CN110928683B - Edge computing resource allocation method based on two types of intensive virtual machines - Google Patents
Edge computing resource allocation method based on two types of intensive virtual machines Download PDFInfo
- Publication number
- CN110928683B CN110928683B CN201911112202.4A CN201911112202A CN110928683B CN 110928683 B CN110928683 B CN 110928683B CN 201911112202 A CN201911112202 A CN 201911112202A CN 110928683 B CN110928683 B CN 110928683B
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- time
- intensive
- data
- cpu
- 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.)
- Active
Links
Images
Classifications
-
- 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]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Based on the two-type intensive virtual machine edge computing resource allocation method, the requirements of IO intensive and CPU intensive virtual machines in an edge server on resource allocation are comprehensively considered, the virtual machine state, the uploading data type weight, the system acceleration ratio, the task completion time and the like are respectively introduced into the two-type virtual machine resource allocation computing algorithm, the IO intensive virtual machine priority queue and the number of CPU cores required by the CPU intensive virtual machine are dynamically generated, the reasonable allocation of the edge server resources can be realized, and the influence on cost and time delay caused by intensive computing, high resource consumption and the like is effectively reduced. The problem that the traditional heuristic method, cost function and the like have great influence on the setting human factors of the priority and are not beneficial to the edge server to carry out resource allocation is solved, and the resource allocation efficiency is finally improved.
Description
Technical Field
The invention belongs to the technical field of edge computing resource allocation, and particularly relates to an edge computing resource allocation method based on two types of intensive virtual machines.
Background
In the industrial manufacturing process, the process is more and more complex, the acquired basic data has multiple sources and heterogeneity, and meanwhile, because the industrial software and hardware platform has stronger independence and closure, the edge calculation deployed at the end close to the manufacturing equipment faces great challenges in data analysis and storage capacity. Therefore, the server resources are fully utilized through the reasonable allocation of the edge server resources, and the speed and efficiency of data storage and calculation are further improved. The calculation of the virtual machine priority is a crucial link for resource allocation of the edge server, and whether the virtual machine priority setting reasonably and directly affects the accuracy of resource allocation of the edge server.
The traditional edge computing resource allocation method mostly assumes that the computing capacity and the storage capacity of an edge computing server are in an unlimited state, the existing resource allocation only considers computing resources, the storage resource allocation problem is less involved, and the setting of the priority usually adopts a cost function or a random method and is greatly influenced by human factors. In fact, the calculation of the priority is closely related to network performance, data characteristics, task types, calculation capacity, storage efficiency and the like, which increases the difficulty of determining the edge calculation resource allocation threshold under the cloud-end collaborative computing architecture. Meanwhile, because the manufactured physical unit has dynamic flexibility, the requirement of the edge server on resources is difficult to accurately reflect by directly using the traditional priority setting method. The above disadvantages may cause that resource allocation is difficult to be adjusted in time during the operation of the edge server, and then the efficiency of edge calculation is affected.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method for distributing computing resources based on the edges of two types of intensive virtual machines, which comprises the steps of respectively processing an IO intensive virtual machine and a CPU intensive virtual machine, and generating an IO intensive virtual machine priority queue by combining a second-order difference method and the like aiming at the IO intensive virtual machine; aiming at the CPU intensive virtual machine, the real-time requirements of the CPU intensive virtual machine are calculated according to the influence of the virtual machine on the physical machine performance, the task completion time and the like, and the real-time requirements are used for edge computing resource allocation, so that the data storage efficiency and the computing speed of an edge end server are improved, the time delay is reduced, and the cost is saved.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
the edge computing resource allocation method based on two types of intensive virtual machines comprises the following specific steps:
A. calculating a certain IO intensive virtual machine ViAt sampling time tiReceiving data tau transmitted from terminal equipmenti,
Wherein, N represents the number of terminal devices sending data to the edge server, N represents the number of data types uploaded by a certain terminal device, and alphajkThe size of the data scale corresponding to the uploaded data type is epsilon, and the epsilon represents errors generated by sampling and calculation;
B. for IO intensive virtual machine ViCarrying out second-order difference on the received data to obtain a data uploading rate increment delta v:
suppose IO intensive virtual machine ViAverage velocity of processed data is viThen, time t 'is required'iWill tiThe data transmitted at any moment are all stored in a database, i.e.
Then the IO intensive virtual machine ViTreatment time t'iIs divided into three parts of t'i,t′i+1,t′i+2It means that three stages are required to store the data in the database to obtain the data uploading rate increment deltav,
Δv=τi+2-2τi+1+τi
wherein, taui+2,τi+1,τiAre respectively corresponding to t'i+2,t′i+1,t′iTime virtual machine ViThe amount of data stored in the database;
C. judging IO intensive virtual machine V according to upload rate increment delta ViAll possible states are represented by the set status ═ {1,2,3,4,5,6}, where 1 is the no data state, 2 is the decrease state, 3 is the steady state, 4 is the increase state, 5 is the early warning state, and 6 is the excess state;
D. compute IO intensive virtual machine ViProbability that current state is the same as previous state
Wherein ξiNum represents the number of times that the current state of the virtual machine is the same as the previous stateiRepresenting virtual machine V as a proper subset of state set statusiA set of state changes during a work session;
E. compute IO intensive virtual machine ViPriority p of stored datai,
Wherein S isi preRepresenting IO intensive virtual machine ViSoft allocated pre-stored space size, Si postRepresenting virtual machine ViAt tiThe size of data stored in a storage node at any moment;
F. compute controller-to-IO intensive virtual machines V in control nodesiPriority P for resource allocationi,PiThe larger the value, the more computing resources need to be allocated to the virtual machine in time, to increase the data storage rate,
wherein, W is a weight set corresponding to different types of data, and W ═ W1,w2,...,wnN represents a terminal deviceThe number of uploaded data types;
G. assigning a priority P to each IO intensive virtual machine in an edge serveriEstablishing a maximum priority queue P, correspondingly updating the resource allocation of the IO intensive virtual machine through the change of the priority queue P, optimizing and reasonably allocating storage resources;
H. calculating a certain CPU intensive virtual machine ViAll tasks in' execute the required Time under parallel pseudo-distributed conditions,
wherein, time represents the average time of executing a single task, C is the number of CPU cores, and T is the CPU intensive virtual machine Vi' the number of tasks to be executed simultaneously;
I. compute CPU intensive virtual machine ViThe acceleration ratio sp of' is such that,
wherein R isnFor CPU-intensive virtual machines Vi' minimum execution time for completing task under parallel pseudo-distributed condition, RdFor CPU-intensive virtual machines Vi' minimum execution time to complete a task under parallel distributed conditions;
J. when the number of the virtual machines exceeds the CPU core number C of the host machine, the physical machine performance is gradually reduced along with the increase of the number of the virtual machines, the influence factor is theta, the relationship between the number G of the virtual machines and the influence factor theta is obtained,
wherein G is the number of virtual machines;
K. defining CPU intensive virtual machine Vi'Time required for all tasks to execute under parallel distribution condition',
wherein sp is the acceleration ratio, theta is the influence factor, and Time is the CPU-intensive virtual machine Vi' all tasks in the system are executed under the condition of parallel pseudo distribution for required time;
l. setting a CPU intensive virtual machine ViIn the method, the upper Time limit Req _ Time ' required to be completed by all tasks is Time _ com + Time _ cor, where Time _ com is the actual completion Time of the task, Time _ cor is the communication delay Time, all tasks can be completed smoothly, and it is necessary to satisfy that Time ' is less than or equal to Req _ Time ', and if not, it is necessary to allocate the computing resources, i.e. the number of CPU cores, and the specific process is as follows: suppose that d tasks b 'of the b tasks that are not completed on time are { b'1,b′2,...,b′dAt execution time of
The corresponding request is completed at an upper time limit of
Then in the present case, the CPU intensive virtual machine Vi' the computational resources to be allocated are as follows:
wherein the content of the first and second substances,for CPU-intensive virtual machines Vi' number of CPU cores to be allocated.
The invention has the beneficial effects that:
the invention comprehensively considers the requirements of IO intensive and CPU intensive virtual machines in the edge server on resource allocation, respectively introduces the virtual machine state, the uploading data type weight, the system acceleration ratio, the task completion time and the like into the two types of virtual machine resource allocation priority calculation algorithms, can realize reasonable allocation of the edge server resources, and thereby effectively reduces the influence on cost and time delay caused by intensive calculation, high resource consumption and the like. The problem that the traditional random method, cost function and the like have great influence on the setting human factors of the priority and are not beneficial to the edge server to carry out resource allocation is solved, and the resource allocation efficiency is finally improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a diagram of a resource allocation result;
FIG. 3 is a timing comparison diagram of the multitask concurrent execution and other methods of the present invention.
Detailed Description
The present invention is described in further detail below with reference to the attached drawing figures.
Referring to fig. 1, the method for distributing edge computing resources based on two types of intensive virtual machines includes the following steps:
A. calculating a certain IO intensive virtual machine Vi∈V={V1,V2,…,VQQ is the number of virtual machines at sampling time tiReceiving data tau transmitted from terminal equipmenti,
Wherein, N represents the number of terminal devices sending data to the edge server, and N represents the data type set S uploaded by a certain terminal devicej={sj1,sj2,…,sjnNumber of elements in }, αjk∈αj={αj1,αj2,...αjk,...αjnThe size of the data scale corresponding to the type of the uploaded data is used, and epsilon represents errors generated by sampling and calculation;
B. for IO intensive virtual machine ViCarrying out second order difference on the received data to obtain a numberData upload rate increment Δ v:
suppose a virtual machine ViAverage velocity of processed data is viThen, time t 'is required'iWill tiThe data transmitted at any moment are all stored in a database, i.e.
Then the IO intensive virtual machine ViTreatment time t'iIs divided into three parts of t'i,t′t+1,t′i+2It means that three stages are required to store the data in the database to obtain the data uploading rate increment deltav,
Δv=τi+2-2τi+1+τi
wherein, taui+2,τi+1,τiAre respectively corresponding to t'i+2,t′i+1,t′iTime virtual machine ViThe amount of data stored in the database;
C. judging IO intensive virtual machine V according to upload rate increment delta ViIs represented by the set status ═ {1,2,3,4,5,6}, where 1 is the no data state, 2 is the decrease state, 3 is the steady state, 4 is the increase state, 5 is the early warning state, and 6 is the excess state; setting: when Δ v ≈ τiWhen the terminal equipment receives the data, the virtual machine does not receive the data sent by the terminal equipment in the time period, and the virtual machine is in a data-free state; when-taui<Δv<When 0, the data uploaded by the terminal equipment in the time slot is reduced, and the data is in a reduced state; when the delta v is approximately equal to 0, the data uploaded by the terminal equipment in the time period basically keeps unchanged and is in a stable state; when Δ v<M0When the time is up, the data uploaded by the terminal equipment in the time slot is increased and is in an increased state; when M is0≤Δv<M1When the time is up, the data uploaded by the terminal equipment in the time period exceeds the early warning value M0The state is an early warning state; when Δ v.gtoreq.M1When the data uploaded by the terminal equipment exceeds the actual physical storage capacity of the storage node in the time period, the data is in an excess state and needs to be addedHard disk space or optimized data storage scheme to address the rate of terminal device data upload, where M1The actual storage size of the storage space of the edge server.
D. Compute IO intensive virtual machine ViProbability that current state is the same as previous state
Wherein ξiNum represents the number of times that the current state of the virtual machine is the same as the previous stateiRepresenting virtual machine V as a proper subset of state set statusiA set of state changes over a period of operation;
E. compute IO intensive virtual machine ViPriority p of stored datai,
Wherein S isi preRepresenting virtual machine ViSoft allocated pre-stored space size, Si postRepresenting virtual machine ViAt tiThe size of data stored in a storage node at any moment;
F. computing a controller-to-virtual machine V in a control nodeiPriority P for resource allocationi,PiThe larger the value, the more computing resources need to be allocated to the virtual machine in time, to increase the data storage rate,
wherein, W is a weight set corresponding to different types of data, and W ═ W1,w2,...,wn};
G. Assigning a priority P to each IO intensive virtual machine in an edge serveriAnd establishes a maximum priority queue P by priorityThe change of the level queue P correspondingly updates the resource configuration of the IO intensive virtual machine, optimizes and reasonably distributes storage resources;
H. calculating a certain CPU intensive virtual machine ViAll tasks in' execute the required Time under parallel pseudo-distributed conditions,
wherein, time represents the average time of executing a single task, C is the number of CPU cores, and T is the CPU intensive virtual machine Vi' the number of tasks to be executed simultaneously;
I. compute CPU intensive virtual machine ViThe acceleration ratio sp of' is such that,
wherein R isnFor CPU-intensive virtual machines Vi' minimum execution time for completing task under parallel pseudo-distributed condition, RdFor CPU-intensive virtual machines Vi' minimum execution time to complete a task under parallel distributed conditions;
J. when the number of the virtual machines exceeds the CPU core number C of the host machine, the physical machine performance is gradually reduced along with the increase of the number of the virtual machines, the influence factor is theta, the relationship between the number G of the virtual machines and the influence factor theta is obtained,
wherein G is the number of virtual machines;
K. defining CPU intensive virtual machine Vi'Time required for all tasks to execute under parallel distribution condition',
wherein sp is the acceleration ratio, theta is the influence factor, and Time is the CPU-intensive virtual machine Vi' all tasks in the system are executed under the condition of parallel pseudo distribution for required time;
l. setting a CPU intensive virtual machine ViIn the method, the upper Time limit Req _ Time ' required to be completed by all tasks is Time _ com + Time _ cor, where Time _ com is the actual completion Time of the task, Time _ cor is the communication delay Time, all tasks can be completed smoothly, and it is necessary to satisfy that Time ' is less than or equal to Req _ Time ', and if not, it is necessary to allocate the computing resources, i.e. the number of CPU cores, and the specific process is as follows: suppose that d tasks b 'of the b tasks that are not completed on time are { b'1,b′2,...,b′dAt execution time of
The corresponding request is completed at an upper time limit of
Then in the present case, the CPU intensive virtual machine Vi' the computational resources to be allocated are as follows:
wherein the content of the first and second substances,for CPU-intensive virtual machines Vi' number of CPU cores to be allocated;
m. to demonstrate the effectiveness of the method of the invention, it is compared with different methods, as further described. The virtual environment of the edge server is created through the VMware VSphere virtualization platform, and comprises 1 Controller node, 4 IO intensive virtual machines and 3 CPU intensive virtual machines.
Firstly, respectively running client services of a resource distributor on 4 IO intensive virtual machines, monitoring IO performance of the virtual machines, obtaining priority through calculation and sending the priority to a Controller node, running server services of the resource distributor on the Controller node, collecting the priority of the IO intensive virtual machines, and responding to resource requests of the IO intensive virtual machines according to a maximum priority list. Compared with the current popular heuristic algorithm, namely a genetic algorithm, the algorithm adopted by the invention can distribute resources for the optimal response of the IO intensive virtual machine under the condition of ensuring sufficient sampling width. The genetic algorithm has uncertainty due to the adoption of random numbers, and the resource response is not reasonable enough, for example, between 7s and 23s, the resource allocation time is too compact, so that part of resources can be repeatedly allocated; the allocation of resources at 60s does not take into account the possibility that data continues to rise at 59s and is not robust enough. Therefore, the algorithm of the present invention can adaptively allocate resources to the IO-intensive virtual machine according to the business requirements, and the specific result is shown in fig. 2.
Next, the Hadoop and Spark distributed storage and computation frameworks were run on the 3 CPU-intensive virtual machines, respectively. Aiming at common CPU intensive application programs such as WordCount, Sort, TeraSort, RandomWriter and the like, test data sets with different scales are generated through HiBench and stored in an HDFS file system and submitted to Spark to execute corresponding tasks. Compared with the current popular genetic algorithm, the completion time of the experiment after the 6 tasks are executed 100 times concurrently is shown in figure 3, wherein, (a) is Task-1WordCount execution time comparison, (b) is Task-2 Sort execution time comparison, (c) is Task-3 TeraSort execution time comparison, (d) is Task-4 RandomWriter execution time comparison, (e) is Task-5 Grep execution time comparison, and (f) is Task-6 TestDFSIO execution time comparison, and the method adopted by the invention has higher operation efficiency.
Claims (1)
1. An edge computing resource allocation method based on two types of intensive virtual machines comprises the following specific steps:
A. calculating a certain IO intensive virtual machine ViAt sampling time tiReceiving data tau transmitted from terminal equipmentiI.e. by
Wherein, N represents the number of terminal devices sending data to the edge server, N represents the number of data types uploaded by a certain terminal device, and alphajkThe size of the data scale corresponding to the uploaded data type is epsilon, and the epsilon represents errors generated by sampling and calculation;
B. for IO intensive virtual machine ViCarrying out second-order difference on the received data to obtain a data uploading rate increment delta v:
first assume IO intensive virtual machine ViAverage velocity of processed data is viThen, it takes time ti' will tiThe data transmitted at any moment are all stored in a database, i.e.
Then the IO intensive virtual machine ViIs processed for a time ti'divided into three portions t'i,t′i+1,t′i+2It means that three stages are required to store the data in the database to obtain the data uploading rate increment deltav,
Δv=τi+2-2τi+1+τi
wherein, taui+2,τi+1,τiAre respectively corresponding to t'i+2,t′i+1,ti' time IO intensive virtual machine ViThe amount of data stored in the database;
C. judging IO intensive virtual machine V according to upload rate increment delta ViAll possible states are represented by the set status ═ {1,2,3,4,5,6}, where 1 is a no-data state,2 is a decrease state, 3 is a steady state, 4 is an increase state, 5 is an early warning state, and 6 is an excess state;
D. compute IO intensive virtual machine ViProbability that current state is the same as previous state
Wherein ξiNum represents the number of times that the current state of the virtual machine is the same as the previous stateiRepresenting virtual machine V as a proper subset of state set statusiA set of state changes during a work session;
E. computer virtual machine ViPriority p of stored datai,
Wherein S isi preRepresenting IO intensive virtual machine ViSoft allocated pre-stored space size, Si postRepresenting virtual machine ViAt tiThe size of data stored in a storage node at any moment;
F. compute controller-to-IO intensive virtual machines V in control nodesiPriority P for resource allocationi,PiThe larger the value, the more computing resources need to be allocated to the virtual machine in time, to increase the data storage rate,
wherein, W is a weight set corresponding to different types of data, and W ═ W1,w2,...,wn};
G. Assigning a priority P to each IO intensive virtual machine in an edge serveriAnd establishing a maximum priority queue P, and changing the queue P according to the priorityThe resource allocation of the IO intensive virtual machine is correspondingly updated, and storage resources are optimized and reasonably allocated;
H. calculating a certain CPU intensive virtual machine ViAll tasks in' execute the required Time under parallel pseudo-distributed conditions,
wherein, time represents the average time of executing a single task, C is the number of CPU cores, and T is the CPU intensive virtual machine Vi' the number of tasks to be executed simultaneously;
I. compute CPU intensive virtual machine ViThe acceleration ratio sp of' is such that,
wherein R isnFor CPU-intensive virtual machines Vi' minimum execution time for completing task under parallel pseudo-distributed condition, RdFor CPU-intensive virtual machines Vi' minimum execution time to complete a task under parallel distributed conditions;
J. when the number of the virtual machines exceeds the CPU core number C of the host machine, the physical machine performance is gradually reduced along with the increase of the number of the virtual machines, the influence factor is theta, the relationship between the number G of the virtual machines and the influence factor theta is obtained,
wherein G is the number of virtual machines;
K. defining CPU intensive virtual machine Vi'Time required for all tasks to execute under parallel distribution condition',
wherein sp is the acceleration ratio, theta is the influence factor, and Time is the CPU-intensive virtual machine Vi' all tasks in the system are executed under the condition of parallel pseudo distribution for required time;
l. setting a CPU intensive virtual machine ViIn the method, the upper Time limit Req _ Time ' required to be completed by all tasks is Time _ com + Time _ cor, where Time _ com is the actual completion Time of the task, Time _ cor is the communication delay Time, all tasks can be completed smoothly, and it is necessary to satisfy that Time ' is less than or equal to Req _ Time ', and if not, it is necessary to allocate the computing resources, i.e. the number of CPU cores, and the specific process is as follows: suppose that d tasks b 'of the b tasks that are not completed on time are { b'1,b′2,...,b′dAt execution time of
The corresponding request is completed at an upper time limit of
Then in the present case, the CPU intensive virtual machine Vi' the computational resources to be allocated are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911112202.4A CN110928683B (en) | 2019-11-14 | 2019-11-14 | Edge computing resource allocation method based on two types of intensive virtual machines |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911112202.4A CN110928683B (en) | 2019-11-14 | 2019-11-14 | Edge computing resource allocation method based on two types of intensive virtual machines |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110928683A CN110928683A (en) | 2020-03-27 |
CN110928683B true CN110928683B (en) | 2022-01-28 |
Family
ID=69854013
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911112202.4A Active CN110928683B (en) | 2019-11-14 | 2019-11-14 | Edge computing resource allocation method based on two types of intensive virtual machines |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110928683B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116684483B (en) * | 2023-08-02 | 2023-09-29 | 北京中电普华信息技术有限公司 | Method for distributing communication resources of edge internet of things proxy and related products |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105718310A (en) * | 2016-01-13 | 2016-06-29 | 上海应用技术学院 | Virtual machine scheduling method for IO dense application under cloud platform |
CN106951330A (en) * | 2017-04-10 | 2017-07-14 | 郑州轻工业学院 | A kind of maximized virtual machine distribution method of cloud service center service utility |
US10037231B1 (en) * | 2017-06-07 | 2018-07-31 | Hong Kong Applied Science and Technology Research Institute Company Limited | Method and system for jointly determining computational offloading and content prefetching in a cellular communication system |
CN108804227A (en) * | 2018-05-23 | 2018-11-13 | 大连理工大学 | The method of the unloading of computation-intensive task and best resource configuration based on mobile cloud computing |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10938736B2 (en) * | 2017-10-18 | 2021-03-02 | Futurewei Technologies, Inc. | Dynamic allocation of edge computing resources in edge computing centers |
-
2019
- 2019-11-14 CN CN201911112202.4A patent/CN110928683B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105718310A (en) * | 2016-01-13 | 2016-06-29 | 上海应用技术学院 | Virtual machine scheduling method for IO dense application under cloud platform |
CN106951330A (en) * | 2017-04-10 | 2017-07-14 | 郑州轻工业学院 | A kind of maximized virtual machine distribution method of cloud service center service utility |
US10037231B1 (en) * | 2017-06-07 | 2018-07-31 | Hong Kong Applied Science and Technology Research Institute Company Limited | Method and system for jointly determining computational offloading and content prefetching in a cellular communication system |
CN108804227A (en) * | 2018-05-23 | 2018-11-13 | 大连理工大学 | The method of the unloading of computation-intensive task and best resource configuration based on mobile cloud computing |
Non-Patent Citations (5)
Title |
---|
Application Aware Workload Allocation for Edge Computing-Based IoT;Qiang Fan等;《 IEEE Internet of Things Journal》;IEEE;20180412;第5卷(第3期);第2146-2153页 * |
Enabling Workload Engineering in Edge, Fog, and Cloud Computing through OpenStack-based Middleware;Giovanni Merlino等;《ACM Transactions on Internet Technology》;ACM;20190430;第19卷(第2期);第1-22页 * |
Workload Allocation Mechanism for Minimum Service Delay in Edge Computing-Based Power Internet of Things;Xudong Niu等;《IEEE Access 》;IEEE;20190531;第7卷;第83771-83784页 * |
基于云端协同计算架构的边缘端I/O密集型虚拟机资源分配方案;赵龙乾等;《计算机应用研究》;四川省计算机研究院;20190930;第37卷(第9期);第2734-2738页 * |
移动边缘计算环境下的动态资源分配策略;朱新峰等;《计算机工程与科学》;国防科技大学计算机学院;20190715(第7期);第1184-1190页 * |
Also Published As
Publication number | Publication date |
---|---|
CN110928683A (en) | 2020-03-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20220391771A1 (en) | Method, apparatus, and computer device and storage medium for distributed training of machine learning model | |
CN109617826B (en) | Storm dynamic load balancing method based on cuckoo search | |
CN107659433B (en) | Cloud resource scheduling method and equipment | |
CN108845874B (en) | Dynamic resource allocation method and server | |
CN111381950A (en) | Task scheduling method and system based on multiple copies for edge computing environment | |
KR20180077218A (en) | Processing computed graphs | |
CN110688219B (en) | Adaptive weight load balancing algorithm based on reverse chaotic cuckoo search | |
CN111274036A (en) | Deep learning task scheduling method based on speed prediction | |
CN109189563B (en) | Resource scheduling method and device, computing equipment and storage medium | |
CN110502323B (en) | Real-time scheduling method for cloud computing tasks | |
CN111131486B (en) | Load adjustment method and device of execution node, server and storage medium | |
CN103401939A (en) | Load balancing method adopting mixing scheduling strategy | |
CN112087332B (en) | Virtual network performance optimization system under cloud edge cooperation | |
CN102281290A (en) | Emulation system and method for a PaaS (Platform-as-a-service) cloud platform | |
CN107562537B (en) | Cloud computing task scheduling method based on universal gravitation search | |
CN114647515A (en) | GPU cluster-oriented dynamic resource scheduling method | |
US9665409B2 (en) | Methods and apparatus facilitating access to storage among multiple computers | |
CN113778691A (en) | Task migration decision method, device and system | |
CN116467082A (en) | Big data-based resource allocation method and system | |
Manikandan et al. | Virtualized load balancer for hybrid cloud using genetic algorithm | |
CN116684420A (en) | Cluster resource scheduling method, device, cluster system and readable storage medium | |
CN110928683B (en) | Edge computing resource allocation method based on two types of intensive virtual machines | |
Zhou et al. | JPAS: Job-progress-aware flow scheduling for deep learning clusters | |
US20210390405A1 (en) | Microservice-based training systems in heterogeneous graphic processor unit (gpu) cluster and operating method thereof | |
CN110743164B (en) | Dynamic resource partitioning method for reducing response delay in cloud game |
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 |