow short circuit constraint and total vector constraint; the cold quantity superposition constraint is used for judging the coincidence quantity of intersection points between the heat dissipation permission directions in the heat dissipation flow channel; the airflow short circuit constraint is used for judging the relative airflow quantity in the heat dissipation flow channel; the total vector constraint is used for judging the direction of the unit vector sum of all heat dissipation permission directions in the heat dissipation flow channel.
And the heat dissipation configuration unit generates a corresponding heat dissipation strategy according to the obtained heat dissipation flow channel, and generates a power limit heat dissipation value according to the rated power of the corresponding refrigeration equipment in the heat dissipation strategy. The last heat dissipation path is set according to the following constraint conditions, one is cold quantity superposition constraint, if a plurality of refrigeration devices have the same orientation in one heat dissipation path, cold quantity waste can be caused, the heat dissipation path can be eliminated, the other is airflow short circuit constraint, if the number of short-circuit airflow is large, heat dissipation efficiency is low, the heat dissipation paths can be deleted from all the heat dissipation paths, and finally, the total vector constraint is required to face an air outlet or vertically upwards, so that gas circulation is easily formed, and heat dissipation is facilitated. Therefore, all heat dissipation paths can be obtained, the power limit heat dissipation value can be calculated according to the rated power of the refrigeration equipment started in each heat dissipation path, the actual heat dissipation power in the heat dissipation instruction can be adjusted according to the actual temperature, the air outlet speed proportion is fixed, but the overall flow speed of the air flow of the overall heat dissipation path can be changed, and the adjustment is carried out through the actual heat and the power.
Further, the environment configuration module includes: the scheduling configuration unit is used for configuring the configuration of the scheduling database, generating a virtual scheduling path according to the communication topological relation between the servers and generating a virtual scheduling task value; the server corresponding to the virtual scheduling path executes to obtain a local processing heat value and a scheduling processing heat value, a corresponding scheduling heat value is generated according to the local processing heat value and the scheduling processing heat value, the virtual scheduling path with the scheduling heat value lower than a scheduling threshold value is screened from the virtual scheduling paths to serve as the scheduling path, and the scheduling path is stored in a scheduling database.
When the virtual scheduling task is executed in the two devices corresponding to the scheduling path, the CPU temperature of the local device is obtained to obtain the local processing heat value, the scheduling processing heat value is obtained according to the obtained CPU temperature of the local device and the change of the CPU temperature of the device executing the virtual scheduling task, the heat value variable quantity can be obtained by calculating the difference value, if the heat value variable quantity is increased greatly, the scheduling loses the original meaning, therefore, the scheduling path under the condition is not stored, and only the virtual scheduling path which meets the condition that the scheduling heat value is lower than the scheduling threshold value in the virtual scheduling path can be used as the scheduling path and stored in the scheduling database.
Further, the heat dissipation fitting module includes: the dissipation fitting algorithm is used for calculating a fitting heat dissipation difference value of each dissipation position, and the heat exchange efficiency value of the heat dissipation strategy is the sum of the fitting heat dissipation difference values corresponding to the heat dissipation strategy; the dissipation fitting algorithm is:
Figure 201678DEST_PATH_IMAGE016
Figure 648839DEST_PATH_IMAGE017
(ii) a Wherein T is d Fitting a heat dissipation difference value for the dissipation position; p t Is a preset reference heat dissipation value;
Figure 104092DEST_PATH_IMAGE018
the branch stage number corresponding to the dissipation position;
Figure 269494DEST_PATH_IMAGE019
the heat dissipation flow ratio corresponding to the dissipation position;
Figure 101183DEST_PATH_IMAGE020
the effective heat dissipation distance between the dissipation position and the heat dissipation flow channel;
Figure 137273DEST_PATH_IMAGE021
and configuring a fitting dissipation grade for the dissipation position according to the fitting dissipation difference value of each dissipation position to generate the scheduling model.
The logic of the fit of the dissipation algorithm is as follows, because the heat dissipation flow channel is known, the effective heat dissipation effect of each dissipation position can be obtained according to the heat dissipation flow channel, because the larger the cold input in the flow channel is, the better the heat dissipation effect of the dissipation position passing by the flow channel is, the heat dissipation fit difference of each dissipation position is calculated through the logic, namely the heat dissipation effect of each heat dissipation strategy corresponding to the position is, and the calculation of the heat dissipation fit difference needs to consider two factors, the first is the effective cold supply in the heat dissipation strategy, and the second is the effective heat generation corresponding to the dissipation position. The provision of cold requires consideration of: the first branch stage number, i.e. the branch of the heat dissipation channel, if the heat dissipation effect of the main channel is due to the heat dissipation effect of the branch, for example, the number of the score stage number of the main channel is defined as 1, and each branch of the main channel is reduced by 0.1, the number of the branch stage can be obtained, andSlthe heat dissipation flow ratio is such that the reaction gas can easily escape from the flow channel, for example, the flow channel is a confluence of three flow channels, but the air flow efficiency is low if only one output is provided, so that the corresponding heat dissipation flow ratio can be obtained by dividing the number of input flow channels by the number of output flow channels. Although the heat dissipation flow path passes through the dissipation position, it is generally less able to cover the dissipation position, so the relative distance between the dissipation position and the heat dissipation flow path is increasedThe larger the cooling capacity, the less susceptible the cooling capacity is, the more the effective heat dissipation distance needs to be considered, and the more the effective heat dissipation distance needs to be according to the upper and lower positional relationship, for example, the heat dissipation channel is below the dissipation position, so that heat is hardly dissipated due to the rise of hot air, and if the heat dissipation channel is above the dissipation position, it is easier, so that the faster the effective heat dissipation distance is reduced for every upward distance, and the slower the effective heat dissipation distance is reduced for every downward unit distance, and the sum of the effective cooling capacity values of the equipment corresponding to the cooling capacity input before the dissipation position, where the effective cooling capacity value is the rated cooling power of the equipment divided by the number of nodes, and if the cooling capacity of the refrigeration equipment passes through 4 dissipation positions before reaching the dissipation position, the effective cooling capacity is correspondingly reduced. Secondly, the influence of heat to the dissipation position, the original heat of the dissipation position can be known and calculated by superposition of a heat production model at the dissipation position, the heat influence of the dissipation position at any moment can be known, the size and the positive and negative conditions of the position corresponding to the fitting heat dissipation value of the heat dissipation strategy can be calculated, if the fitting heat dissipation value is overlarge, the waste of large cold quantity is explained for changing the position, if the fitting heat dissipation value is a negative value, the position can not be cooled well, the server can not exhaust heat, and through calculating the sum of the fitting heat dissipation values (absolute value), whether the heat dissipation strategy is matched with the corresponding heat production prediction result can be judged, the higher the fitting heat dissipation value is, and the more serious heat waste or loss is explained.
Further, the scheduling selection algorithm is as follows:
Figure 29005DEST_PATH_IMAGE022
wherein
Figure 681703DEST_PATH_IMAGE023
In order to schedule the total power consumption value,
Figure 51505DEST_PATH_IMAGE024
is a preset energy consumption proportion parameter,
Figure 207680DEST_PATH_IMAGE025
is the scheduling efficiency value.
Furthermore, the scheduling fitting unit is configured with a crawler scheduling strategy and used for analyzing the scheduling model to obtain a corresponding scheduling path; the crawler scheduling policy comprises: a topology configuration step, which is used for establishing a crawler scheduling topology through a scheduling path by taking a corresponding server in the scheduling model, of which the fitting escape level is higher than a preset level standard, as an initial coordinate and taking a corresponding server in the scheduling model, of which the fitting escape level is lower than the preset level standard, as an end coordinate; a crawler configuration step, which is used for generating a plurality of virtual crawler tasks and configuring a reference pheromone for each virtual crawler task; when the virtual crawler task passes through an escape position, the escape position obtains corresponding reference pheromones, and each escape position is configured with an information attenuation algorithm for reducing the pheromones of the escape position along with time; a crawler executing step, which is used for continuously configuring a virtual crawler task for each initial coordinate until reaching the end coordinate; and a pheromone triggering step, wherein when the pheromone of the escape position reaches the corresponding pheromone triggering threshold value, the escape position is taken as an end point coordinate, the escape position with the highest associated priority value is taken as a starting coordinate to determine a scheduling path, the scheduling path is taken as the starting coordinate to be deleted from the crawler scheduling topology, and meanwhile, the fitting escape grade taken as the end point coordinate is increased.
A crawler algorithm is adopted in a scheduling fitting unit, the crawler task randomly moves until a terminal coordinate is found by configuration, nodes with higher pheromones are easier to select when a crawler runs, so a server which is used as a middle node is easier to achieve a triggering condition of the pheromone, the pheromone is attenuated according to time, the attenuation factors comprise a scheduling heat exchange value of an escape position corresponding to a scheduling path and a difference value of all scheduling heat exchange values divided by a fitting escape grade to obtain a basis of attenuation of the pheromone, the attenuation is slower when the grade difference is larger, the heat dissipation effect is easier to improve, the waste is more serious when the scheduling heat exchange value is higher, the attenuation is faster, and therefore when the pheromone of the escape position reaches a triggering threshold value of the corresponding pheromone, the escape position is used as the terminal coordinate, the escape position with the highest associated priority value is used as a starting coordinate to determine the scheduling path, and deleting the coordinates as the initial coordinates from the crawler scheduling topology, and increasing the fitting escape level as the end point coordinates.
Further, the information attenuation algorithm is as follows:
Figure 270314DEST_PATH_IMAGE026
(ii) a Wherein
Figure 144729DEST_PATH_IMAGE027
Is the pheromone at the last moment,
Figure 583800DEST_PATH_IMAGE028
is the pheromone at the next time instant,
Figure 860061DEST_PATH_IMAGE029
in order to be a pre-set attenuation parameter,
Figure 828017DEST_PATH_IMAGE030
the scheduling heat exchange value of the k-th scheduling path corresponding to the escape position,
Figure 189728DEST_PATH_IMAGE031
the difference of the fitted escape level of the k-th scheduling path corresponding to the escape position.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, the environment data model is established through the environment data generated by the plurality of monitoring units, so that the pertinence is stronger during heat dissipation, waste is avoided, and secondly, the heat generation condition of the machine room can be predicted through the heat generation model of the machine room generated by the environment data model, so that the actual heat generation point of the machine room can be known, and the follow-up heat dissipation is facilitated to be more targeted. And finally, resource scheduling processing is carried out on the server with overhigh load, and the balance of the heat value and the heat exchange efficiency value generated by scheduling is comprehensively considered, so that the optimal scheduling strategy is selected to carry out temperature adjustment on the machine room of the data center, and the energy consumption of the refrigeration equipment is reduced to the maximum extent.
Drawings
Fig. 1 is a schematic block diagram of a system according to embodiment 1 of the present invention.
FIG. 2 is a schematic flow chart of steps S111-S113 in embodiment 1 of the present invention.
FIG. 3 is a flowchart illustrating steps S211-S214 in embodiment 1 of the present invention.
Detailed Description
The drawings are only for purposes of illustration and are not to be construed as limiting the invention. For a better understanding of the following embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
Example 1
The embodiment provides a multipoint energy consumption detection and dynamic regulation system in a data center machine room, which is used for selecting an optimal heat dissipation strategy and a scheduling strategy to control refrigeration equipment in the machine room to regulate the temperature in the machine room, so that the energy consumption of non-IT equipment (mainly refrigeration equipment) in the machine room is effectively reduced.
As shown in fig. 1, the system includes: the environment monitoring system comprises an environment monitoring module 100, an environment configuration module 200, a heat dissipation fitting module 300 and an execution module 400.
The environment monitoring module 100 includes a plurality of monitoring units 110 disposed at different locations of a data center room;
the monitoring unit 110 is used for monitoring the environmental information change and feeding back the environmental data.
In particular embodiments, the environmental information changes monitored by the monitoring unit 110 include temperature, humidity of the data center room environment, and temperature and humidity outside the data center. The humidity mainly influences heat conduction and a temperature safety threshold, and the temperature directly reflects the heat distribution condition in the machine room. The monitoring units 110 are temperature sensors and are distributed at each dissipation position of the computer room, where the dissipation position refers to a position where the servers commonly dissipate heat. Because the data center server is more, the server is from taking CPU temperature monitoring for monitor server CPU temperature, temperature sensor corresponds to the server setting, and each loss position corresponds a plurality of servers, all sets up temperature sensor at every loss position according to the physical position, also can set up temperature sensor in other suitable positions, in order to obtain the more accurate temperature information in loss position. The plurality of monitoring units 110 are humidity sensors, and are disposed in the machine room to monitor humidity conditions at different positions, and can be disposed outdoors to analyze heat discharge efficiency of the refrigeration equipment.
The environment monitoring module 100 is configured to establish an actual measurement environment data model according to environment data fed back by monitoring units disposed at different positions, and generate a machine room heat generation model according to the actual measurement environment data model. The positions of the monitoring units 110 are known, and the environment monitoring module 100 can construct an environment data model according to the position data of the data center loaded in advance, and meanwhile, the environment data model changes at any time and generates a machine room heat generation model according to the actually measured environment data model.
Specifically, the environment monitoring module 100 includes a model conversion unit 120 configured with a model conversion strategy, and configured to generate a machine room heat generation model according to the measured environment data model.
As shown in fig. 2, the model transformation strategy includes the following steps:
s111, heat generation prediction: historical power information of each server is obtained from a historical load database, and a heat production prediction function is generated according to the historical power information of each server;
in the actual application process, the actual heat production of each server needs to be predicted by continuously repeating the steps and then the current ambient temperature condition is combined, so that an optimal strategy is given, a heat production model of the machine room needs to be continuously generated, the interval time for repeating the steps can be preset, for example, set to be 5 minutes, and the prediction of the actual heat production of the next stage of the machine room is generated every five minutes. The heat generation value is related to three factors, the heat generation condition of each server in the next stage needs to be predicted, the historical heat generation condition of the server needs to be analyzed, the regions possibly corresponding to each server are different, the physical interfaces are different, and the contents of corresponding services are different, so the regularity of the server is different.
The heat production prediction function is generated according to the historical power information of each server, corresponding prediction can be obtained by analyzing according to time, period and waveform characteristics theoretically, but two problems may occur, namely waveform characteristics which cannot be identified, large operation data amount and excessive related samples, and if the samples are subjected to cluster analysis directly, some abnormal data cannot be eliminated, so that the prediction accuracy is caused to be in a problem. Based on this, the heat generation prediction step of step S110 is configured with a cluster analysis sub-strategy, from which a final heat generation prediction function is generated. The cluster analysis sub-strategy is specifically used for acquiring the average power and the average CPU temperature of the servers in the characteristic time periods from the historical power information according to the characteristic time periods, performing cluster analysis on the average power and the average CPU temperature through a cluster analysis algorithm to establish a grouped index for the servers in each characteristic time period, taking the servers with the completely same grouped index as the same server type group, calculating the average heat generation waveform of each server type group according to the historical power information, fitting the average heat generation waveform according to the reference heat generation waveform to obtain a fitted heat generation waveform, and connecting the heat generation fitted waveforms in different characteristic time periods to obtain a heat generation prediction function.
The characteristic time period refers to the date corresponding to the extracted one time period and the extracted one time period, and the cluster analysis sub-strategy comprises the steps of acquiring multiple groups of power waveforms and temperature waveforms of each characteristic time period according to the characteristic time periods, and then analyzing the average power and the CPU temperature of the server in each characteristic time period, so as to obtain the power value and the CPU temperature value of each server, wherein for example, the power of the X1 server in the A1 time period is B1, and the CPU temperature is C1. And after obtaining discrete numerical values of a plurality of servers, carrying out cluster analysis on the discrete point values in all the same time periods. In a specific embodiment, the cluster analysis algorithm may be a DBSCAN cluster analysis algorithm. The above servers can be classified into a plurality of categories by setting the neighborhood radius, and it should be noted that, for X1, the category belonging to a time period a1 is not necessarily the same as the category belonging to a time period a2, so that the category number of each server in each characteristic time period is obtained, for example, 132 time periods exist, then there are 132 numbers for server X1, and similarly, there are 132 numbers for server X2, if the number of server X1 is identical to the number of server X2, then 2 groups are classified into the same type group, it should be noted that although there are more numbers, it is not necessary that all characteristic time periods will have a plurality of groups, for example, for a server, when the throughput is 100 units and the throughput is 0 unit, the generated heat is almost different, so the server temperature will not rise suddenly, and in the divided characteristic time periods, there will be a plurality of such characteristic periods, because the server is used in the morning at night, so that the situations of over-high power and over-high CPU temperature are few, and the statistical power and CPU data are to avoid that one machine room may have a plurality of types of servers, may process information of the same data amount, and generate different heat, so that both the two characteristics need to be acquired, so that the servers of different types are not divided into different characteristic groups in the cluster analysis algorithm. The neighborhood radius is preferably set between 1.2 and 1.5 depending on the effect of real server power and CPU temperature on the heat production results. After the grouping is obtained, the average waveform of all historical data in the group can be obtained to obtain the characteristic sample, so that the number of the samples is increased, the influence of abnormal historical data is reduced, a plurality of servers can share the sample to reduce the operation amount when the data is processed, and meanwhile, the data level difference caused by different types of the servers can be avoided. Therefore, the heat generation change of each server in the future time can be judged according to the samples, the waveform is marked as a heat generation prediction function, and the obtained heat generation prediction function can reflect the change of the heat generated by the servers along with the time.
Preferably, in order to continuously optimize the processing strategy according to the newly input sample, in general, the iteration threshold and the thermal prediction value do not deviate too much, and if the deviation is too much, the model needs to be modified, based on which the environment monitoring module 100 further includes: and the model correcting unit 130 is configured to determine whether a difference between the iterative thermal value and the thermal prediction value at the same time is greater than a first deviation threshold, or whether the difference is greater than a second deviation threshold, if so, correct the corresponding reference thermal waveform to reduce the difference between the iterative thermal value and the thermal prediction value, and if so, correct the neighborhood radius parameter in the corresponding cluster analysis algorithm to generate a new server type group.
S112, heat dissipation iteration step: calculating a theoretical heat value of each dissipation position through a thermal dissipation algorithm, substituting the theoretical heat value into a thermal iteration algorithm to calculate an iterative heat value at the next moment, and acquiring the iterative heat value of a corresponding server in a preset period to establish a heat value data set;
in this step, the heat dissipation algorithm is:
Figure 901332DEST_PATH_IMAGE001
Figure 297678DEST_PATH_IMAGE002
。
wherein the content of the first and second substances,
Figure 436536DEST_PATH_IMAGE003
the theoretical thermal value of the corresponding dissipation location at time t,
Figure 285543DEST_PATH_IMAGE004
the weight parameter for the nth server associated with the escape location is
Figure 800838DEST_PATH_IMAGE032
，
Figure 51691DEST_PATH_IMAGE033
A is a preset first heat exchange parameter, b is a preset second heat exchange parameter, c is a preset third heat exchange parameter,
Figure 627029DEST_PATH_IMAGE007
the temperature values for the escape locations in the environmental data model,
Figure 697753DEST_PATH_IMAGE008
is the indoor ambient temperature value in the ambient data model,
Figure 16739DEST_PATH_IMAGE009
the outdoor environment temperature value in the environment data model.
In the practical application process, the temperature of the data center is also influenced by the environmental conditions at the moment, and the temperature of each server is also influenced by the position of the server. The heat dissipation iteration step 120 is configured with a heat dissipation algorithm and a heat iteration algorithm, calculates a theoretical heat value of each dissipation position through the heat dissipation algorithm, brings the theoretical heat value into the heat iteration algorithm to calculate an iteration heat value at the next moment, and obtains an iteration heat value corresponding to the server in a preset period to establish a heat value data set. The purpose of the thermal dissipation algorithm is to calculate the heat corresponding to the dissipation position, and since the dissipation heat at the dissipation position is determined by the heat production of each server, the measured temperature of the current dissipation position and the overall heat dissipation efficiency, the theoretical heat value at time t can be obtained by calculating the function, wherein the weight corresponding to each server needs to be considered
Figure 122098DEST_PATH_IMAGE014
The longer the distance and the larger the height deviation are, the smaller the heat influence on the dissipation position is, so that the corresponding weight is configured according to the position of each server according to the relation between the distance and the height, and a/b/c are preset heat exchange parameters which are related to the environment humidity, and the larger the humidity is, the smaller the heat exchange parameters areThe corresponding temperature value is converted into a heat operation unit to enable the magnitude of the numerical value to be equal, the heat of the dissipation position is subtracted from the heat generated by the actual server to obtain a first heat difference, the heat diffused to other spaces by the server can obtain a second heat difference, the second heat difference needs to consider opening ventilation, the heat diffusion under the refrigeration condition is not considered, so that the heat effect of each server on the dissipation position can be obtained, the heat of the dissipation position is estimated, the heat generated by the server is iterated through a heat iteration algorithm, as the function of the heat generated by the original server only considers the heat generated by the server processing data, the dissipation heat is considered at the moment, the server can be more accurately predicted by combining environmental factors, the dissipation position is an important position in the heat dissipation step and is also a key position of the model, and as the heat dissipation flow channel mainly circulates through the dissipation position, that is to say, the heat change of loss position has directly influenced heat exchange efficiency, and original function is neglected the heat production of loss position direct monitoring server itself, so through the thermal iteration algorithm, thereby calculate the heat stack of server through the influence of loss position reality, more reliable.
The hot iteration algorithm is
Figure 868337DEST_PATH_IMAGE034
；
Wherein the content of the first and second substances,
Figure 426357DEST_PATH_IMAGE035
the heat generation prediction value at time t1 of the nth server associated with the escape location corresponding to the heat generation prediction function, t 1 And t, a preset iteration time interval is formed, d is a preset fourth heat exchange parameter, e is a preset fifth heat exchange parameter,
Figure 549034DEST_PATH_IMAGE012
the dissipation weight of the dissipation location is
Figure 508900DEST_PATH_IMAGE036
And m is the total number of the escape positions of the system, and the heat production predicted value generated by the heat iteration algorithm is used asIs the iterative heating value.
The heat iteration algorithm is to calculate the heat value of the heat generated by the server in the primitive function, and the value is a predicted value, so the heat value can be directly obtained in the primitive function, then the heat dissipation at the previous moment is calculated, and meanwhile, the inverse influence of the heat at the dissipation position at the previous moment is subtracted, namely, the theoretically required equivalent cold quantity of the heat taken away from the dissipation position is obtained, the higher the heat at the dissipation position is, the closer the dissipation position is to the heat exhaust port, the lower the height of the dissipation position is, the easier the heat is taken away from the dissipation position, and the heat at the dissipation position can be calculated, and the weight is obtained
Figure 160461DEST_PATH_IMAGE037
According to the distribution of the height and the distance of the heat exhaust opening, d/e is also a preset heat exchange parameter and is related to humidity, so that the heat influence on the server from the heat change of the dissipation position can be calculated, the heat of the server is equivalent to the heat of the dissipation position, and then a heat dissipation strategy, such as the setting of a heat dissipation flow channel and the setting of power, only needs to consider the influence on the dissipation position, so that the heat influence value of the server under the effect of no cold is obtained, and the heat dissipation strategy can be guided more accurately.
S113, model generation: a machine room heat production model is generated based on the marking of the locations for each thermal value data set.
The environment configuration module 200 is configured with a heat dissipation policy database 21 and a scheduling database 22.
The heat dissipation strategy database 21 stores heat dissipation strategies, and each heat dissipation strategy corresponds to a heat dissipation channel index and a power limiting heat dissipation value.
The heat dissipation strategy comprises a plurality of heat dissipation instructions, each heat dissipation instruction corresponds to heat dissipation equipment of the data center machine room, and the heat dissipation instructions comprise heat dissipation power, air outlet direction and air outlet speed.
Specifically, the environment configuration module 200 includes:
the heat dissipation configuration unit 210 is configured to obtain control information of each refrigeration device in the machine room, obtain a heat dissipation permission direction from the control information, associate the heat dissipation permission directions of different refrigeration devices to obtain a heat dissipation path, screen the heat dissipation path according to preset path screening conditions to obtain a heat dissipation flow channel, generate a corresponding heat dissipation policy according to the obtained heat dissipation flow channel, and generate a power limit heat dissipation value according to a rated power of the corresponding refrigeration device in the heat dissipation policy.
The path screening conditions comprise cold quantity superposition constraint, airflow short circuit constraint and total vector constraint.
The cold volume superposition constraint is used for judging the coincidence quantity of intersection points between the heat dissipation permission directions in the heat dissipation flow channel. In this constraint, if a plurality of refrigeration devices have the same orientation in a heat dissipation path, which would result in waste of cooling energy, such a heat dissipation path can be eliminated.
The airflow short circuit constraint is used to determine the relative airflow number in the heat dissipation flow channel. In this constraint, if the number of short-circuited air flows in the heat dissipation flow path is large, the heat dissipation efficiency is low, and the heat dissipation path may be deleted from all the heat dissipation paths.
The overall vector constraint is used for judging the direction of the unit vector sum of all heat dissipation permission directions in the heat dissipation flow channel. In this constraint, the heat dissipation flow channel needs to face the air outlet or be vertically upward, so that gas circulation is easily formed, and heat dissipation is facilitated.
The scheduling database 22 is configured with a plurality of scheduling paths, each scheduling path corresponds to a scheduling heat exchange value, and the scheduling paths reflect the data scheduling relationship between servers.
The environment configuration module 200 further includes a scheduling configuration unit 220, configured to configure the configuration of the scheduling database 22, generate a virtual scheduling path according to the communication topology relationship between the servers, and generate a virtual scheduling task, so that the server corresponding to the virtual scheduling path executes the virtual scheduling task to obtain a local processing heat value and a scheduling processing heat value. And generating a corresponding scheduling heat exchange value according to the local processing heat value and the scheduling processing heat value, screening the virtual scheduling paths with the scheduling heat exchange value lower than the scheduling threshold value as the scheduling paths, and storing the scheduling paths into the scheduling database 22.
When the virtual scheduling task is executed in the two devices corresponding to the scheduling path, the CPU temperature of the local device is obtained to obtain the local processing heat value, the scheduling processing heat value is obtained according to the obtained CPU temperature of the local device and the change of the CPU temperature of the device executing the virtual scheduling task, the heat value variable quantity can be obtained by calculating the difference value, if the heat value variable quantity is increased greatly, the scheduling loses the original meaning, therefore, the scheduling path under the condition is not stored, and only the virtual scheduling path which meets the condition that the scheduling heat value is lower than the scheduling threshold value in the virtual scheduling path can be used as the scheduling path and stored in the scheduling database.
The heat dissipation fitting module 300 includes a heat dissipation limit determination unit 310, a heat dissipation fitting unit 320, a schedule fitting unit 330, and a policy selection unit 340.
The heat dissipation limit judgment unit 310 calculates an expected heat dissipation value according to the heat generation model of the machine room, and takes a heat dissipation strategy in which the corresponding power limit heat dissipation value is higher than the expected heat dissipation value as a first strategy group.
The heat dissipation strategy is screened as a first strategy group based on the power limit heat dissipation value, if the maximum power of the heat dissipation strategy still cannot meet the heat dissipation requirement, the heat dissipation strategy is not adopted, the expected heat dissipation value can be calculated through a heat production model of the machine room, because the distributed heat is known, the safe temperature of each server is known, the expected heat dissipation value can be obtained by combining the position relation, so that part of the heat dissipation strategy is eliminated, and the fitting calculation amount is reduced.
The heat dissipation fitting unit 320 fits the heat dissipation channel indexes of the heat dissipation strategies in the first strategy group with the heat generation model of the machine room to calculate the heat exchange efficiency value of each heat dissipation strategy and obtain a corresponding scheduling model.
The dissipation fitting unit 320 is configured with a dissipation fitting algorithm for calculating a fitted dissipation difference value for each dissipation position. And the heat exchange efficiency value of the heat dissipation strategy is the sum of the fitting heat dissipation difference values corresponding to the heat dissipation strategy.
The dissipation fitting algorithm is:
Figure 471357DEST_PATH_IMAGE016
Figure 132145DEST_PATH_IMAGE017
；
wherein T is d Fitting a heat dissipation difference value for the dissipation position; p t Is a preset reference heat dissipation value;
Figure 946517DEST_PATH_IMAGE038
the branch stage number corresponding to the dissipation position;
Figure 34559DEST_PATH_IMAGE039
the heat dissipation flow ratio corresponding to the dissipation position;
Figure 567172DEST_PATH_IMAGE020
the effective heat dissipation distance between the dissipation position and the heat dissipation flow channel;
Figure 766072DEST_PATH_IMAGE040
and configuring a fitting dissipation grade for the dissipation position according to the fitting dissipation difference value of each dissipation position to generate the scheduling model.
The logic of the fit of the dissipation algorithm is as follows, because the heat dissipation flow channel is known, the effective heat dissipation effect of each dissipation position can be obtained according to the heat dissipation flow channel, because the larger the cold input in the flow channel is, the better the heat dissipation effect of the dissipation position passing by the flow channel is, the heat dissipation fit difference of each dissipation position is calculated through the logic, namely the heat dissipation effect of each heat dissipation strategy corresponding to the position is, and the calculation of the heat dissipation fit difference needs to consider two factors, the first is the effective cold supply in the heat dissipation strategy, and the second is the effective heat generation corresponding to the dissipation position. The provision of cold requires consideration of: first, branch stages, that is, branches of the heat dissipation channel, can be obtained if the heat dissipation effect of the main channel is due to the heat dissipation effect of the branches, for example, the number of stages of scores of the main channel is defined as 1, and the number of stages of branches is reduced by 0.1 for each branch of the main channel,whileSlThe heat dissipation flow ratio is such that the reaction gas can easily escape from the flow channel, for example, the flow channel is a confluence of three flow channels, but the air flow efficiency is low if only one output is provided, so that the corresponding heat dissipation flow ratio can be obtained by dividing the number of input flow channels by the number of output flow channels. Although the heat dissipation flow channel passes through the dissipation position, the heat dissipation position is generally rarely covered, so the larger the relative distance between the dissipation position and the heat dissipation flow channel is, the less the cold is affected, so the effective heat dissipation distance needs to be considered, and the effective heat dissipation distance needs to be according to the upper and lower position relation, for example, the heat dissipation flow channel is below the dissipation position, because hot air rises, heat is difficult to be dissipated, if the heat dissipation flow channel is above the dissipation position, the heat dissipation is easier, so the effective heat dissipation distance is faster to reduce every upward distance, and the effective heat dissipation distance is slower to reduce every downward unit distance, and the sum of the effective cold values of the equipment corresponding to the cold input before the dissipation position, wherein the effective cold value is the rated refrigeration power of the equipment divided by the number of nodes, and if the cold of the refrigeration equipment passes through 4 dissipation positions before reaching the dissipation position, the effective cooling power is reduced accordingly. Secondly, the influence of heat to the dissipation position, the original heat of the dissipation position can be known and calculated by superposition of a heat production model at the dissipation position, the heat influence of the dissipation position at any moment can be known, the size and the positive and negative conditions of the position corresponding to the fitting heat dissipation value of the heat dissipation strategy can be calculated, if the fitting heat dissipation value is overlarge, the waste of large cold quantity is explained for changing the position, if the fitting heat dissipation value is a negative value, the position can not be cooled well, the server can not exhaust heat, and through calculating the sum of the fitting heat dissipation values (absolute value), whether the heat dissipation strategy is matched with the corresponding heat production prediction result can be judged, the higher the fitting heat dissipation value is, and the more serious heat waste or loss is explained.
The scheduling fitting unit 330 obtains a corresponding scheduling path by parsing each scheduling model, generates a corresponding scheduling policy, and obtains a scheduling efficiency value of each heat dissipation policy by summing the scheduling heat exchange values corresponding to each scheduling path.
The scheduling fitting unit 330 is configured with a crawler scheduling policy for parsing the scheduling model to obtain a corresponding scheduling path.
As shown in fig. 3, the crawler scheduling policy includes:
s211, topology configuration: taking a corresponding server with the fitting escape grade higher than a preset grade standard in the scheduling model as a starting coordinate, taking a corresponding server with the fitting escape grade lower than the preset grade standard in the scheduling model as an end coordinate, and establishing a crawler scheduling topology through a scheduling path;
s212, crawler configuration: generating a plurality of virtual crawler tasks, and configuring a reference pheromone for each virtual crawler task; when the virtual crawler task passes through an escape position, the escape position obtains corresponding reference pheromones, and each escape position is configured with an information attenuation algorithm for reducing the pheromones of the escape position along with time;
s213, crawler execution: continuously configuring a virtual crawler task for each initial coordinate until a terminal coordinate is reached;
in this step, the transmission node of each virtual crawler task is randomly generated according to the pheromone, and the higher the pheromone of the target node is, the higher the probability of being selected is.
S214, pheromone triggering step: when the pheromone of the escape position reaches the corresponding pheromone triggering threshold, the escape position is taken as an end point coordinate, the escape position with the highest associated priority value is taken as a starting coordinate to determine a scheduling path, the scheduling path is deleted from the crawler scheduling topology as the starting coordinate, and meanwhile, the fitting escape grade as the end point coordinate is increased.
In this step, the information triggering threshold is negatively correlated with the fitting escape level, and the associated priority value corresponding to the escape position is the product of the fitting escape level of the escape position and the completion number of the virtual crawler task.
A crawler algorithm is adopted in the scheduling fitting unit 330, the crawler task is configured to move randomly until a terminal coordinate is found, and a node with higher pheromone is easier to select when the crawler runs, so a server which is used as a middle node is easier to achieve a triggering condition of the pheromone, the pheromone is attenuated according to time, the attenuation factors comprise a scheduling heat exchange value of a scheduling path corresponding to an escape position and a difference value of all scheduling heat exchange values divided by a fitting escape grade to obtain a basis of attenuation of the pheromone, the larger the grade difference is, the easier the improvement of the heat dissipation effect is, the slower the attenuation is, the higher the scheduling heat exchange value is, the more serious the waste is, the faster the attenuation is, therefore, when the pheromone of the escape position reaches a triggering threshold value of the corresponding pheromone, the escape position is used as the terminal coordinate, the escape position with the highest associated priority value is used as a starting coordinate to determine the scheduling path, and deleting the coordinates as the initial coordinates from the crawler scheduling topology, and increasing the fitting escape level as the end point coordinates.
The strategy selection module 340 is configured with a scheduling selection algorithm, the scheduling selection algorithm is configured to calculate a scheduling total power consumption value of each heat dissipation strategy according to the heat exchange efficiency value and the scheduling efficiency value of each heat dissipation strategy, and select the heat dissipation strategy with the smallest scheduling total power consumption value as the target heat dissipation strategy.
Specifically, the scheduling selection algorithm is
Figure 700530DEST_PATH_IMAGE041
In which
Figure 959473DEST_PATH_IMAGE042
In order to schedule the total power consumption value,
Figure 713802DEST_PATH_IMAGE043
is a preset energy consumption proportion parameter,
Figure 981972DEST_PATH_IMAGE044
is the scheduling efficiency value.
The execution module 400 is configured to execute a target heat dissipation policy and a scheduling policy corresponding to the same heat dissipation policy.
The execution module 400 outputs the corresponding heat dissipation policy and scheduling policy, so that when data congestion occurs in the system, scheduling can be performed in time without repeatedly determining a scheduling path, and heat exchange can be performed in advance when the heat is high.
It should be understood that the above-mentioned embodiments of the present invention are only examples for clearly illustrating the technical solutions of the present invention, and are not intended to limit the specific embodiments of the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention claims should be included in the protection scope of the present invention claims.

Claims (11)

1. A multipoint energy consumption detection and dynamic regulation system in a data center machine room is characterized by comprising an environment monitoring module, an environment configuration module, a heat dissipation fitting module and an execution module;
the environment monitoring module comprises a plurality of monitoring units arranged at different positions of a data center machine room;
the monitoring unit is used for monitoring the change of environmental information and feeding back environmental data;
the environment monitoring module is used for establishing an actual measurement environment data model according to environment data fed back by monitoring units at different positions and generating a heat generation model of the machine room according to the actual measurement environment data model;
the environment configuration module is configured with a heat dissipation strategy database and a scheduling database;
the heat dissipation strategy database stores heat dissipation strategies, and each heat dissipation strategy corresponds to a heat dissipation flow channel index and a power limiting heat dissipation value;
the scheduling database is configured with a plurality of scheduling paths, and each scheduling path corresponds to a scheduling heat exchange value;
the heat dissipation fitting module comprises a heat dissipation limitation judging unit, a heat dissipation fitting unit, a scheduling fitting unit and a strategy selecting unit;
the heat dissipation limitation judging unit is used for calculating to obtain an expected heat dissipation value according to the heat production model of the machine room, and taking a heat dissipation strategy with the corresponding power limitation heat dissipation value higher than the expected heat dissipation value as a first strategy group;
the heat dissipation fitting unit is used for fitting the heat dissipation flow channel indexes of the heat dissipation strategies in the first strategy group with the heat production model of the machine room so as to calculate the heat exchange efficiency value of each heat dissipation strategy and obtain a corresponding scheduling model;
the scheduling fitting unit is used for obtaining a corresponding scheduling path according to the analysis of each scheduling model, generating a corresponding scheduling strategy, and summing a scheduling heat exchange value corresponding to each scheduling path to obtain a scheduling efficiency value of each heat dissipation strategy;
the strategy selection module is configured with a scheduling selection algorithm, the scheduling selection algorithm is used for calculating a scheduling total energy consumption value of each heat dissipation strategy according to the heat exchange efficiency value and the scheduling efficiency value of each heat dissipation strategy, and the heat dissipation strategy with the minimum scheduling total energy consumption value is selected as a target heat dissipation strategy;
the execution module is used for executing the target heat dissipation strategy and the scheduling strategy corresponding to the same heat dissipation strategy.
2. The system for detecting and dynamically adjusting the multipoint energy consumption in the data center machine room according to claim 1, wherein the environment monitoring module comprises a model conversion unit configured with a model conversion strategy;
the model conversion strategy is used for generating a machine room heat generation model according to the actually measured environment data model;
the model conversion strategy comprises:
the heat generation prediction step comprises the steps of obtaining historical power information of each server from a historical load database, and generating a heat generation prediction function according to the historical power information of each server;
the heat dissipation iteration step comprises the steps of calculating the theoretical heat value of each dissipation position through a thermal dissipation algorithm, substituting the theoretical heat value into the thermal iteration algorithm to calculate the iteration heat value at the next moment, and acquiring the iteration heat value of a corresponding server in a preset period to establish a heat value data set;
a model generating step including generating the machine room heat production model based on marking a location for each thermal value data set.
3. The system of claim 1, wherein the thermal runaway algorithm is as follows:
Figure 422899DEST_PATH_IMAGE001
Figure 938194DEST_PATH_IMAGE002
；
wherein the content of the first and second substances,
Figure 189047DEST_PATH_IMAGE003
the theoretical thermal value of the corresponding dissipation location at time t,
Figure 498805DEST_PATH_IMAGE004
the weight parameter for the nth server associated with the escape location is
Figure 835109DEST_PATH_IMAGE005
，
Figure 154094DEST_PATH_IMAGE006
A is a preset first heat exchange parameter, b is a preset second heat exchange parameter, c is a preset third heat exchange parameter,
Figure 259454DEST_PATH_IMAGE007
the temperature values for the escape locations in the environmental data model,
Figure 5693DEST_PATH_IMAGE008
is the indoor ambient temperature value in the ambient data model,
Figure 298134DEST_PATH_IMAGE009
an outdoor environment temperature value in the environment data model;
the hot iteration algorithm is
Figure 420811DEST_PATH_IMAGE010
；
Wherein the content of the first and second substances,
Figure 380676DEST_PATH_IMAGE011
the corresponding heat generation prediction function for the nth server associated with the escape location is at t 1 Predicted heat production at time, t 1 And t, a preset iteration time interval is formed, d is a preset fourth heat exchange parameter, e is a preset fifth heat exchange parameter,
Figure 297817DEST_PATH_IMAGE012
the dissipation weight of the dissipation location is
Figure 343133DEST_PATH_IMAGE013
And m is the total number of the dissipation positions of the system, and the heat production predicted value generated by the heat iteration algorithm is used as the iteration heat value.
4. The system of claim 3, wherein the heat generation prediction step further comprises:
the cluster analysis sub-strategy is used for acquiring the average power and the average CPU temperature of the servers in the characteristic time periods from the historical power information according to the characteristic time periods, performing cluster analysis on the average power and the average CPU temperature through a cluster analysis algorithm to establish a grouped index for the servers in each characteristic time period, taking the servers with the completely same grouped index as the same server type group, calculating the average heat generation waveform of each server type group according to the historical power information, fitting the average heat generation waveform according to the reference heat generation waveform to obtain a fitted heat generation waveform, and connecting the heat generation fitted waveforms in different characteristic time periods to obtain a heat generation prediction function.
5. The system of claim 4, wherein the environment monitoring module comprises:
and the model correction unit is used for judging whether the difference value between the iterative heat value and the heat prediction value at the same moment is greater than a first deviation threshold value or not, or whether the difference value is greater than a second deviation threshold value or not, if so, correcting the corresponding reference heat generation waveform to reduce the difference value between the iterative heat value and the heat prediction value, and if so, correcting the neighborhood radius parameter in the corresponding cluster analysis algorithm to generate a new server type group.
6. The system of claim 1, wherein the environment configuration module comprises:
the heat dissipation configuration unit is used for acquiring control information of each refrigeration device in the machine room, acquiring heat dissipation permission directions from the control information, associating the heat dissipation permission directions of different refrigeration devices to acquire heat dissipation paths, screening the heat dissipation paths through preset path screening conditions to acquire heat dissipation channels, generating a corresponding heat dissipation strategy according to the acquired heat dissipation channels, and generating power limit heat dissipation values according to rated powers of the corresponding refrigeration devices in the heat dissipation strategies;
the path screening conditions comprise cold quantity superposition constraint, airflow short circuit constraint and total vector constraint;
the cold quantity superposition constraint is used for judging the coincidence quantity of intersection points between the heat dissipation permission directions in the heat dissipation flow channel;
the airflow short circuit constraint is used for judging the relative airflow quantity in the heat dissipation flow channel;
the total vector constraint is used for judging the direction of the unit vector sum of all heat dissipation permission directions in the heat dissipation flow channel.
7. The system of claim 1, wherein the environment configuration module comprises:
the scheduling configuration unit is used for configuring the configuration of the scheduling database, generating a virtual scheduling path according to the communication topological relation between the servers, and generating a virtual scheduling task so as to enable the server corresponding to the virtual scheduling path to execute the virtual scheduling task and obtain a local processing heat value and a scheduling processing heat value;
and generating a corresponding scheduling heat exchange value according to the local processing heat value and the scheduling processing heat value, screening the virtual scheduling paths with the scheduling heat exchange value lower than the scheduling threshold value as the scheduling paths, and storing the scheduling paths into a scheduling database.
8. The system of claim 3, wherein the heat dissipation fitting unit comprises:
the dissipation fitting algorithm is used for calculating a fitting heat dissipation difference value of each dissipation position, and the heat exchange efficiency value of the heat dissipation strategy is the sum of the fitting heat dissipation difference values corresponding to the heat dissipation strategy;
the dissipation fitting algorithm is:
Figure 3922DEST_PATH_IMAGE014
Figure 818294DEST_PATH_IMAGE015
；
wherein T is d Fitting a heat dissipation difference value for the dissipation position; p t Is a preset reference heat dissipation value;
Figure 171915DEST_PATH_IMAGE016
the branch stage number corresponding to the dissipation position;
Figure 704527DEST_PATH_IMAGE017
the heat dissipation flow ratio corresponding to the dissipation position;
Figure 903428DEST_PATH_IMAGE018
the effective heat dissipation distance between the dissipation position and the heat dissipation flow channel;
Figure 572306DEST_PATH_IMAGE019
and configuring a fitting dissipation grade for the dissipation position according to the fitting dissipation difference value of each dissipation position to generate the scheduling model.
9. The system according to claim 1, wherein the scheduling selection algorithm is
Figure 96829DEST_PATH_IMAGE020
Wherein
Figure 116737DEST_PATH_IMAGE021
In order to schedule the total power consumption value,
Figure 119328DEST_PATH_IMAGE022
is a preset energy consumption proportion parameter,
Figure 642713DEST_PATH_IMAGE023
is the scheduling efficiency value.
10. The system for multi-point energy consumption detection and dynamic adjustment in a data center room of claim 8,
the scheduling fitting unit is configured with a crawler scheduling strategy and used for analyzing a scheduling model to obtain a corresponding scheduling path;
the crawler scheduling policy comprises:
a topology configuration step, which is used for establishing a crawler scheduling topology through a scheduling path by taking a corresponding server in the scheduling model, of which the fitting escape level is higher than a preset level standard, as an initial coordinate and taking a corresponding server in the scheduling model, of which the fitting escape level is lower than the preset level standard, as an end coordinate;
a crawler configuration step, which is used for generating a plurality of virtual crawler tasks and configuring a reference pheromone for each virtual crawler task;
when the virtual crawler task passes through an escape position, the escape position obtains corresponding reference pheromones, and each escape position is configured with an information attenuation algorithm for reducing the pheromones of the escape position along with time;
a crawler executing step, which is used for continuously configuring a virtual crawler task for each initial coordinate until reaching the end coordinate;
and a pheromone triggering step, wherein when the pheromone of the escape position reaches the corresponding pheromone triggering threshold value, the escape position is taken as an end point coordinate, the escape position with the highest associated priority value is taken as a starting coordinate to determine a scheduling path, the scheduling path is taken as the starting coordinate to be deleted from the crawler scheduling topology, and meanwhile, the fitting escape grade taken as the end point coordinate is increased.
11. The system according to claim 10, wherein the information attenuation algorithm is
Figure 338137DEST_PATH_IMAGE024
；
Wherein
Figure 579763DEST_PATH_IMAGE025
Is the pheromone at the last moment,
Figure 120465DEST_PATH_IMAGE026
is the pheromone at the next time instant,
Figure 29515DEST_PATH_IMAGE027
in order to be a pre-set attenuation parameter,
Figure 630261DEST_PATH_IMAGE028
the scheduling heat exchange value of the k-th scheduling path corresponding to the escape position,
Figure 624762DEST_PATH_IMAGE029
the difference of the fitted escape level of the k-th scheduling path corresponding to the escape position.
CN202210814549.9A 2022-07-12 2022-07-12 Multi-point energy consumption detection and dynamic regulation system in data center machine room Active CN114901057B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210814549.9A CN114901057B (en) 2022-07-12 2022-07-12 Multi-point energy consumption detection and dynamic regulation system in data center machine room

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210814549.9A CN114901057B (en) 2022-07-12 2022-07-12 Multi-point energy consumption detection and dynamic regulation system in data center machine room

Publications (2)

Publication Number Publication Date
CN114901057A true CN114901057A (en) 2022-08-12
CN114901057B CN114901057B (en) 2022-09-27

Family

ID=82729254

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210814549.9A Active CN114901057B (en) 2022-07-12 2022-07-12 Multi-point energy consumption detection and dynamic regulation system in data center machine room

Country Status (1)

Country Link
CN (1) CN114901057B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116471821A (en) * 2023-06-19 2023-07-21 广州豪特节能环保科技股份有限公司 Method, system, equipment and medium for dynamic control energy conservation of data center
CN116820214A (en) * 2023-08-30 2023-09-29 深圳智锐通科技有限公司 Intelligent regulation and control method and system for ventilation and heat dissipation of server

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100076607A1 (en) * 2008-08-08 2010-03-25 Osman Ahmed Data center thermal performance optimization using distributed cooling systems
CN109714400A (en) * 2018-12-12 2019-05-03 华南理工大学 A kind of energy optimization resource scheduling system and its method towards container cluster
CN110705062A (en) * 2019-09-20 2020-01-17 苏州智博汇能电子科技股份有限公司 Cabinet energy consumption remote statistical metering method based on 5G
CN111795488A (en) * 2020-07-10 2020-10-20 哈尔滨工程大学 Intelligent temperature regulation and control system and method for distributed machine room
CN113551373A (en) * 2021-07-19 2021-10-26 江苏中堃数据技术有限公司 Data center air conditioner energy-saving control method based on federal reinforcement learning
CN113778215A (en) * 2021-04-28 2021-12-10 龙坤(无锡)智慧科技有限公司 Method for realizing data center PUE prediction and consumption reduction strategy based on big data

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100076607A1 (en) * 2008-08-08 2010-03-25 Osman Ahmed Data center thermal performance optimization using distributed cooling systems
CN109714400A (en) * 2018-12-12 2019-05-03 华南理工大学 A kind of energy optimization resource scheduling system and its method towards container cluster
CN110705062A (en) * 2019-09-20 2020-01-17 苏州智博汇能电子科技股份有限公司 Cabinet energy consumption remote statistical metering method based on 5G
CN111795488A (en) * 2020-07-10 2020-10-20 哈尔滨工程大学 Intelligent temperature regulation and control system and method for distributed machine room
CN113778215A (en) * 2021-04-28 2021-12-10 龙坤(无锡)智慧科技有限公司 Method for realizing data center PUE prediction and consumption reduction strategy based on big data
CN113551373A (en) * 2021-07-19 2021-10-26 江苏中堃数据技术有限公司 Data center air conditioner energy-saving control method based on federal reinforcement learning

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116471821A (en) * 2023-06-19 2023-07-21 广州豪特节能环保科技股份有限公司 Method, system, equipment and medium for dynamic control energy conservation of data center
CN116471821B (en) * 2023-06-19 2023-09-15 广州豪特节能环保科技股份有限公司 Method, system, equipment and medium for dynamic control energy conservation of data center
CN116820214A (en) * 2023-08-30 2023-09-29 深圳智锐通科技有限公司 Intelligent regulation and control method and system for ventilation and heat dissipation of server
CN116820214B (en) * 2023-08-30 2024-01-05 深圳智锐通科技有限公司 Intelligent regulation and control method and system for ventilation and heat dissipation of server

Also Published As

Publication number Publication date
CN114901057B (en) 2022-09-27

Similar Documents

Publication Publication Date Title
CN114901057B (en) Multi-point energy consumption detection and dynamic regulation system in data center machine room
CN109800066B (en) Energy-saving scheduling method and system for data center
CN104698843B (en) A kind of data center's energy-saving control method based on Model Predictive Control
Andradóttir et al. Dynamic server allocation for queueing networks with flexible servers
KR101867487B1 (en) System and method of migration based fuzzy in cloud environment
CN103218678B (en) The operation management method of information processing system
KR101754536B1 (en) Method and apparatus for optimum control of refrigerator using building energy management system
CN115095907B (en) Intelligent heat supply energy-saving regulation and control method and system based on deep learning and storage medium
CN107133286B (en) Method and system for generating and analyzing three-dimensional graph of temperature parameter distribution field of machine room
Wei et al. Research on cloud design resources scheduling based on genetic algorithm
CN107543277A (en) A kind of temprature control method, apparatus and system
CN205919586U (en) Temperature control device and temperature control system
Zheng et al. FCTcon: Dynamic control of flow completion time in data center networks for power efficiency
CN113225994B (en) Intelligent air conditioner control method facing data center
CN111461925A (en) Power distribution network fault analysis method, device, equipment and storage medium
CN113853096A (en) Internet of things linkage-based IDC machine room temperature regulation and control method and system
CN116991070A (en) Comprehensive performance intelligent management method and system for multi-equipment terminal based on Internet of things
CN114980666B (en) Control method and system for liquid cooling heat dissipation system of data center
CN116578134A (en) Universal base station temperature control method and system based on reinforcement learning
CN107911763B (en) Intelligent power distribution and utilization communication network EPON network planning method based on QoS
CN113094149B (en) Data center virtual machine placement method, system, medium and equipment
CN113965445B (en) Positioning method and device for quality difference root cause, computer equipment and storage medium
CN114325141B (en) Base station electric quantity abnormality monitoring method and device and readable storage medium
CN110955728A (en) Power consumption data transmission method, server and storage medium
CN112241812A (en) Low-voltage distribution network topology identification method based on unilateral optimization and genetic algorithm cooperation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant