CN109298919B - Multi-core scheduling method of soft real-time system for high-utilization-rate task set - Google Patents
Multi-core scheduling method of soft real-time system for high-utilization-rate task set Download PDFInfo
- Publication number
- CN109298919B CN109298919B CN201810980910.9A CN201810980910A CN109298919B CN 109298919 B CN109298919 B CN 109298919B CN 201810980910 A CN201810980910 A CN 201810980910A CN 109298919 B CN109298919 B CN 109298919B
- Authority
- CN
- China
- Prior art keywords
- tasks
- task
- utilization
- time
- real
- 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
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multi Processors (AREA)
Abstract
The invention relates to a multi-core scheduling method of a soft real-time system facing a high-utilization-rate task set, which is characterized in that the utilization rate of all tasks is calculated aiming at the real-time parameters of periodic tasks given by the soft real-time system under an embedded multi-core platform, the basis for judging that the tasks belong to a high-utilization-rate factor set is given according to the task utilization rate, a task set which can meet the condition that more than 2 processor utilization rates are full utilization rates is found out, then the tasks in the set are scheduled in real time, the real-time periodic tasks in different sets are distributed to be executed on processor cores in different groups during scheduling, the priority sequence during task scheduling is specified, and then the real-time periodic tasks are scheduled according to the corresponding real-time scheduling process. The invention not only reduces the task migration degree to the minimum, but also has high utilization rate of the system processor core.
Description
Technical Field
The invention belongs to the technical field of embedded system multi-core, and particularly relates to a multi-core scheduling method for a high-utilization-rate task set-oriented soft real-time system.
Background
Based on embedded system software with multiple cores, a process scheduling strategy is one of key technologies. How to design a proper scheduling algorithm under a multi-core platform is a focus of current research, so that the scheduling algorithm can meet the requirements of real-time performance and safety of critical tasks, and also fully utilizes strong computing power provided by multi-core. Originally, the static fixed priority real-time Scheduling algorithm based on the single core, such as RMS (Rate-monobonic Scheduling) and dms (dead mononic Scheduling), and the dynamic priority Scheduling algorithm, such as edf (early delay first) and mlf (minimum delay first), have all proved to be excellent Scheduling algorithms, but these algorithms cannot adapt to the requirement of multiple cores because they cannot implement parallel Scheduling. On the basis of a scheduling method of a multi-core architecture, real-time periodic task models are roughly divided into two main categories: one is a task scheduling model independent of each other, and the other is a task scheduling model with a dependency relationship. For the real-time periodic tasks which are independent of each other, the scheduling algorithm under the model has two main categories: a global scheduling algorithm and a task division scheduling algorithm. The global scheduling algorithm has the advantages of high core utilization rate and the disadvantages of thread migration for load balancing, and the migration cost is high. The task division scheduling algorithm has the advantages of achieving load balance and enabling the core utilization rate to be low. Therefore, what method is adopted to achieve both the core utilization rate improvement and the load balancing is a focus of research in the industry. To this end, current researchers employ two approaches: one approach is to consider task partitioning as a binning problem and set up various task partitioning conditions by employing heuristic methods to improve core utilization. However, how to set up task division conditions to optimize the core utilization rate becomes a difficult problem; the other is to combine task partitioning with global to reduce migration overhead. Most typically, such as the semi-partitioned scheduling algorithm, the general idea of this class of algorithms is to divide the scheduling algorithm into two parts: partitioning and scheduling. In the partitioning phase, most tasks are assigned to one of the processors, such tasks being called non-partitionable tasks (non-split tasks); while other small parts of the task may be split into two or more sub-tasks and distributed to different processors, called split tasks. And once the partition phase terminates, the scheduling phase begins executing the corresponding tasks and subtasks on the designated processor. During run time, each partitioned task will be migrated to a different processor. The semi-division scheduling algorithm is superior to a simple task division and a global scheduling algorithm in theory and practical application, but how to divide tasks into division tasks and division tasks into non-divisible tasks can improve the core utilization rate and make load balance and low cost difficult. Although the existing half-partition algorithm has advantages, no good solution exists for a real-time system with high execution rate to completely meet the requirements.
Disclosure of Invention
The application provides a multi-core scheduling method for a high-utilization-rate task set-oriented soft real-time system, which solves the problem of real-time scheduling of periodic tasks in the soft real-time system with tasks all belonging to a high-utilization-rate factor set under an embedded multi-core platform, and reduces task switching overhead and context switching times caused by migration in the prior art.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a multi-core scheduling method for a soft real-time system facing a high-utilization-rate task set is characterized in that the utilization rates of all tasks are calculated according to real-time parameters of periodic tasks given by the soft real-time system under an embedded multi-core platform, a basis for judging that the tasks belong to a high-utilization-rate factor set is given according to the task utilization rates, a task set which can meet the condition that more than 2 processor utilization rates are full utilization rates is found out, then the tasks in the set are scheduled in real time, the real-time periodic tasks in different sets are distributed to processor cores in different groups to be executed during scheduling, the priority sequence during task scheduling is specified, and then the real-time periodic tasks are scheduled according to corresponding real-time scheduling processes.
Further, the method specifically comprises the following steps:
if yes, then finding out the minimum K from the rest task utilization factorminDividing the full-utilization real-time periodic tasks into a group of tasks, wherein K is more than or equal to 2min≤M-KminAnd assigning the set of tasks to K by a split task methodminRepeatedly judging M-K on each processor coreminWhether the value of (d) is greater than or equal to 2;
if not, dividing the rest tasks into one group, distributing the rest tasks to the rest 2 processor cores by a task dividing method, and finishing the scheduling method.
Further, the task segmentation method specifically includes the following steps:
step 101, arranging the tasks in the queue according to the order of the utilization rate, and finding out the task tau with the maximum utilization rateiAssigning and deleting the jth ═ 1 processor core from the queue;
step 102, if j is less than or equal to KminCalculating the residual utilization factor U of the jth processorjJudgment of Uj>Whether 0 is true;
if U is presentj>If 0 is true, the task τ with the maximum utilization will be selectediThe division value is calculated according to the following formula:will tauiWorst execution time e ofiAccording to fi,jIs divided into the jth processor core, i.e., e is executed on the jth corei×fi,jLength of time of (e), remaining ofi×(1-fi,j) The execution time of (2) is divided to other processors;
if U is presentjIf the result is more than 0, the j is added with 1 to return after the j processor is distributed;
if j is>KminAnd then, the situation that the group of tasks are distributed completely and the corresponding processor cores of the distributed tasks are returned is explained.
The invention has the beneficial effects that:
the invention defines a semi-partitioned scheduling method based on minimum mobility aiming at the periodic tasks of the soft real-time system belonging to the high utilization factor set. The method is simple and easy to implement, not only can achieve high utilization rate and load balance, but also can reduce unnecessary task switching overhead, and has wide application prospect and important value in the safety field of embedded equipment, particularly in some airborne fields.
Drawings
FIG. 1 is a result of an EDF-fm scheduling algorithm distributed over four cores in the prior art;
FIG. 2 is a result of an EDF-os scheduling algorithm distributed over four cores in the prior art;
FIG. 3 is a main flow diagram of the method of the present invention;
FIG. 4 is a flow chart of a task segmentation algorithm in the method of the present invention;
FIG. 5 is a flow chart of the process of queuing up the full utilization rate of a task K in the method of the present invention
FIG. 6 is the result of a set of embodiments distributing over four cores under the method of the present invention;
FIG. 7 is a set of embodiments of execution on four cores under the method of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. Wherein like elements in different embodiments are numbered with like associated elements. In the following description, numerous details are set forth in order to provide a better understanding of the present application. However, those skilled in the art will readily recognize that some of the features may be omitted or replaced with other elements, materials, methods in different instances. In some instances, certain operations related to the present application have not been shown or described in detail in order to avoid obscuring the core of the present application from excessive description, and it is not necessary for those skilled in the art to describe these operations in detail, so that they may be fully understood from the description in the specification and the general knowledge in the art.
The invention firstly provides a task scheduling model of a multi-core real-time system and a definition of real-time periodic tasks belonging to a high-utilization-rate task set. Secondly, aiming at real-time periodic Tasks of the system, an Earliest-delay-First-based scheduling and Split Tasks (EDF-MSTL) scheduling method based on minimum migration degree and splitting degree is provided, the method comprehensively considers the number of the Split Tasks, the Tasks with high utilization rate are distributed to the same processor core as much as possible, and the migration number of the Split Tasks among different processors is reduced to the minimum.
1. Multi-core real-time system task scheduling model
A typical real-time system consists of several processes with time constraints, called tasks. In most real-time systems, tasks occur repeatedly, i.e., each task is activated periodically. Each activation is referred to as a job (job). Each job is required to be completed within a specific time period. A task is one that persists for a long time and can be activated an unlimited number of times, if not specified otherwise. The earliest periodic task models were proposed by Liu and Layland, which are described in detail as follows:
let a real-time system t be composed of n periodic tasks, denoted as τ ═ τ1,τ2LL,τn}. Each periodic task contains four parameters. I.e. taui(ri,ei,pi,di) (1. ltoreq. i. ltoreq. n), where riIndicating release time (release time), which is the time that the processor core can respond; e.g. of the typeiRepresenting a task TiWorst Case Execution Time (Worst-Case Execution Time, WCET); p is a radical ofiIs τiThe cycle time of (a); diIndicating a time limit (deadline). Herein requires ei≤di≤pi. If d isi=piThe system is called an implicit deadlines (implicit deadlines) system, and the task is called an implicit deadlines task. If d isi<piThis system is called a constrained thread system (dts) system, and its tasks are called including-timed tasks. An arbitrary time limit (arb) if both have no mandatory constraintsAn iterative decode line system). Usually by τi(ri,ei,pi,di) Denotes an inclusion or arbitrary time-limited system, andi(ri,ei,pi) Indicating an implicit time-limited system.
In a periodic task system, all tasks issue their first jobs at the same time, which is called synchronous system, otherwise called asynchronous system. Tau isiJ.gtoreq.1, j.i,jAnd (4) showing. The first job of a task can be published at any time. Will Ji,jIs recorded as ri,jIts absolute time limit di,jIs defined as ri,j+di。Ji,jIs recorded as ei,j。
The periodic task model can only effectively model time-driven processes and cannot effectively model processes with uncertain period intervals caused by external events. For example, in a radar real-time tracking system, one or more tasks are involved, and these tasks are repeatedly activated, each time the activation is required to be performed within a specific time. Each activation of a task is determined by an interactive event generated by a system in the external environment, rather than a fixed cycle time, and this type of task model is called a Sporadic task model. The Sporadic task model differs from the periodic model in that the two consecutive jobs issue time intervals are not fixed, with the minimum time interval becoming the period of the task.
The Sporadic task model refers to n repeatedly occurring tasks τ ═ { τ ═ τ1,τ2LL,τnEach tauiThere are three parameter features: e.g. of the typei(e i0 or more) indicates WCET; period pi>eiIndicates a τiMinimum interval between two successive jobs. Time limit di≥eiIndicates τiAfter it is issued to the maximum time interval when it is completed. Tau isiIs executed sequentially, and only one job is executed at the same time. The periodic task model can be regarded as a special case of the Sporadic task model, namely the taskThe continuous job release in the service is fixed by piThe time unit is divided. The present invention focuses on a system where both periodic tasks and Sporadic tasks exist with implicit deadlines.
The multi-core model is described as follows, P ═ { P ═ P1,P2LL,PMIs a set of M processor cores with the same processing power. Within a certain time period, to a certain processor PmTask τ oniIs activated, the time slice of the j-th job of the task executed on the processor is marked as Ti,j,m。
From the above, the periodic task in the real-time system has four important attributes-the release time riTime limit diWorst execution time eiAnd period pi。
2. Correlation definitions, theorems and conclusions
Define 2 a real-time system for a given set of periodic tasks, ifThe system is said to be a single full utilization system for the set of periodic tasks.
Define 3 a real-time system τ for a given set of periodic tasks, ifWherein M is the number of processor cores of the system, ifThen for the set of periodic tasks, the system is referred to as a K-full utilization system.
Define 4 a real-time system for a given set of periodic tasks, ifAnd if M is the number of the processor cores of the system, the system is called a full-utilization system for the group of periodic tasks.
Theorem for existence task utilization U (tau)i) S of e (0.5,1)HA system, if it is a fully-utilized system, then a necessary condition that the system can be scheduled is that there must be a task that can be partitioned.
And (3) proving that: let τ be { τ1,τ2LL,τnIs one SHN periodic tasks in the system and for a fully-utilized systemSuppose a task τ thereiniFactor of utilization ofThen the sum of the utilization factors of the task and any one task is greater than 1, and in this case, if the system needs to be scheduled, the task cannot share the processor with other tasks and needs to be allocated to one processor core separately, but because of the full utilization system, the sum of the utilization factors of other tasks will occur
The system composed of other tasks becomes an over-full utilization system, and therefore scheduling cannot be achieved. Thus, the system requires tasks that can be segmented if they can be scheduled.
3. High utilization task set example
In a system with a high utilization task set, real-time periodic tasks exist as shown in table 1. It is assumed that each task can be made a split task.
TABLE 1 high utilization task set example
The tasks in the above example are distributed to four processor cores according to the existing EDF-fm (early-delay-First-base-selected-or-scheduling) algorithm and EDF-os (early-delay-First-base-selected-time-base-scheduled) scheduling algorithm in the prior art, as shown in fig. 1 and fig. 2. As can be seen from FIG. 1, the task that is partitioned in the system is τ2,τ3,τ5The three tasks are migrated 1 time respectively, so the migration degree under the scheduling algorithmDegree of task segmentation
The task reordering according to EDF-os is shown in table 2, and the corresponding scheduling result is shown in fig. 2.
TABLE 2 EDF-os Algorithm scheduling sequences
As can be seen from FIG. 2, the task that is partitioned in the system is τ5,τ6The two tasks are migrated 2 times and 1 time respectively, so the migration degree under the scheduling algorithmDegree of task segmentation
The scheduling algorithm designed by the invention adopts a scheduling method (Earliest-delay-First-based scheduling and Split Tasks, EDF-MSTL) based on minimum mobility and segmentation, and the main flow of the scheduling method is as follows:
1. firstly, finding out the utilization factor capable of meeting the minimum K from all tasksmin(2≤KminLess than or equal to M, wherein M is the number of processor cores) into a group of tasks and distributing the group of tasks to K by a task partitioning methodminAnd a processor.
2. Judgment of M-KminIf the value of (b) is greater than or equal to 2, then finding the smallest K that can be satisfied from the remaining task utilization factorsmin(2≤Kmin≤M-Kmin) The full-utilization real-time period tasks are divided into a group of tasks, and the group of tasks are distributed to K processor cores by a task dividing method. And (5) repeating the step (2).
If not, dividing the rest tasks into one group, and distributing the tasks to the rest 2 processor cores by a task dividing method. The scheduling algorithm ends. The specific flow chart is shown in fig. 3.
Wherein a given set of tasks is assigned to KminOn each processor core, the task partitioning algorithm is as follows:
1. firstly, arranging the tasks in a queue according to the sequence of the utilization rate, and finding out the task tau with the maximum utilization rateiThe jth 1 processor core is assigned and removed from the queue.
2. If j is less than or equal to KminComputing the jth processor according to definition 8Remaining utilization factor UjJudgment of Uj>If 0 is true, then the task τ with the maximum utilization will be selectediThe division value is calculated according to the following formula:will tauiWorst execution time e ofiAccording to fi,jIs divided to the jth processor core. I.e. executing e on the jth corei×fi,jLength of time of (e), remaining ofi×(1-fi,j) Is divided over the other processors. If U is presentj>If 0 is not established, the j indicates that the j processor is already allocated, and j is added with 1 and returns; if j is>KminAnd then, the situation that the group of tasks are distributed completely and the corresponding processor cores of the distributed tasks are returned is explained. The specific flow chart is shown in fig. 4.
Results of example testing
According to the scheduling method, the test cases are scheduled, and two groups of 2 full-utilization tasks can be obtained: as shown in table 3. The result of the scheduling correspondingly allocated to the 4 processor cores is shown in fig. 6.
TABLE 3 MSTL Algorithm scheduling sequences
It can be seen from fig. 6 that the task set is divided into two sets with 2 full utilization rates, one of the three tasks in each set can be divided as a split task and allocated to two processor cores, the number of task migrations in a common period is only 2, other tasks are all allocated to the same processor core, and the corresponding migration degree under the scheduling algorithmDegree of task segmentationSystem overviewThe degree of migration and task segmentation of (2) is minimal. The case where the corresponding tasks are executed on four cores in one common cycle is shown in fig. 7.
The present invention has been described in terms of specific examples, which are provided to aid understanding of the invention and are not intended to be limiting. For a person skilled in the art to which the invention pertains, several simple deductions, modifications or substitutions may be made according to the idea of the invention.
Claims (1)
1. A multi-core scheduling method of a soft real-time system facing a high-utilization-rate task set is characterized in that the utilization rate of all tasks is calculated according to real-time parameters of periodic tasks given by the soft real-time system under an embedded multi-core platform, the basis for judging that the tasks belong to a high-utilization-rate factor set is given according to the task utilization rate, a task set which can meet the condition that more than 2 processor utilization rates are full utilization rates is found out, then the tasks in the set are scheduled in real time, the real-time periodic tasks in different sets are distributed to processor cores in different groups to be executed during scheduling, the priority sequence during task scheduling is specified, and then the real-time periodic tasks are scheduled according to corresponding real-time scheduling processes;
the method specifically comprises the following steps:
step 1, from the utilization factor of all tasks, finding out the task with the maximum utilization factor, dividing the task into a high utilization factor set, then finding out the utilization factor of other tasks and the task with the high utilization factor to judge the full utilization of the processor, namely finding out the task which can meet the minimum KminDividing the real-time periodic tasks with full utilization rate into a group of tasks, wherein K is more than or equal to 2minM is less than or equal to M, M is the number of processor cores, and the group of tasks with high utilization rate and other tasks are distributed to K through a task partitioning methodminA processor;
step 2, judging M-KminWhether the value of (d) is greater than or equal to 2;
if yes, then finding out the minimum K from the rest task utilization factorminDividing the full-utilization real-time periodic tasks into a group of tasks with the utilization rate less than or equal to 2Kmin≤M-KminAnd assigning the set of tasks to K by a split task methodminRepeatedly judging M-K on each processor coreminWhether the value of (d) is greater than or equal to 2;
if not, dividing the rest tasks into a group, distributing the group to the rest 2 processor cores by a task dividing method, and ending the scheduling method;
the task segmentation method specifically comprises the following steps:
step 101, arranging the tasks in the queue according to the order of the utilization rate, and finding out the task tau with the maximum utilization rateiAssigning and deleting the jth ═ 1 processor core from the queue;
step 102, if j is less than or equal to KminCalculating the residual utilization factor U of the jth processorjJudgment of Uj>Whether 0 is true;
if U is presentj>If 0 is true, the task τ with the maximum utilization will be selectediThe division value is calculated according to the following formula:will tauiWorst execution time e ofiAccording to fi,jIs divided into the jth processor core, i.e., e is executed on the jth corei×fi,jLength of time of (e), remaining ofi×(1-fi,j) The execution time of (2) is divided to other processors;
if U is presentj>If 0 is not established, the j indicates that the j processor is already allocated, and j is added with 1 and returns;
if j is>KminAnd then, the situation that the group of tasks are distributed completely and the corresponding processor cores of the distributed tasks are returned is explained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810980910.9A CN109298919B (en) | 2018-08-27 | 2018-08-27 | Multi-core scheduling method of soft real-time system for high-utilization-rate task set |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810980910.9A CN109298919B (en) | 2018-08-27 | 2018-08-27 | Multi-core scheduling method of soft real-time system for high-utilization-rate task set |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109298919A CN109298919A (en) | 2019-02-01 |
CN109298919B true CN109298919B (en) | 2021-09-07 |
Family
ID=65165454
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810980910.9A Active CN109298919B (en) | 2018-08-27 | 2018-08-27 | Multi-core scheduling method of soft real-time system for high-utilization-rate task set |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109298919B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110879746B (en) * | 2019-11-11 | 2023-03-31 | 西北工业大学 | High-safety key task scheduling method based on inter-core migration |
KR20210101071A (en) | 2020-02-07 | 2021-08-18 | 삼성전자주식회사 | Electronic device for scheduling based on heterogeneous multi-processor and operating method thereof |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101256515A (en) * | 2008-03-11 | 2008-09-03 | 浙江大学 | Method for implementing load equalization of multicore processor operating system |
CN105700941A (en) * | 2015-12-18 | 2016-06-22 | 西安工业大学 | Three-dimensional scheduler model and scheduling algorithm thereof |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7725897B2 (en) * | 2004-11-24 | 2010-05-25 | Kabushiki Kaisha Toshiba | Systems and methods for performing real-time processing using multiple processors |
US8510741B2 (en) * | 2007-03-28 | 2013-08-13 | Massachusetts Institute Of Technology | Computing the processor desires of jobs in an adaptively parallel scheduling environment |
-
2018
- 2018-08-27 CN CN201810980910.9A patent/CN109298919B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101256515A (en) * | 2008-03-11 | 2008-09-03 | 浙江大学 | Method for implementing load equalization of multicore processor operating system |
CN105700941A (en) * | 2015-12-18 | 2016-06-22 | 西安工业大学 | Three-dimensional scheduler model and scheduling algorithm thereof |
Non-Patent Citations (3)
Title |
---|
Base-Utilization partitioning algorithm for multiprocessors;Huang Shujuan;《2011 IEEE International Conference on Computer Science and Automation Engineering》;20110612;全文 * |
pCOMPATS: Period-Compatible Task Allocation and Splitting on Multi-core Processors;Arvind Kandhalu等;《2012 IEEE 18th Real Time and Embedded Technology and Applications Symposium》;20120517;全文 * |
基于利用率和负载均衡的多核实时调度算法研究;黄姝娟等;《西北工业大学学报》;20120229;正文第1-7页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109298919A (en) | 2019-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9916183B2 (en) | Scheduling mapreduce jobs in a cluster of dynamically available servers | |
Ghazalie et al. | Aperiodic servers in a deadline scheduling environment | |
CN102591712B (en) | Decoupling parallel scheduling method for rely tasks in cloud computing | |
US8812639B2 (en) | Job managing device, job managing method and job managing program | |
CN103927225A (en) | Multi-core framework Internet information processing and optimizing method | |
Kang et al. | Lalarand: Flexible layer-by-layer cpu/gpu scheduling for real-time dnn tasks | |
WO2012028213A1 (en) | Re-scheduling workload in a hybrid computing environment | |
Dubey et al. | A priority based job scheduling algorithm using IBA and EASY algorithm for cloud metaschedular | |
CN109298919B (en) | Multi-core scheduling method of soft real-time system for high-utilization-rate task set | |
Liu et al. | Supporting soft real-time parallel applications on multicore processors | |
US20210390405A1 (en) | Microservice-based training systems in heterogeneous graphic processor unit (gpu) cluster and operating method thereof | |
Alhussian et al. | An unfair semi-greedy real-time multiprocessor scheduling algorithm | |
Azmi et al. | Scheduling grid jobs using priority rule algorithms and gap filling techniques | |
Liu et al. | Scheduling Parallel Jobs Using Migration and Consolidation in the Cloud. | |
Lemerre et al. | Equivalence between schedule representations: Theory and applications | |
Sirohi et al. | Improvised round robin (CPU) scheduling algorithm | |
Qin et al. | Dependent task scheduling algorithm in distributed system | |
KR101639947B1 (en) | Hadoop preemptive deadline constraint scheduling method, execution program thereof method and recorded medium of the program | |
Alhussian et al. | An efficient real-time multiprocessor scheduling algorithm | |
Azhen et al. | Adaptive real-time scheduling for mixed task sets based on total bandwidth server | |
Yi et al. | An effective algorithm of jobs scheduling in clusters | |
Li et al. | H-pfsp: Efficient hybrid parallel pfsp protected scheduling for mapreduce system | |
Zhao et al. | Effective straggler mitigation with cross-layer interference-aware optimization | |
Rajguru et al. | A performance analysis of task scheduling algorithms using qualitative parameters | |
Lin et al. | LAG-Based Analysis for Preemptive Global Scheduling with Dynamic Cache Allocation |
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 |