CN1845075A - Service oriented high-performance grid computing job scheduling method - Google Patents
Service oriented high-performance grid computing job scheduling method Download PDFInfo
- Publication number
- CN1845075A CN1845075A CN 200610026430 CN200610026430A CN1845075A CN 1845075 A CN1845075 A CN 1845075A CN 200610026430 CN200610026430 CN 200610026430 CN 200610026430 A CN200610026430 A CN 200610026430A CN 1845075 A CN1845075 A CN 1845075A
- Authority
- CN
- China
- Prior art keywords
- subtask
- resource
- resource service
- oriented
- service group
- 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
Abstract
The invention relates to a network high-performance calculation transfer method faced to service. It uses face-service method; the grid system is formed by distributed resource server group; according to the character of high-performance calculation application, and the character of cooperation treatment between different calculating stations in the grid calculation, decomposing one high-performance calculation application into one calculation subtasks with data input and output relationships; using direction non-cycle graph to express the data relativity between the subtasks; using improved dynamic priority transfer algorism to synchronously match the ready subtask and usable resource server group; then transferring selected subtask to selected resource server group; the invention has considered the condition that one subtask can be executed only in some special resource server group, while it can realize the high-efficiency transfer of high-performance calculation in the grid range.
Description
Technical field
What the present invention relates to is a kind of method of field of computer technology, particularly is a kind of service-oriented high-performance grid computing job scheduling method.
Background technology
Along with network technology and development of computer, and to the ever-increasing demand of high-performance calculation, on the basis that cluster calculates, early stage unit's calculating and computing grid have in recent years appearred.Because the successful Application of grid computing in high-performance calculation, facilitated grid computing further to be used and deepen in many-sides such as information service, data processing.Simultaneously, obtained significant progress under the Web service technology promotion in the industry cycle.Based on gridding technique and Web service, by IBM and Globus alliance WSRF (WS-Resource Framework) has been proposed, to integrate the advantage of Web service and gridding technique, be implemented in the resource sharing in trans-regional and the mechanism, make full use of all kinds of computational resources.No matter based on the WSRF of Web service and gridding technique, or other Enterprise SOA (SOA), all need be at dynamically, establish effective ways and the mechanism that how to construct, dispose and to use oriented application under the open computing environment.On the basis of traditional calculations grid, high-performance calculation is used also will have benefited from Enterprise SOA.Similar with traditional framework, under Enterprise SOA, problems such as the organization and administration of same needs solution resource, the decomposition of task and scheduling.
Finding by prior art documents, serves as mainly to have proposed a kind of service-oriented grid job management method (surplus petrel is looked into gift with surplus petrel, Li Wei. a kind of service-oriented grid job management mechanism, computer research and development, 2003,40 (12): 1770-1774).This method is based on a kind of service-oriented task management mechanism, and it is as the agency of user capture gridding resource (service), for the user provides transparent, that have nothing to do with the resource physical location and has the job service interface that session is supported.And the notion of having introduced service level agreement (SLA) represents the different mesh services ranks of user's request, and job management system then realizes that according to customizable service level configuration (SLAP) is mapped to concrete task management behavior with every QoS characteristic of stipulating among the SLA.This task management mechanism has been applied in Vega grid system software, and support flexibly and effectively can be provided for the application based on service grid environment.This method is primarily aimed at the issued transaction operation under the grid environment, does not consider the characteristics that high-performance calculation is used, and therefore is not suitable for the high-performance grid computing application.
Summary of the invention
The present invention is directed to the deficiencies in the prior art, in conjunction with high-performance calculation application and grid computing characteristics, a kind of service-oriented high-performance grid computing job scheduling method is proposed, achieve the effective scheduling of high-performance calculation operation under grid environment, improve the utilization factor of resource, reduce the execution time of operation.
The present invention is achieved by the following technical solutions, the present invention adopts service-oriented mode, grid system is made of the resource service group that distributes, the characteristics of collaborative process operation between the various computing website in the feature of using according to high-performance calculation, the grid computing are used a high-performance calculation and are decomposed into one group of calculating subtask that has the data input/output relation.Adopt oriented no circular chart to represent the data dependence between each subtask in the high-performance calculation application, adopt follow-on dynamic priority scheduling algorithm to mate ready subtask and available resource service group simultaneously, with selected subtask scheduling in selected resource service group, consider a situation that the subtask can only be carried out on some specific resource service groups, realized the efficient scheduling of the high-performance calculation operation in the grid scope.
The inventive method comprises that the following step poly-:
(1) determines ready task { v in the oriented no circular chart of operation (DAG)
i}
A common high-performance calculation application task can resolve into a plurality of subtasks, and at the characteristics of wide area network, computing grid is applicable to that usually the less high-performance calculation of the traffic is used between the subtask.Have certain data dependence between these subtasks, wherein the input and output correlation circumstance is more, and therefore, a plurality of subtasks of a high-performance calculation application can be represented with oriented no circular chart (DAG).
Oriented no circular chart G=(V, E), vertex set V={v wherein
1, v
I..., v
mThe expression high-performance calculation uses a plurality of subtasks be decomposed, e
Ij=(v
i, v
j) ∈ E represents from subtask v
iTo subtask v
jCommunication, and | e
Ij| then represent the traffic, communication can be divided into two big classes: a class is the file data communication of big data, and another kind of is the supplemental characteristic communication of small data.It should be noted that, for a subtask, be not all can move in all the resource management territories in grid environment, need specific scientific engineering computing storehouse as some subtasks, and this scientific engineering computing storehouse exist only on some specific calculating nodes.
If subtask v in the oriented no circular chart
iAll preceding thread subtasks are all to carry out, and then this subtask is a ready task.
(2) obtain the real-time information { R of current available resource service groups
j}
If high-performance calculation used be divided into a plurality of subtasks, by the corresponding calculated program solution, these subtasks are to belong to the computation-intensive task usually respectively, therefore only can move a program copy usually on a particular computational.The computational resource of this calculation procedure and specific run has promptly constituted the resource service group.
Grid environment can be defined as by a series of resource service group and constitute, and wherein each resource service group is made up of software of calculation and specific hardware computational resource, can be visited by the external world very expediently by service encapsulates mechanism.The set of resource service group is expressed as R={R formally
1, R
2..., R
n.Expense function C: V * R → R represents that one is calculated the time overhead that carry out the subtask, subtask v on the resource service group
iAt resource service group R
jThe time overhead of last execution is defined as C (v
i, R
j).For the resource service group that can not carry out certain subtask, define its time expense for infinitely great.
By the information service in the inquiry grid system, can obtain the real-time information { R of corresponding resource service group
j.
Based on the setting in (1) and step poly-(2), high-performance grid computing application scheduling problem can be expressed as the mapping problems of figure, and be about to an oriented no circular chart and be mapped in the set of resource service group, be target to minimize total execution time.
(3) calculate MDL (v
i, R
J|j ∈ A (vi), ∑ (t))
Job scheduling method adopts a kind of improved dynamic priority scheduling algorithm (DLS), and the DLS algorithm is to carry out scheduler task by task priority, and the subtask of having dispatched in the priority that the characteristics of DLS algorithm are a subtask and the task image is relevant.
Improved dynamic priority (Dynamic Level) is expressed as MDL (v
i, R
j, ∑ (t)), it has reflected under state set ∑ (t) situation, subtask v
iBe dispatched to resource service group R
jOn matching degree, state set ∑ (t) expression is the status information of all the resource service groups and the communication resource etc. during t constantly.Form ground, dynamic priority MDL can be defined as:
Wherein, t is the current time.First in the formula represents subtask v
iIn the static priority of task image, its value is from subtask v
iThe maximal value of execution time summation to the different paths of task image terminal point, for the task image that a high-performance calculation is used, this is a static information.t
d(v
i, R
j) be illustrated under state set ∑ (t) situation subtask v
iThe pot life the earliest of required all data.t
a(R
j) expression resource service group free time the earliest.Δ (v
i, R
j) then be the difference of expression resource service group processing power, be defined as:
Δ(v
i，R
j)＝C(v
i)-C(v
i，R
j)
Wherein, C (v
i) expression subtask v
iThe mean value of execution time on all resource service groups.The execution time is the maximal value of finite time value on all resource service groups if the mean value that calculates for infinitely great, then is taken at.A (v
i) expression can subtasking v
iThe set of all available resources service groups.Calculating Δ (v
i, R
j) time, C (v
i) be defined as:
Wherein, ‖ A (v
i) ‖ represents set A (v
i) in the number of element.
(4) determine to make the resource service group R that the MDL value is maximum
JWith ready task v
I
(5) scheduling ready task v
ITo resource service group R
J
(6) wait for scheduling events next time, and repeat said process.
Wherein, resource is meant all kinds of computational resources, can be physically cpu resource, storage resources, also can be in logic database, engineering calculation storehouse etc.; Web service is a kind of interface, and it has described on network the operational set that can conduct interviews by related protocol; And the resource service group is meant the operational set on specific resources; Task is to refer to high-performance calculation to use, and the subtask refers to the computing module after decomposition is used in high-performance calculation.
In the job scheduling method that the present invention proposes, MDL is a target with the maximization dynamic priority, mate ready subtask and available resource service group, and with this subtask scheduling in selected resource service group, the status information of update system then, recomputate MDL, determine the scheduling of next subtask.Its advantage is that subtask and resource service group select at the same time, is better than choosing separately subtask or resource service group.
The job scheduling method that the present invention proposes, be to use at high-performance calculation, adopt service-oriented mode, pass through net point infrastructure, for in wide scope, realize collaborative find the solution the high-performance calculation problem provide a kind of efficiently, method flexibly, considered simultaneously to finish the situation that a computational tasks needs the dedicated computing resource, be different from situation that computational tasks can be carried out in the prior art on all computational resources.Therefore, the job scheduling method of proposition can be advantageously applied to the actual conditions of high-performance calculation in the grid environment, can obtain more performance.
Description of drawings
Fig. 1 is service-oriented stratification resource management architecture structural drawing
Fig. 2 uses the oriented no circular chart synoptic diagram of a plurality of subtasks formation of decomposing for high-performance calculation
Fig. 3 is a task scheduling process flow diagram of the present invention
Embodiment
For realizing the computational resource of efficient Tissue distribution, isomery, autonomy, grid system adopts following organizational form: grid portal, global resource administration and supervision authorities and local resource management layer, as shown in Figure 1.
1) grid portal
On the upper strata of institutional framework is grid portal, also is terminal user and the mutual assembly of high-performance grid computing environment.Grid portal provide instrument for the terminal user so that under grid environment the extensive performance application of cooperative scheduling, by succinct graphical interfaces or Web interface, the terminal user can obtain information on services very expediently, for example computing ability, scientific and engineering storehouse, system load etc. also can be called calculation services, configuration parameter, monitoring intermediate result simultaneously expediently and be downloaded final calculation result etc.Theoretically, the terminal user need not pay close attention to grid agreement and net point infrastructure; On the other hand, the developer in performance application field also need not pay close attention to the configuration of application software under the grid environment.The terminal user can utilize a plurality of high-performance calculation application software to finish large-scale science calculation task by grid platform very expediently.
2) global resource administration and supervision authorities
In the global resource administration and supervision authorities, need to realize the parsing of services request, the decomposition of calculation task, the cooperative scheduling that calculates the subtask and information service etc.
Agency service (broker service) is an important composition part of grid middleware, resolves the user who comes by the grid portal transmission by its and asks, and finally according to the different application problem, be a plurality of subtasks with a large-scale high-performance calculation task division.For example, airplane complete machine simulation task can be decomposed into a plurality of subtasks, and promptly a plurality of functional modules, each functional module are calculated the fuselage of aircraft, two wings, tailplane and vertical tails respectively.Task is resolved and decomposition is closely related with the corresponding calculated task type, the different different decomposition strategy of high-performance calculation application need.
Usually there is certain data dependence between a plurality of subtasks that are decomposed, sets and have the input and output relevance between these subtasks.These subtasks can be carried out by a plurality of application software that distribute on the geography, stride management domain respectively by collaborative distribution services.Information service then provides the status information of current total system.
3) local resource management layer
In order to manage local resource, need two class Web services: a class is the job factory service, and another kind of is job management service.The job factory service is created Service Instance for the request of high-performance calculation subtask, and establishes the corresponding calculated resource.The resource information of local resource management domain is provided to the external world simultaneously, comprises hardware asset information and computing application software service.Job management service then is to manage the Service Instance that has been created, and in conjunction with the computational resource of establishing, forms corresponding specific requested resource service groups, and issue has been submitted to the status information of the operation of background dispatching system.
The specific service of this two class is realized by the computational resource of special use, be service that high-performance calculation subtask request is created then be dynamic attachment in current available resource, promptly be dynamically to specify according to request type and current available resource information by local resource management layer.
Concrete scheduling process is as follows:
Grid user is submitted a high-performance calculation operation to by grid portal, is decomposed according to the feature of computing application by the agency service in the global resource administration and supervision authorities then, forms the oriented no cycle task figure of corresponding task.As shown in Figure 2, this computational tasks is by 7 subtask v
iForm, i=1,2 ..., 7, the communication between them is represented on the limit between each subtask, has also reflected the required precedence of they execution.
The scheduling framework of operation as shown in Figure 3, the collaborative distribution services in the global resource administration and supervision authorities is inquired about the real-time information of resource service group in the current grid, i.e. R={R to information service
1, R
2..., R
nInformation.
Collaborative distribution services defines ready subtask in no cycle task figure, i.e. the pairing subtask of node that forerunner's node has been carried out in oriented no circular chart.At each subtask, calculate MDL (v respectively
i, R
J|j ∈ A (vi), ∑ (t)), and the ready subtask v of selected scheduling
ITo resource service group R
J, promptly in these subtasks and resource service group, resource service group R
JWith ready subtask v
IMake the MDL value maximum.Collaborative distribution services repeats above-mentioned scheduling process according to scheduling events, all is scheduled on the corresponding resource service group until all subtasks.
Resource service group R
JJob management service obtains to carry out corresponding subtask v in the pairing local resource management layer
IRequest after, by the corresponding Service Instance of job factory service-creation, determine relevant hardware resource and software, start the corresponding calculated service.The term of execution of task, job management service also responds the query requests from the user simultaneously, and subtask v is provided
IThe execution intermediateness.Wait for corresponding subtask v
ICarry out after the end, job management service returns to collaborative distribution services with result of calculation.
Integrate a plurality of subtasks (subtask v that this high-performance calculation is used by collaborative distribution services
i, i=1,2 ..., 7) result of calculation, and be back to grid portal, by grid portal result of calculation is returned to the user at last.
Claims (7)
1, a kind of service-oriented high-performance grid computing job scheduling method, it is characterized in that, adopt service-oriented mode, grid system is made of the resource service group that distributes, feature according to the high-performance calculation application, the characteristics of collaborative process operation between the various computing website in the grid computing, a high-performance calculation application is decomposed into one group of calculating subtask that has the data input/output relation, represent the data dependence between each subtask in the high-performance calculation application with oriented no circular chart, and adopt dynamic priority scheduling algorithm to mate ready subtask and available resource service group simultaneously, selected subtask scheduling in selected resource service group, is realized the efficient scheduling of the high-performance calculation operation in the grid scope.
2, service-oriented high-performance grid computing job scheduling method according to claim 1 is characterized in that, comprises that the following step poly-:
(1) determines ready task { v among the oriented no circular chart DAG of operation
i}
Represent a plurality of subtasks that high-performance calculation is used with oriented no circular chart, and oriented no circular chart G=(V, E), vertex set V={v wherein
1, v
1..., v
mThe expression high-performance calculation uses a plurality of subtasks be decomposed, e
Ij=(v
i, v
j) ∈ E represents from subtask v
iTo subtask v
jCommunication, and | e
Ij| then represent the traffic, if subtask v in the oriented no circular chart
iAll preceding thread subtasks are all to carry out, and then this subtask is a ready task;
(2) obtain the real-time information { R of current available resource service groups
j}
Grid environment is defined as by a series of resource service group and constitutes, and the set of resource service group is expressed as R={R formally
1, R
2..., R
n, expense function C: V * R → R represents that one is calculated the time overhead that carry out the subtask, subtask v on the resource service group
iAt resource service group R
jThe time overhead of last execution is defined as C (v
i, R
j), by the information service in the inquiry grid system, obtain the real-time information { R of corresponding resource service group
j; (3) calculate
Adopt dynamic priority scheduling algorithm, dynamic priority is expressed as MDL (v
i, R
j, ∑ (t)), it has reflected under state set ∑ (t) situation, subtask v
iBe dispatched to resource service group R
jOn matching degree, state set ∑ (t) expression is the status information of all the resource service groups and communication resource during t constantly, dynamic priority MDL is defined as:
Wherein, t is the current time, and the first in the formula represents subtask v
iIn the static priority of task image, its value is from subtask v
iThe maximal value of execution time summation to the different paths of task image terminal point, t
d(v
i, R
j) be illustrated under state set ∑ (t) situation subtask v
iThe pot life the earliest of required all data, t
a(R
j) expression resource service group free time the earliest,
(4) determine to make the resource service group R that the MDL value is maximum
JWith ready task v
I
(5) scheduling ready task v
ITo resource service group R
J
(6) wait for scheduling events next time, and repeat said process.
3, service-oriented high-performance grid computing job scheduling method according to claim 2, it is characterized in that, in the described step (1), for a subtask, be not all can move in all the resource management territories in grid environment, some subtasks need corresponding scientific engineering computing storehouse, and this scientific engineering computing storehouse exists only on the corresponding calculated node.
4, service-oriented high-performance grid computing job scheduling method according to claim 2 is characterized in that, in the described step (2), for the resource service group that can not carry out certain subtask, defines its time expense for infinitely great.
5, according to claim 1 or 2 or 4 described service-oriented high-performance grid computing job scheduling methods, it is characterized in that, described resource service group, be meant: if a high-performance calculation application is divided into a plurality of subtasks, respectively by the corresponding calculated program solution, these subtasks belong to the computation-intensive task, therefore only can move a program copy on a computational resource, and the computational resource of this calculation procedure and its operation has promptly constituted the resource service group.
6, service-oriented high-performance grid computing job scheduling method according to claim 2 is characterized in that, in the described step (3), and Δ (v
i, R
j) be the difference of expression resource service group processing power, be defined as:
Δ(v
i，R
j)＝ C(v
i)-C(v
i，R
j)
Wherein, C (v
i) expression subtask v
iThe mean value of execution time on all resource service groups, the execution time is the maximal value of finite time value on all resource service groups if the mean value that calculates for infinitely great, then is taken at, A (v
iBut) expression subtasking v
iThe set of all available resources service groups.
7, service-oriented high-performance grid computing job scheduling method according to claim 6 is characterized in that, is calculating Δ (v
i, R
j) time, C (v
i) be defined as:
Wherein, ‖ A (v
i) ‖ represents set A (v
i) in the number of element.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200610026430 CN1845075A (en) | 2006-05-11 | 2006-05-11 | Service oriented high-performance grid computing job scheduling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200610026430 CN1845075A (en) | 2006-05-11 | 2006-05-11 | Service oriented high-performance grid computing job scheduling method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN1845075A true CN1845075A (en) | 2006-10-11 |
Family
ID=37064003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200610026430 Pending CN1845075A (en) | 2006-05-11 | 2006-05-11 | Service oriented high-performance grid computing job scheduling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1845075A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100456703C (en) * | 2007-08-02 | 2009-01-28 | 上海交通大学 | Electric power computation gridding application system |
| CN101692208B (en) * | 2009-10-15 | 2011-03-30 | 北京交通大学 | Task scheduling method for processing real-time traffic information |
| CN102364447A (en) * | 2011-10-28 | 2012-02-29 | 北京航空航天大学 | Operation scheduling method for optimizing communication energy consumption among multiple tasks |
| CN103092683A (en) * | 2011-11-07 | 2013-05-08 | Sap股份公司 | Scheduling used for analyzing data and based on elicitation method |
| CN103324534A (en) * | 2012-03-22 | 2013-09-25 | 阿里巴巴集团控股有限公司 | Operation scheduling method and operation scheduler |
| CN104142855A (en) * | 2013-05-10 | 2014-11-12 | 中国电信股份有限公司 | Dynamic task scheduling method and device |
| CN107038070A (en) * | 2017-04-10 | 2017-08-11 | 郑州轻工业学院 | The Parallel Task Scheduling method that reliability is perceived is performed under a kind of cloud environment |
| CN107273193A (en) * | 2017-04-28 | 2017-10-20 | 中国科学院信息工程研究所 | A kind of data processing method and system towards many Computational frames based on DAG |
| CN109561143A (en) * | 2018-11-26 | 2019-04-02 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Airborne SOA platform association sex service publication and selection method |
| CN111782389A (en) * | 2020-06-22 | 2020-10-16 | 中科边缘智慧信息科技（苏州）有限公司 | Task scheduling system and method under mobile edge information service network |
| CN112398911A (en) * | 2020-10-22 | 2021-02-23 | 成都中讯创新科技股份有限公司 | Multi-channel network scheduling method based on FC network |
| CN113485818A (en) * | 2021-08-03 | 2021-10-08 | 北京八分量信息科技有限公司 | Heterogeneous task scheduling method and device and related products |
-
2006
- 2006-05-11 CN CN 200610026430 patent/CN1845075A/en active Pending
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100456703C (en) * | 2007-08-02 | 2009-01-28 | 上海交通大学 | Electric power computation gridding application system |
| CN101692208B (en) * | 2009-10-15 | 2011-03-30 | 北京交通大学 | Task scheduling method for processing real-time traffic information |
| CN102364447A (en) * | 2011-10-28 | 2012-02-29 | 北京航空航天大学 | Operation scheduling method for optimizing communication energy consumption among multiple tasks |
| CN103092683A (en) * | 2011-11-07 | 2013-05-08 | Sap股份公司 | Scheduling used for analyzing data and based on elicitation method |
| CN103092683B (en) * | 2011-11-07 | 2017-12-26 | Sap欧洲公司 | For data analysis based on didactic scheduling |
| CN103324534A (en) * | 2012-03-22 | 2013-09-25 | 阿里巴巴集团控股有限公司 | Operation scheduling method and operation scheduler |
| CN104142855A (en) * | 2013-05-10 | 2014-11-12 | 中国电信股份有限公司 | Dynamic task scheduling method and device |
| CN104142855B (en) * | 2013-05-10 | 2017-07-07 | 中国电信股份有限公司 | The dynamic dispatching method and device of task |
| CN107038070B (en) * | 2017-04-10 | 2021-04-16 | 郑州轻工业学院 | Parallel task scheduling method for sensing execution reliability in cloud environment |
| CN107038070A (en) * | 2017-04-10 | 2017-08-11 | 郑州轻工业学院 | The Parallel Task Scheduling method that reliability is perceived is performed under a kind of cloud environment |
| CN107273193A (en) * | 2017-04-28 | 2017-10-20 | 中国科学院信息工程研究所 | A kind of data processing method and system towards many Computational frames based on DAG |
| CN109561143A (en) * | 2018-11-26 | 2019-04-02 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Airborne SOA platform association sex service publication and selection method |
| CN109561143B (en) * | 2018-11-26 | 2021-05-07 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Method for issuing and selecting relevance service of airborne SOA (service oriented architecture) platform |
| CN111782389A (en) * | 2020-06-22 | 2020-10-16 | 中科边缘智慧信息科技（苏州）有限公司 | Task scheduling system and method under mobile edge information service network |
| CN112398911A (en) * | 2020-10-22 | 2021-02-23 | 成都中讯创新科技股份有限公司 | Multi-channel network scheduling method based on FC network |
| CN112398911B (en) * | 2020-10-22 | 2022-07-15 | 成都中讯创新科技股份有限公司 | Multichannel network scheduling method based on FC network |
| CN113485818A (en) * | 2021-08-03 | 2021-10-08 | 北京八分量信息科技有限公司 | Heterogeneous task scheduling method and device and related products |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1845075A (en) | Service oriented high-performance grid computing job scheduling method | |
| Balaji et al. | Power aware virtual machine placement in IaaS cloud using discrete firefly algorithm | |
| CN100570569C (en) | Operation cross-domain control method under the grid computing environment | |
| CN104050042B (en) | The resource allocation methods and device of ETL operations | |
| CN101067758A (en) | Energy consumption management method for inserting system | |
| CN107341205A (en) | A kind of intelligent distribution system based on big data platform | |
| CN101034362A (en) | Method for implementing network job scheduling using mobile proxy | |
| CN101068157A (en) | Grid task dispatching method based on time and expense restriction | |
| Patni et al. | Load balancing strategies for grid computing | |
| CN104112049A (en) | P2P (peer-to-peer) architecture based cross-data-center MapReduce task scheduling system and P2P architecture based cross-data-center MapReduce task scheduling method | |
| Wang et al. | Integer particle swarm optimization based task scheduling for device-edge-cloud cooperative computing to improve SLA satisfaction | |
| He et al. | Energy-efficient framework for virtual machine consolidation in cloud data centers | |
| Jagadish Kumar et al. | Hybrid gradient descent golden eagle optimization (HGDGEO) algorithm-based efficient heterogeneous resource scheduling for big data processing on clouds | |
| Senger | Improving scalability of Bag-of-Tasks applications running on master–slave platforms | |
| Anan et al. | SLA-based optimization of energy efficiency for green cloud computing | |
| CN103268261A (en) | Hierarchical computing resource management method suitable for large-scale high-performance computer | |
| Wang et al. | A survey of system scheduling for hpc and big data | |
| Gao et al. | A dispatching-rule-based task scheduling policy for MapReduce with multi-type jobs in heterogeneous environments | |
| CN113448736A (en) | Task mapping method for approximate computation task on multi-core heterogeneous processing platform based on energy and QoS joint optimization | |
| Afgan et al. | Dynamic BLAST–a grid enabled BLAST | |
| Lu et al. | Grid load balancing scheduling algorithm based on statistics thinking | |
| Shih et al. | A performance-based approach to dynamic workload distribution for master-slave applications on grid environments | |
| Ng et al. | Grid enabled optimization | |
| Yu | Construction of load balancing scheduling model for cloud computing task based on chaotic ant colony algorithm | |
| CN1617126A (en) | Computer group system and its operation managing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |