CN112887345A - Node load balancing scheduling method for edge computing environment - Google Patents
Node load balancing scheduling method for edge computing environment Download PDFInfo
- Publication number
- CN112887345A CN112887345A CN201911196737.4A CN201911196737A CN112887345A CN 112887345 A CN112887345 A CN 112887345A CN 201911196737 A CN201911196737 A CN 201911196737A CN 112887345 A CN112887345 A CN 112887345A
- Authority
- CN
- China
- Prior art keywords
- task
- node
- nodes
- load
- current
- 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.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
A node load balancing scheduling method of an edge computing environment is disclosed, which comprises the steps of obtaining the CPU utilization rate and the I/O resource utilization rate of edge nodes around a base station according to the latest completion time of a task sent by a receiving terminal of the base station; calculating the obtained average load rate of the surrounding nodes, and calculating the load rate of each node; and putting the nodes with the load rates less than or equal to the average load rate into a first queue in a positive sequence mode, putting the nodes with the load rates greater than the average load rate into a second queue in a positive sequence mode, and finally putting the current task to the nodes of the first queue or the nodes of the second queue meeting the current task resources for execution. The invention screens the nodes, thereby avoiding the problem of unbalanced load caused by selecting the nearest node only by a common method; when the nodes are placed under the task, the adopted measurement index is the load rate rather than the load, so that the task can be distributed to the nodes with less resources under the condition of large resource difference among the nodes, and the load balance is really achieved.
Description
Technical Field
The invention relates to a technology in the field of wireless communication, in particular to a node load balancing scheduling method of an edge computing environment.
Background
Edge computing is used as a novel distributed multi-deployment architecture, and compared with cloud computing, all terminals under the edge computing have similar distances and similar propagation delays from computing nodes. In the conventional task scheduling method, a node is selected from edge nodes which are close to a task initiating place as a node for processing the task, and factors such as propagation delay, transmission data volume, network environment, computing capacity of the node, latest completion time of the task and the like need to be considered while the node is selected. However, the scheduling method does not use the current load of the node as a node selection index, and due to the non-uniformity of terminal distribution, part of the nodes may be overloaded and part of the nodes may be idle. Because a single edge compute node has limited computational, network I/O, and storage resources, these resources should be utilized to the maximum extent possible during scheduling.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a node load balancing scheduling method for an edge computing environment, which can correctly distribute tasks according to the load rate of peripheral edge nodes in the edge computing environment.
The invention is realized by the following technical scheme:
the invention relates to a node load balancing scheduling method of an edge computing environment, which comprises the steps of calculating a propagation distance which can be reached before the task is completed at the latest according to the task type, the latest completion time, required computing resources and I/O resources; acquiring all node indexes within the distance that the task can propagate in the edge computing cluster monitoring system, computing the average value of the node indexes of all nodes, and computing the average load rate L according to the average valueaAnd calculating the load rate of the single node by the node indexes of the single node, respectively judging whether each node can load the current task and adaptively transferring the lower task to the corresponding node, thereby realizing balanced scheduling.
The invention relates to a system for realizing the method, which comprises the following steps: the task receiving module, the task allocation module, the task transfer module and the task recovery module, wherein: the task receiving module receives a task transmitted by a terminal through a network and distributes the task to the task distribution module, and the task distribution module calculates to obtain an optimal node for task transfer according to a distribution strategy and outputs task information and corresponding optimal node information to the task transfer module; the task transferring module transfers the task to the optimal node, and after the node finishes executing the task, the corresponding result information of the task is obtained and fed back to the task recycling module; and the task recovery module transmits the task result information back to the terminal through the network.
Technical effects
The invention integrally solves the problem of unbalanced load of the node server in the edge computing environment; the technical effects produced thereby include: the tasks are distributed according to the task type, the node load rate and other indexes, so that the possibility of overloading the server is greatly reduced, and the task execution efficiency is greatly improved.
Compared with the prior art, the method screens the nodes, and avoids the problem of load imbalance caused by selecting the nearest node only by a common method; when the nodes are placed under the task, the adopted measurement index is the load rate rather than the load, so that the task can be distributed to the nodes with less resources under the condition of large resource difference among the nodes, and the load balance is really achieved.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a flow chart of a discrimination of a node put in a queue;
fig. 3 is a flowchart of task selection drop node.
Detailed Description
As shown in fig. 1, this embodiment specifically includes the following steps:
step 1: and acquiring the task type, the latest completion time, the required computing resource and the I/O resource sent by the receiving terminal.
The task types comprise: compute intensive and I/O intensive, for the ith task, there is a computation coefficient ciAnd I/O coefficient miWhen computationally intensive, then: c. Ci≥miWhen the I/O is intensive, then: m isi≥ciAnd c is and ci+mi=2。
Step 2: the propagation distance that can be reached before the latest completion time of the task is calculated.
The propagation distance R ═ Ti-Tn) sv, wherein: v is the propagation velocity of electromagnetic wave in the optical fiber is 2.0 x 105km/s,TiFor the latest completion time of the task, TnIs the current time, s(s)<1) Is the loss factor.
And step 3: and acquiring n node indexes within the distance which can be propagated by the task in the edge computing cluster monitoring system.
The propagation distance is particularly within a circle with a radius R around the base station.
And 4, step 4: and (3) calculating the average value of the node indexes of the n nodes obtained in the step (2), calculating the average load rate La according to the average value, wherein the load rate is used as a measurement index of the current load degree of the nodes, the load rate is determined by the current state of the nodes and the current task information, and tasks are distributed to correct nodes by comparing the load rate with the average load rate of the nodes in the radius R of the base station, so that load balance can be really achieved.
The node load rate is specifically determined according to the CPU utilization rate of the node, the I/O resource utilization rate of the node and the calculation coefficient c of the current taskiAnd I/O coefficient miDetermining:wherein: n is a radical ofj(c) Denotes the CPU utilization, N, of the jth nodej(m) represents the I/O resource utilization of the jth node, ciCoefficient of calculation, m, representing the current taskiRepresenting the I/O coefficient of the current task. The average load rate is composed of the CPU utilization rate of the node within the radius R by taking the base station receiving the current task as the center, the I/O resource utilization rate of the node and the calculation coefficient c of the current taskiAnd I/O coefficient miDetermining:wherein: Nj(c) denotes the CPU utilization, N, of the jth nodej(m) represents the I/O resource utilization of the jth node, ciCoefficient of calculation, m, representing the current taskiI/O coefficient, C, representing the current taskaRepresenting the average CPU utilization, M, of a nodeaRepresenting the average I/O resource usage of the node.
And 5: and (3) calculating the load rate of a single node j according to the node indexes of the single node j in the nodes obtained in the step (2), putting the nodes with the load rates smaller than or equal to the average load rate into a first queue in a positive sequence, and putting the other nodes into a second queue in a positive sequence.
The load rate of the single node j satisfies the following conditions:wherein: n is a radical ofj(c) Denotes the CPU utilization, N, of the jth nodej(m) represents the I/O resource utilization of the jth node, ciCoefficient of calculation, m, representing the current taskiRepresenting the I/O coefficient of the current task.
As shown in FIG. 2, the load rate L of a single nodejLess than or equal to the average load rate LaInto a first queue Q1, the load rate L of a single nodejGreater than the average load rate LaInto a second queue Q2.
Step 6: as shown in fig. 3, the nodes are sequentially taken out from the first queue, whether the node can load the current task is determined, when the node capable of loading the current task exists, the task is put to the node, and when the node capable of loading the current task does not exist, the node in the second queue is determined.
The judgment requires that: a isi<ajAnd b isi<bjWherein: a isiRepresenting the computational resources required by the current task i, ajRepresenting the remaining computational resources of node j, biIndicating the I/O resources required by the current task I, bjRepresenting the remaining I/O resources of node j.
The node indexes are as follows: CPU utilization and node I/O resource usage.
In practical application, the situation that the current task can be loaded without nodes may be encountered, if the situation is encountered, the searching range of the nodes can be expanded, the nodes capable of loading the current task are searched, and then the task is put down, and the specific operation is as follows: expanding the propagation distance calculated in the step 2 by 2 times as follows: 2R, executing the step 2 to the step 6 again, if no node can load the current task, in order to avoid overlong time delay, the current task needs to be abandoned.
Compared with the prior art, the method reduces the number of idle nodes and the number of overload nodes in the edge computing environment, and because the method distributes the tasks according to the load rate and has a larger candidate range, the tasks can be uniformly distributed on the nodes around the base station receiving the tasks, so that the number of the idle nodes and the number of the overload nodes are reduced simultaneously. Due to the reduction of the number of the overload nodes, the task processing efficiency is improved, and the terminal user obtains better experience.
The foregoing embodiments may be modified in many different ways by those skilled in the art without departing from the spirit and scope of the invention, which is defined by the appended claims and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
Claims (6)
1. A node load balancing scheduling method of an edge computing environment is characterized in that a propagation distance which can be reached before the task is completed at the latest is calculated according to the task type, the latest completion time, required computing resources and I/O resources; acquiring all node indexes within the distance that the task can propagate in the edge computing cluster monitoring system, computing the average value of the node indexes of all nodes, and computing the average load rate L according to the average valueaThe node indexes of the single nodes calculate the load rate of the single nodes, whether each node can load the current task or not is judged, and the adaptive lower task is transferred to the corresponding node, so that balanced scheduling is realized;
the node indexes are as follows: CPU utilization and node I/O resource utilization;
the task type packageComprises the following steps: compute intensive and I/O intensive, for the ith task, there is a computation coefficient ciAnd I/O coefficient miWhen computationally intensive, then: c. Ci≥miWhen the I/O is intensive, then: m isi≥ciAnd c isi+mi=2;
The load rate of each node meets the following conditions:wherein: n is a radical ofj(c) Denotes the CPU utilization, N, of the jth nodej(m) represents the I/O resource utilization of the jth node, ciCoefficient of calculation, m, representing the current taskiRepresenting the I/O coefficient of the current task.
2. The method of claim 1, wherein said propagation distance R ═ (T)i-Tn) sv, wherein: v is the propagation velocity of electromagnetic wave in the optical fiber is 2.0 x 105km/s,TiFor the latest completion time of the task, TnS is the loss factor for the current time.
3. The method of claim 1, wherein the average load rate isWherein:Nj(c) denotes the CPU utilization, N, of the jth nodej(m) represents the I/O resource utilization of the jth node, ciCoefficient of calculation, m, representing the current taskiI/O coefficient, C, representing the current taskaRepresenting the average CPU utilization, M, of a nodeaRepresenting the average I/O resource usage of the node.
4. The method of claim 1, wherein said separate determinations are: according to the load rate of each node, putting the nodes with the load rate less than or equal to the average load rate into a first queue in a positive sequence mode, and putting the other nodes into a second queue in a positive sequence mode; traversing the nodes in sequence from the first queue, judging whether the nodes can load the current task, when the nodes capable of loading the current task exist, transferring the task to the nodes, when the nodes capable of loading the current task do not exist, judging in the nodes of the second queue, when the nodes capable of loading the current task still do not exist, expanding the searching range of the nodes to 2 times 2R of the original searching range and recalculating the propagation distance which can be reached before the latest completion time of the task, and when the nodes can not load the current task still, abandoning the current task until the nodes capable of loading the current task are found and then transferring the task.
5. The method of claim 1 or 4, wherein said determining requires that: a isi<ajAnd b isi<bjWherein: a isiRepresenting the computational resources required by the current task i, ajRepresenting the remaining computational resources of node j, biIndicating the I/O resources required by the current task I, bjRepresenting the remaining I/O resources of node j.
6. A system for implementing the method of any preceding claim, comprising: the task receiving module, the task allocation module, the task transfer module and the task recovery module, wherein: the task receiving module receives a task transmitted by a terminal through a network and distributes the task to the task distribution module, and the task distribution module calculates to obtain an optimal node for task transfer according to a distribution strategy and outputs task information and corresponding optimal node information to the task transfer module; the task transferring module transfers the task to the optimal node, and after the node finishes executing the task, the corresponding result information of the task is obtained and fed back to the task recycling module; and the task recovery module transmits the task result information back to the terminal through the network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911196737.4A CN112887345A (en) | 2019-11-29 | 2019-11-29 | Node load balancing scheduling method for edge computing environment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911196737.4A CN112887345A (en) | 2019-11-29 | 2019-11-29 | Node load balancing scheduling method for edge computing environment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112887345A true CN112887345A (en) | 2021-06-01 |
Family
ID=76038348
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911196737.4A Pending CN112887345A (en) | 2019-11-29 | 2019-11-29 | Node load balancing scheduling method for edge computing environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112887345A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210176174A1 (en) * | 2019-12-05 | 2021-06-10 | Institute For Information Industry | Load balancing device and method for an edge computing network |
CN114745563A (en) * | 2022-04-11 | 2022-07-12 | 中国联合网络通信集团有限公司 | Method, device and system for selecting edge computing node to process live broadcast task |
CN115396515A (en) * | 2022-08-19 | 2022-11-25 | 中国联合网络通信集团有限公司 | Resource scheduling method, device and storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1615397A1 (en) * | 2004-04-29 | 2006-01-11 | Alcatel | Device to distribute the load using multiple criteria for a peripheral equipment of a label switched network |
CN103902379A (en) * | 2012-12-25 | 2014-07-02 | 中国移动通信集团公司 | Task scheduling method and device and server cluster |
CN105760227A (en) * | 2016-02-04 | 2016-07-13 | 中国联合网络通信集团有限公司 | Method and system for resource scheduling in cloud environment |
CN106095572A (en) * | 2016-06-08 | 2016-11-09 | 东方网力科技股份有限公司 | The Dispatching System of a kind of big data process and method |
US20170201597A1 (en) * | 2016-01-12 | 2017-07-13 | Cisco Technology, Inc. | Attaching service level agreements to application containers and enabling service assurance |
CN109617989A (en) * | 2018-12-28 | 2019-04-12 | 浙江省公众信息产业有限公司 | For loading method, apparatus, system and the computer-readable medium of distribution |
CN109783235A (en) * | 2018-12-29 | 2019-05-21 | 西安交通大学 | A kind of load equilibration scheduling method based on principle of maximum entropy |
CN110493825A (en) * | 2019-07-18 | 2019-11-22 | 广东工业大学 | The load-balancing method of multitask under a kind of edge calculations framework |
CN110492493A (en) * | 2019-08-16 | 2019-11-22 | 清华大学 | A kind of power system reactive power compensation method for optimizing configuration |
-
2019
- 2019-11-29 CN CN201911196737.4A patent/CN112887345A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1615397A1 (en) * | 2004-04-29 | 2006-01-11 | Alcatel | Device to distribute the load using multiple criteria for a peripheral equipment of a label switched network |
CN103902379A (en) * | 2012-12-25 | 2014-07-02 | 中国移动通信集团公司 | Task scheduling method and device and server cluster |
US20170201597A1 (en) * | 2016-01-12 | 2017-07-13 | Cisco Technology, Inc. | Attaching service level agreements to application containers and enabling service assurance |
CN105760227A (en) * | 2016-02-04 | 2016-07-13 | 中国联合网络通信集团有限公司 | Method and system for resource scheduling in cloud environment |
CN106095572A (en) * | 2016-06-08 | 2016-11-09 | 东方网力科技股份有限公司 | The Dispatching System of a kind of big data process and method |
CN109617989A (en) * | 2018-12-28 | 2019-04-12 | 浙江省公众信息产业有限公司 | For loading method, apparatus, system and the computer-readable medium of distribution |
CN109783235A (en) * | 2018-12-29 | 2019-05-21 | 西安交通大学 | A kind of load equilibration scheduling method based on principle of maximum entropy |
CN110493825A (en) * | 2019-07-18 | 2019-11-22 | 广东工业大学 | The load-balancing method of multitask under a kind of edge calculations framework |
CN110492493A (en) * | 2019-08-16 | 2019-11-22 | 清华大学 | A kind of power system reactive power compensation method for optimizing configuration |
Non-Patent Citations (1)
Title |
---|
乐光学等: ""边缘计算多约束可信协同任务迁移策略"", 《电信科学》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210176174A1 (en) * | 2019-12-05 | 2021-06-10 | Institute For Information Industry | Load balancing device and method for an edge computing network |
CN114745563A (en) * | 2022-04-11 | 2022-07-12 | 中国联合网络通信集团有限公司 | Method, device and system for selecting edge computing node to process live broadcast task |
CN114745563B (en) * | 2022-04-11 | 2024-01-30 | 中国联合网络通信集团有限公司 | Method, device and system for processing live broadcast task by selecting edge computing node |
CN115396515A (en) * | 2022-08-19 | 2022-11-25 | 中国联合网络通信集团有限公司 | Resource scheduling method, device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112887345A (en) | Node load balancing scheduling method for edge computing environment | |
CN110557732B (en) | Vehicle edge computing network task unloading load balancing system and balancing method | |
CN109561148A (en) | Distributed task dispatching method in edge calculations network based on directed acyclic graph | |
CN112153145A (en) | Method and device for unloading calculation tasks facing Internet of vehicles in 5G edge environment | |
CN112600895B (en) | Service scheduling method, system, terminal and storage medium for mobile edge calculation | |
CN108446179B (en) | Weight priority Task scheduling method based on load balancing mechanism | |
CN113115252B (en) | Delay sensitive task distributed mobile edge computing resource scheduling method and system | |
CN111641973A (en) | Load balancing method based on fog node cooperation in fog computing network | |
CN108089918B (en) | Graph computation load balancing method for heterogeneous server structure | |
CN108829512A (en) | A kind of cloud central hardware accelerates distribution method, system and the cloud center of calculating power | |
CN108255427B (en) | Data storage and dynamic migration method and device | |
CN115629865B (en) | Deep learning inference task scheduling method based on edge calculation | |
CN110570075A (en) | Power business edge calculation task allocation method and device | |
CN113961264B (en) | Intelligent unloading algorithm and system for video monitoring cloud edge cooperation | |
CN114595049A (en) | Cloud-edge cooperative task scheduling method and device | |
CN113918240A (en) | Task unloading method and device | |
CN105138408A (en) | Virtual machine migration method and device based on Open Stack | |
CN113329432B (en) | Edge service arrangement method and system based on multi-objective optimization | |
CN112511652B (en) | Cooperative computing task allocation method under edge computing | |
CN110933000A (en) | Distributed data multi-stage aggregation method, device, server and storage medium | |
CN114429195A (en) | Performance optimization method and device for hybrid expert model training | |
CN113778675A (en) | Calculation task distribution system and method based on block chain network | |
CN113504949A (en) | Task unloading and parameter optimization method and system for MAR client in edge computing | |
WO2021000694A1 (en) | Method for deploying services and scheduling apparatus | |
CN117032977A (en) | Mixed part application resource allocation method and device, computer equipment and storage medium |
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 | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20220812 |
|
AD01 | Patent right deemed abandoned |