CN117707741A - Energy consumption balanced scheduling method and system based on spatial position - Google Patents
Energy consumption balanced scheduling method and system based on spatial position Download PDFInfo
- Publication number
- CN117707741A CN117707741A CN202410159989.4A CN202410159989A CN117707741A CN 117707741 A CN117707741 A CN 117707741A CN 202410159989 A CN202410159989 A CN 202410159989A CN 117707741 A CN117707741 A CN 117707741A
- Authority
- CN
- China
- Prior art keywords
- power consumption
- scheduling
- cabinet
- node
- job
- 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
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000013507 mapping Methods 0.000 claims abstract description 16
- 230000002159 abnormal effect Effects 0.000 claims abstract description 13
- 238000012216 screening Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000005012 migration Effects 0.000 claims description 2
- 238000013508 migration Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
Classifications
-
- 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
Landscapes
- Power Sources (AREA)
Abstract
The invention relates to the technical field of job scheduling of high-performance calculation, and discloses an energy consumption balanced scheduling method and system based on a spatial position; wherein the method comprises the following steps: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet; mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption; screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy; scheduling the job to be scheduled to one or more screened computing nodes; and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption. The energy consumption of the whole cluster can be effectively balanced, and the generation of hot spots in a machine room is avoided.
Description
Technical Field
The invention relates to the technical field of job scheduling of high-performance calculation, in particular to an energy consumption balanced scheduling method and system based on spatial positions.
Background
The statements in this section merely relate to the background of the present disclosure and may not necessarily constitute prior art.
High performance computing (High Performance Computing, HPC) technology is widely used in computationally intensive applications such as energy, biology, weather, scientific research, geological exploration, etc., the scale and complexity of computing cluster systems are continually increasing, and energy consumption problems are increasingly prominent. The job scheduling is used as a key link of cluster system operation, and the strategy has great influence on the energy consumption of the clusters.
The cluster job scheduling system is responsible for carrying out job and resource allocation on limited system resources according to application programs submitted by users. The policies and algorithms of the job scheduling system have a significant impact on the performance and energy efficiency of high performance computing HPC clusters. Traditional job scheduling strategies mainly focus on performance indexes such as resource utilization rate, job throughput rate, job waiting time and the like. In order to optimize the energy consumption of the computing system, researchers begin to explore how to perform energy-saving scheduling and load balancing on high-performance computing HPC cluster task scheduling on the premise of guaranteeing various performance requirements in the study of job scheduling technology, so as to achieve the rationality and effectiveness of scheduling. Currently, many studies utilize machine learning methods to construct energy consumption prediction models and guide scheduling decisions based on the characteristics of the job and the state of the system. In addition, a plurality of targets such as performance, energy consumption, reliability and the like of the operation are considered in the operation scheduling algorithm based on multi-target optimization, and meta-heuristic methods such as a genetic algorithm, a particle swarm algorithm, a simulated annealing algorithm and the like are used to minimize the total energy consumption and the peak energy consumption of the operation. However, these researches ignore the power consumption difference of nodes at different positions of the high-performance computing HPC machine room and the energy consumption characteristics of the whole machine room, and cannot solve the problem of uneven cluster energy consumption distribution. Therefore, how to balance the overall power consumption of the high-performance computing HPC cluster in the job scheduling process, and avoid the generation of hot spots, is the key point of current research.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides an energy consumption balanced scheduling method and system based on space positions; by adding the attribute of the spatial position of the physical node into the power consumption information of the computing node, when the HPC cluster performs job scheduling, a scheduling strategy is adopted to select a position area with lower current power consumption to distribute the computing node for the job in the job queue according to the power consumption distribution of each position in the current machine room, so that the whole energy consumption of the cluster can be effectively balanced, and the generation of hot spots in the machine room can be avoided.
In one aspect, an energy consumption balance scheduling method based on spatial location is provided, including: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet; mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption; screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy; scheduling the job to be scheduled to one or more screened computing nodes; and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
In another aspect, an energy consumption balance scheduling system based on spatial location is provided, including: an acquisition module configured to: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet; a mapping module configured to: mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption; a screening module configured to: screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy; a scheduling module configured to: scheduling the job to be scheduled to one or more screened computing nodes; a tuning module configured to: and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
The technical scheme has the following advantages or beneficial effects: according to the invention, through the position-power consumption data characteristics of the cabinet and the nodes in the HPC machine room in the two-dimensional space dimension, the computing node with lower local position power consumption is selected to allocate resources for the job to be scheduled. In addition, a secondary scheduling optimization strategy is adopted for the abnormal situation of serious unbalance of energy consumption in the space position. By using the scheduling strategy of the invention, workload can be reasonably distributed on the nodes, and regional cabinet heat aggregation caused by overhigh power consumption of the local server is avoided, so that the whole energy of the cluster is uniformly distributed, and the hot spot problem of the HPC cluster is effectively solved. Meanwhile, the energy efficiency and the performance of the HPC cluster are considered, and the resource utilization rate and the stability of the cluster are improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
Fig. 1 is a flow chart of a method according to a first embodiment.
Fig. 2 is a plan view of a machine room cluster system according to the first embodiment.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
In a first embodiment, as shown in fig. 1, the present embodiment provides an energy consumption balancing scheduling method based on spatial location, including: s101: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet; s102: mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption; s103: screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy; s104: scheduling the job to be scheduled to one or more screened computing nodes; s105: and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
Further, S101: the method for acquiring the power consumption data and the space position data of the cabinet in the high-performance computer room and the power consumption data and the space position data of the computing nodes in the cabinet specifically comprises the following steps: numbering the cabinets according to columns, mapping a two-dimensional plane of the machine room into a cabinet position matrix form, wherein each element of the cabinet position matrix corresponds to a unique position identifier of the cabinet one by one; the location of a computing node in a rack is represented using the location of the rack in which the node is located.
Specifically, the power consumption of the cabinet level and the node level is periodically collected by the sensor, and the time interval for collecting the data can be specified according to the needs, and in this embodiment, the time interval for collecting the data is consistent with the time interval for scheduling each round of cluster operation.
As shown in fig. 2, the cabinets are numbered in rows, and the cabinets are shared in the machine roomColumn cabinets, each column having +>A plurality of cabinets; mapping the two-dimensional plane of the machine room into a matrix form, wherein each element of the matrix corresponds to a unique position identifier of the cabinet one by one, and the position information of the nodes is represented by the position of the cabinet where the nodes are located. The space position data only needs to be collected once, and the cabinet or the node is collected again after being updated and maintained.
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Representing a cabinet position matrix>Representing the number of columns of cabinets in the machine room>Representing the number of cabinets in each row of cabinets, mapping the two-dimensional plane of the machine room to +.>Is a matrix form of (a); in addition respectively useRepresents 1 st to->And (5) arranging cabinets.
In addition, the location information of the nodes uses the location representation of the cabinets in which they are located to number the nodes within each cabinet. Such asNodes in the cabinet are all +.>Position (S)>The node position set in the cabinet is。
Further, the step S102: mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption, wherein the method comprises the following steps: according to the power consumption of the cabinetAnd position->Data, integrating the two data, thereby aggregating the two data in two-dimensional space dimension to obtain position and power consumption +.>Relation, position and power consumption matrix->And position->The elements of the matrix are in one-to-one correspondence.
The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Representing a location and power consumption matrix.
In addition, the power consumption and the position data of the nodes are aggregated, and the position information of the nodes is represented by the position of the cabinet where the nodes are located, so that the result of aggregation is the corresponding relation between the position numbers of the nodes and the power consumption of the nodes. Such asNode set in cabinetThe power consumption value set is +.>。
Further, the step S103: screening the computing nodes which can be used for executing the job tasks according to the cabinet position, the power consumption relation and the job scheduling strategy, wherein the method comprises the following steps: s103-1: calculating the power consumption deviation degree of each row of cabinets in the position matrix, and selecting a row of cabinets with the minimum power consumption deviation degree; the power consumption deviation degree refers to: the power consumption of the column cabinet deviates from the average power consumption degree of the cluster; s103-2: selecting all single cabinets with power consumption smaller than the average power consumption of the cabinets in the current column from a column of cabinets with the smallest power consumption deviation degree; s103-3: and selecting the computing nodes with the power consumption smaller than the set node power consumption threshold from all the selected single cabinets as the object of job scheduling.
Further, the S103-1: calculating the power consumption deviation degree of each row of cabinets in the position matrix, and selecting a row of cabinets with the minimum power consumption deviation degree; the power consumption deviation degree refers to: the power consumption of the column cabinet, the degree of deviation from the average power consumption of the cluster, comprises:power consumption deviation degree of column cabinet>Expressed as: />The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>For the number of cabinets of each column, +.>Is->Power consumption value of location, ">Average power consumption for the cluster.
Cluster average power consumptionExpressed as: />The method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>For the number of columns of cabinets in the machine room>For the number of cabinets of each column, +.>Is->Power consumption value of the location.
Selecting the minimum power consumption deviation degreeColumn cabinet, denoted: />The method comprises the steps of carrying out a first treatment on the surface of the Wherein,respectively->Power consumption of columns deviates.
Specifically, the degree of deviation of power consumptionDefined as the degree to which column cabinet power consumption deviates from cluster average power consumption +.>I.e. the average power consumption of all cabinets. Degree of Power consumption deviation->The greater the number of (2) and greater than 0, the greater the degree to which the power consumption of the row of cabinets deviates upwardly from the average power consumption of the cluster, i.e., the higher the power consumption of the row of cabinets is above the average level; />The number of (2) is equal to 0, representing that the power consumption of the row of cabinets is equal to the average level; />The smaller and smaller the value of 0, the greater the degree to which the power consumption representing the row of cabinets deviates downward from the cluster average power consumption, i.e., the lower the power consumption of the row of cabinets is below the average level. And selecting a row of cabinets with the smallest power consumption deviation degree, namely selecting a row of cabinets with the lowest relative power consumption level.
Further, the step S103-2: selecting all single cabinets with power consumption smaller than the average power consumption of the cabinets in the current column from a row of cabinets with the smallest power consumption deviation degree, wherein the method comprises the following steps: calculation ofPower consumption mean value of column cabinet->Expressed as:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is->The number of cabinets in the column cabinet,/->Is->Power consumption value of the location.
SelectingColumn power consumption is less than->Is added to the position set。
Further, the step S103-3: selecting a computing node with the power consumption smaller than a set node power consumption threshold from all the selected single cabinets as an object of job scheduling, wherein the method specifically comprises the following steps: from a collection of positionsCounting the node positions in the cabinets at all positions to obtain a node position set +.>The method comprises the steps of carrying out a first treatment on the surface of the Each node has a unique node name; according to the position set of the node->Positioning to the node and obtaining the running state of the node to obtain a node position set +.>Is a list of node names.
Node power consumption thresholdThe node reaches a power consumption value in a full load state, and for the node with power consumption exceeding a threshold value, the node is not allocated to a new job; according to node position set->And the relation between the position and the power consumption, obtaining a node position set which can be used for executing the job task>Position set +.>Is a list of node names.
Further, the step S104: scheduling the job to be scheduled to the screened one or more computing nodes, including: when each round of scheduling is performed, the jobs waiting in line are sequentially popped out, and a position set is randomly selected according to the job requirementsIs assigned to the current job: before distributing the nodes, judging whether the nodes meet the resources required by the operation or not; if the node meets the resources required by the job, the selected node is allocated for the current job; otherwise, the current job is skipped, and nodes are allocated for the next job.
Further, the step S105: after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption, including: after each round of scheduling is finished, traversing the cabinets at each position, and judging whether cabinets with power consumption exceeding a threshold value exist or not; if yes, performing secondary scheduling and optimizing; otherwise, the secondary scheduling is not performed.
Specifically, the abnormal energy consumption means that the power consumption of the cabinet exceeds a set threshold, and the difference between the power consumption value of the cabinet at the current position and the power consumption value of the cabinet at any other position is greater than the set threshold.
Specifically, scheduling optimization refers to migrating a job of a computing node whose power consumption exceeds a set threshold to other computing nodes for further execution.
Further, the step S105: after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption, including: s105-1: traversing a location and power consumption matrixDetermining that the power consumption exceeds a set cabinet power consumption threshold +.>Is a position of (2); associating each element of the matrix with a set cabinet power consumption threshold +.>Comparing, determining that the power consumption is greater than the set cabinet power consumption threshold +.>Is arranged at the cabinet position; s105-2: for each power consumption greater than +.>The nodes in the cabinet are ordered in descending order according to the power consumption value; after ordering the nodes in the cabinet, a node sequence is obtained; if the cabinet position determined in step S105-1 is not unique, corresponding to a plurality of node sequences; s105-3: for each node sequence, choose front +.>% nodes; s105-4: for each node selected, the job running on it is migrated to the location set +.>Is a node in (a): during the migration of the job, the position set +.>Randomly selecting the nodes in the list, and judging whether the selected nodes meet the resources required by the operation; if the node meets the resources required by the operation, the current operation is migrated; otherwise, the current operation is abandoned.
Specifically, S105-3: large-scale secondary scheduling reduces the operating efficiency of the clustered system, and thereforeThe value of (2) should not be too large. In this embodiment, <' > a->Set to 5.
The second embodiment provides an energy consumption balance scheduling system based on spatial location, including: an acquisition module configured to: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet; a mapping module configured to: mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption; a screening module configured to: screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy; a scheduling module configured to: scheduling the job to be scheduled to one or more screened computing nodes; a tuning module configured to: and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
Here, it should be noted that the above-mentioned obtaining module, mapping module, screening module, scheduling module and tuning module correspond to steps S101 to S105 in the first embodiment, and the above-mentioned modules are the same as examples and application scenarios implemented by the corresponding steps, but are not limited to the disclosure of the first embodiment. It should be noted that the modules described above may be implemented as part of a system in a computer system, such as a set of computer-executable instructions. The foregoing embodiments are directed to various embodiments, and details of one embodiment may be found in the related description of another embodiment.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The energy consumption balanced scheduling method based on the space position is characterized by comprising the following steps of:
acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet;
mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption;
screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy;
scheduling the job to be scheduled to one or more screened computing nodes;
and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
2. The spatial location based energy consumption balanced dispatching method of claim 1, wherein the acquiring the power consumption data and the spatial location data of the cabinet in the high performance computer room, and the power consumption data and the spatial location data of the computing node in the cabinet specifically comprises:
numbering the cabinets according to columns, mapping a two-dimensional plane of the machine room into a cabinet position matrix form, wherein each element of the cabinet position matrix corresponds to a unique position identifier of the cabinet one by one; the location of a computing node in a rack is represented using the location of the rack in which the node is located.
3. The spatial location based energy consumption balanced scheduling method according to claim 1, wherein the screening of the computing nodes available for executing the job task according to the cabinet location and power consumption relationship and the job scheduling policy comprises:
calculating the power consumption deviation degree of each row of cabinets in the position matrix, and selecting a row of cabinets with the minimum power consumption deviation degree; the power consumption deviation degree refers to: the power consumption of the column cabinet deviates from the average power consumption degree of the cluster;
selecting all single cabinets with power consumption smaller than the average power consumption of the cabinets in the current column from a column of cabinets with the smallest power consumption deviation degree;
and selecting the computing nodes with the power consumption smaller than the set node power consumption threshold from all the selected single cabinets as the object of job scheduling.
4. The spatial location-based energy consumption balanced scheduling method according to claim 3, wherein the power consumption deviation degree of each row of cabinets in the location matrix is calculated, and a row of cabinets with the minimum power consumption deviation degree is selected; the power consumption deviation degree refers to: the power consumption of the column cabinet, the degree of deviation from the average power consumption of the cluster, comprises:
power consumption deviation degree of column cabinet>The calculation formula of (2) is as follows:
；
wherein,for the number of cabinets of each column, +.>Is->Power consumption value of location, ">Average power consumption of the cluster;
cluster average power consumptionThe calculation formula of (2) is as follows:
；
wherein,for the number of columns of cabinets in the machine room>For the number of cabinets of each column, +.>Is->A power consumption value for the location;
selecting the minimum power consumption deviation degreeThe column cabinet has the following formula:
；
wherein,respectively->Power consumption of columns deviates.
5. The spatial location based energy consumption balanced scheduling method according to claim 3, wherein selecting all single cabinets with power consumption smaller than the average power consumption of the current cabinet in a row of cabinets with minimum power consumption deviation comprises:
calculation ofWith cabinets in columnsMean value of power consumption->The formula is as follows:
；
wherein,is->The number of cabinets in the column cabinet,/->Is->A power consumption value for the location;
selectingColumn power consumption is less than->Is added to the set of locations>。
6. The spatial location-based energy consumption balanced scheduling method according to claim 3, wherein in all selected single cabinets, a computing node with a computing node power consumption smaller than a set node power consumption threshold is selected as an object of job scheduling, and specifically comprises:
from a collection of positionsCounting the positions of nodes in the cabinets at all positions to obtain the nodesPoint location setThe method comprises the steps of carrying out a first treatment on the surface of the Each node has a unique node name; according to the position set of the node->Positioning to the node and obtaining the running state of the node to obtain a node position set +.>Is a list of node names;
node power consumption thresholdThe node reaches a power consumption value in a full load state, and for the node with power consumption exceeding a threshold value, the node is not allocated to a new job; according to node position set->And the relation between the position and the power consumption, obtaining a node position set which can be used for executing the job task>Position set +.>Is a list of node names.
7. The spatial location based energy consumption balanced scheduling method according to claim 1, wherein scheduling the job to be scheduled to the one or more screened computing nodes comprises:
when each round of scheduling is performed, the jobs waiting in line are sequentially popped out, and a position set is randomly selected according to the job requirementsIs assigned to the current job:
before distributing the nodes, judging whether the nodes meet the resources required by the operation or not; if the node meets the resources required by the job, the selected node is allocated for the current job; otherwise, the current job is skipped, and nodes are allocated for the next job.
8. The spatial location-based energy consumption balance scheduling method of claim 1, wherein scheduling and optimizing the computing node with abnormal energy consumption after each round of scheduling is finished comprises:
after each round of scheduling is finished, traversing the cabinets at each position, and judging whether cabinets with power consumption exceeding a threshold value exist or not; if yes, performing secondary scheduling and optimizing; otherwise, the secondary scheduling is not performed.
9. The spatial location-based energy consumption balance scheduling method of claim 1, wherein scheduling and optimizing the computing node with abnormal energy consumption after each round of scheduling is finished comprises:
traversing a location and power consumption matrixDetermining that the power consumption exceeds a set cabinet power consumption threshold +.>Is a position of (2); associating each element of the matrix with a set cabinet power consumption threshold +.>Comparing, determining that the power consumption is greater than the set cabinet power consumption threshold +.>Is arranged at the cabinet position;
for each power consumption greater thanThe nodes in the cabinet are ordered in descending order according to the power consumption value; after ordering the nodes in the cabinet, a node order is obtainedA column; if the determined cabinet position is not unique, corresponding to a plurality of node sequences correspondingly;
for each node sequence, select the front% nodes;
for each selected node, the job running thereon is migrated to the location setIs a node in (a): during the migration of the job, the position set +.>Randomly selecting the nodes in the list, and judging whether the selected nodes meet the resources required by the operation; if the node meets the resources required by the operation, the current operation is migrated; otherwise, the current operation is abandoned.
10. The energy consumption balanced scheduling system based on the space position is characterized by comprising the following components:
an acquisition module configured to: acquiring power consumption data and space position data of a cabinet in a high-performance computer room, and power consumption data and space position data of a computing node in the cabinet;
a mapping module configured to: mapping the acquired data to a two-dimensional space dimension to obtain a relation between the cabinet position and the power consumption;
a screening module configured to: screening computing nodes which can be used for executing the job task according to the cabinet position, the power consumption relation and the job scheduling strategy;
a scheduling module configured to: scheduling the job to be scheduled to one or more screened computing nodes;
a tuning module configured to: and after each round of scheduling is finished, scheduling and optimizing the computing nodes with abnormal energy consumption.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410159989.4A CN117707741B (en) | 2024-02-05 | 2024-02-05 | Energy consumption balanced scheduling method and system based on spatial position |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410159989.4A CN117707741B (en) | 2024-02-05 | 2024-02-05 | Energy consumption balanced scheduling method and system based on spatial position |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117707741A true CN117707741A (en) | 2024-03-15 |
| CN117707741B CN117707741B (en) | 2024-05-24 |
Family
ID=90151928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410159989.4A Active CN117707741B (en) | 2024-02-05 | 2024-02-05 | Energy consumption balanced scheduling method and system based on spatial position |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117707741B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080046895A1 (en) * | 2006-08-15 | 2008-02-21 | International Business Machines Corporation | Affinity dispatching load balancer with precise CPU consumption data |
| CN102622273A (en) * | 2012-02-23 | 2012-08-01 | 中国人民解放军国防科学技术大学 | Self-learning load prediction based cluster on-demand starting method |
| US20140046974A1 (en) * | 2012-08-13 | 2014-02-13 | Hulu Llc | Job Dispatcher of Transcoding Jobs for Media Programs |
| CN106844051A (en) * | 2017-01-19 | 2017-06-13 | 河海大学 | The loading commissions migration algorithm of optimised power consumption in a kind of edge calculations environment |
| CN108521325A (en) * | 2018-03-27 | 2018-09-11 | 林喆昊 | A kind of anti-side-channel attack algorithm suitable for system data Life cycle |
| CN114816699A (en) * | 2022-04-11 | 2022-07-29 | 山东省计算中心（国家超级计算济南中心） | Data center job scheduling method and system based on temperature prediction |
| CN116880968A (en) * | 2023-06-05 | 2023-10-13 | 超聚变数字技术有限公司 | Job scheduling method and scheduling system |
-
2024
- 2024-02-05 CN CN202410159989.4A patent/CN117707741B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080046895A1 (en) * | 2006-08-15 | 2008-02-21 | International Business Machines Corporation | Affinity dispatching load balancer with precise CPU consumption data |
| CN102622273A (en) * | 2012-02-23 | 2012-08-01 | 中国人民解放军国防科学技术大学 | Self-learning load prediction based cluster on-demand starting method |
| US20140046974A1 (en) * | 2012-08-13 | 2014-02-13 | Hulu Llc | Job Dispatcher of Transcoding Jobs for Media Programs |
| CN106844051A (en) * | 2017-01-19 | 2017-06-13 | 河海大学 | The loading commissions migration algorithm of optimised power consumption in a kind of edge calculations environment |
| CN108521325A (en) * | 2018-03-27 | 2018-09-11 | 林喆昊 | A kind of anti-side-channel attack algorithm suitable for system data Life cycle |
| CN114816699A (en) * | 2022-04-11 | 2022-07-29 | 山东省计算中心（国家超级计算济南中心） | Data center job scheduling method and system based on temperature prediction |
| CN116880968A (en) * | 2023-06-05 | 2023-10-13 | 超聚变数字技术有限公司 | Job scheduling method and scheduling system |
Non-Patent Citations (2)
| Title |
|---|
| 李谦;伍卫国;曹莹方;王蕾;: "面向能耗优化的云渲染***任务调度策略", 西安交通大学学报, no. 02, 29 February 2016 (2016-02-29) * |
| 顾文斌等: "基于改进MLEA算法的多目标离散车间节能调度", 计算机集成制造***, 28 June 2023 (2023-06-28) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117707741B (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | Energy and migration cost-aware dynamic virtual machine consolidation in heterogeneous cloud datacenters | |
| CN101916209B (en) | Cluster task resource allocation method for multi-core processor | |
| Chen et al. | How does the workload look like in production cloud? analysis and clustering of workloads on alibaba cluster trace | |
| CN103345514B (en) | Streaming data processing method under big data environment | |
| EP2212806B1 (en) | Allocation of resources for concurrent query execution via adaptive segmentation | |
| CN107357652B (en) | Cloud computing task scheduling method based on segmentation ordering and standard deviation adjustment factor | |
| CN102932279A (en) | Multidimensional resource scheduling system and method for cloud environment data center | |
| CN105245617A (en) | Container-based server resource supply method | |
| CN102984137A (en) | Multi-target server scheduling method based on multi-target genetic algorithm | |
| Phan et al. | Evolutionary multiobjective optimization for green clouds | |
| Zhao et al. | A data placement strategy for data-intensive scientific workflows in cloud | |
| CN104050042A (en) | Resource allocation method and resource allocation device for ETL (Extraction-Transformation-Loading) jobs | |
| CN101539800A (en) | Low-power job schedule calculating method | |
| Li et al. | An effective scheduling strategy based on hypergraph partition in geographically distributed datacenters | |
| CN103699433A (en) | Method and system for performing dynamic adjustment on number of tasks in Hadoop platform | |
| CN105975345A (en) | Video frame data dynamic equilibrium memory management method based on distributed memory | |
| CN106371924A (en) | Task scheduling method for maximizing MapReduce cluster energy consumption | |
| Fan et al. | Intelligent resource scheduling based on locality principle in data center networks | |
| Delavar et al. | A synthetic heuristic algorithm for independent task scheduling in cloud systems | |
| CN110362379A (en) | Based on the dispatching method of virtual machine for improving ant group algorithm | |
| Zhao et al. | Heuristic data placement for data-intensive applications in heterogeneous cloud | |
| Li et al. | An energy-aware scheduling algorithm for big data applications in Spark | |
| Wang et al. | Dependency-aware network adaptive scheduling of data-intensive parallel jobs | |
| CN115220900A (en) | Energy-saving scheduling method and system based on operation power consumption prediction | |
| CN117707741B (en) | Energy consumption balanced scheduling method and system based on spatial position |
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 |