CN115328647A - Compute intensive service elastic expansion method based on task queue - Google Patents

Compute intensive service elastic expansion method based on task queue Download PDF

Info

Publication number
CN115328647A
CN115328647A CN202210871366.0A CN202210871366A CN115328647A CN 115328647 A CN115328647 A CN 115328647A CN 202210871366 A CN202210871366 A CN 202210871366A CN 115328647 A CN115328647 A CN 115328647A
Authority
CN
China
Prior art keywords
task
instances
time
tasks
queue
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
Application number
CN202210871366.0A
Other languages
Chinese (zh)
Inventor
于劲松
梁思远
周金浛
唐荻音
周倜
苗毅
陶来发
刘浩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihang University
Original Assignee
Beihang University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beihang University filed Critical Beihang University
Priority to CN202210871366.0A priority Critical patent/CN115328647A/en
Publication of CN115328647A publication Critical patent/CN115328647A/en
Pending legal-status Critical Current

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]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • 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
    • G06F9/54—Interprogram communication
    • G06F9/546—Message passing systems or structures, e.g. queues
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/54—Indexing scheme relating to G06F9/54
    • G06F2209/548—Queue
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a flexible scaling method for computationally intensive services based on task queues, for flexible scaling of computationally intensive but delay-tolerant services. The method comprises (S101) storing independently executable computing tasks in a task queue; (S102) forming a task time consumption estimation mapping; (S103) estimating the time consumption of the task; (S104) increasing and decreasing instances and predicting the on-time completion condition of the task to obtain the minimum number of the instances; (S105) increasing or decreasing the instances according to the minimum number of instances. The compute-intensive service elastic expansion method based on the task queue provided by the invention arranges the tasks in a queue form, provides the minimum number of instances within the range of meeting the specified delay requirement by combining prediction and adjustment, saves computing resources to the greatest extent while meeting the business requirement, and meets the compute-intensive service elastic expansion requirement.

Description

Compute intensive service elastic expansion method based on task queue
Technical Field
The invention relates to a computer resource elastic expansion method, in particular to a calculation intensive service elastic expansion method based on a task queue.
Background
In order to satisfy business requirements under the condition of saving computer resources as much as possible, a computer system or a computer cluster needs to automatically adjust the number of service instances used for business according to a certain policy, which is called elastic expansion. When the business requirement is increased, the elastic expansion automatically increases the service instance, and more computing resources are distributed to meet the requirement of processing the business requirement; when the business requirement is reduced, the elastic expansion automatically reduces the service instances, and releases the occupied computing resources to other applications, so as to realize the efficient utilization of the resources. The existing elastic expansion method comprises the following steps: 1) The method based on the alarm is that the occupancy rates of resources such as a CPU of the service instance are continuously monitored, if the occupancy rates are continuously too high, the service instance is increased, and if the occupancy rates are continuously too low, the service instance is decreased; 2) A planning-based approach analyzes and predicts peak periods of traffic, increases instances of service during peak periods, and decreases instances of service at other times.
IO intensive services can use the two resilient scaling methods described above, such as Web services. In this type of service, waiting for reads and writes accounts for most of the traffic processing, and the CPU duty of a single thread is not high. Therefore, they often employ multithreading to increase CPU utilization. When the service processing load is increased, the CPU occupancy rate is increased, so that the alarm-based method can realize timely and effective elastic expansion and contraction by continuously monitoring the occupancy rate of CPU resources. On the other hand, the Web service logic is simple, the single service is processed quickly, the load is mainly related to the number of services such as user access amount, and the time and the size of the peak value of the user access amount are analyzed and estimated, and the elastic expansion and contraction can be realized timely and effectively by applying a plan-based method.
However, computationally intensive services cannot use both of the elastic scaling methods described above. In the computation-intensive services such as regular log analysis, data mining, machine learning model training and the like, the operation occupies most of the time of business processing, and the CPU continuously runs in a high proportion and is irrelevant to the business load. Thus, the alarm-based approach fails to increase or decrease service instances by continuously monitoring the occupancy of CPU resources. Also, the individual traffic handling time of a computationally intensive service is difficult to estimate, and thus a plan-based approach is also difficult to allocate service instances by analyzing and predicting traffic peaks. Therefore, how to reasonably perform elastic scaling is the key to the computationally intensive service to reasonably allocate computer resources.
Disclosure of Invention
The invention provides a task queue-based elastic expansion method for compute-intensive services, which adopts a prediction method to provide the minimum number of instances meeting the specified business processing requirements and can be suitable for the elastic expansion of the compute-intensive services.
In order to achieve the purpose, the invention proposes the following technical scheme:
the compute-intensive service elastic scaling method based on the task queue comprises the following steps:
(S101) storing independently executable computing tasks in a task queue, wherein each task is given a latest allowed completion time, and the tasks are sorted according to the latest allowed completion time in the queue;
(S102) recording the historical task time consumption, the task type and the quantized values of the task time consumption influence factors to form a task time consumption estimation mapping;
(S103) estimating task time consumptions of all tasks being executed and in the queue by using the task time consumption estimation map;
(S104) predicting the predicted completion time of all tasks according to the last task completion time and all task consumed time of all current examples;
(S105) increasing and decreasing instances and predicting the task on-time completion condition until the minimum number of instances for enabling the tasks in all the queues to be completed on time is obtained;
(S106) increasing or decreasing the instances according to the minimum number of instances.
Specifically, the task time consumption estimation mapping in the step (S102) is formed by performing multiple linear regression on different types of calculation tasks.
Further, the step (S105) includes: if a task with the predicted completion time later than the latest allowed completion time exists in the prediction, circularly executing and adding one instance for predicting again until the predicted completion time of all the tasks in the task queue is earlier than the latest allowed completion time, and taking the minimum number of instances as the number of instances when the circulation exits; otherwise, the loop execution reduces one instance to predict again until there is a task whose predicted completion time is later than the latest allowed completion time, and the minimum number of instances is taken as the number of instances at the loop exit plus one.
The beneficial effect of this application is:
the compute-intensive service elastic expansion method based on the task queue provided by the invention arranges the tasks in a queue form, provides the minimum number of instances within the range of meeting the specified delay requirement by combining prediction and adjustment, saves computing resources to the greatest extent while meeting the business requirement, and meets the compute-intensive service elastic expansion requirement.
Drawings
FIG. 1 is a flowchart of a compute intensive service elastic scaling method based on task queues according to an embodiment of the present invention;
FIG. 2 is a flowchart of a method for forming a task time consumption estimation map according to an embodiment of the present invention;
fig. 3 is a flowchart of a method for predicting predicted completion time of all tasks according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and not by way of limitation with respect to the scope of the invention.
In order that the above objects, features and advantages of the present application can be more clearly understood, the present invention will be further described in detail with reference to the accompanying drawings and examples. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. The specific embodiments described herein are offered by way of illustration and disclosure only and are not intended to limit the present application. All other embodiments that can be derived by one of ordinary skill in the art from the description of the embodiments are intended to be within the scope of the present disclosure.
One embodiment of the invention is applied to data mining services, and the service types comprise threshold mining, cluster analysis and prediction analysis. The time consumption of the threshold value mining task is only related to the sample size, the time consumption of the clustering analysis task is related to the sample size and the sample dimension, and the time consumption of the prediction analysis task is related to the prediction step number.
As shown in fig. 1, the compute intensive service elastic scaling method based on task queue includes the following steps:
(S101) storing independently executable computing tasks in a task queue, wherein each task is given a latest allowed completion time, and the tasks are sorted according to the latest allowed completion time in the queue;
(S102) recording the historical task time consumption, the task type and the quantized values of the task time consumption influence factors to form a task time consumption estimation mapping;
(S103) estimating task time consumptions of all tasks being executed and in the queue by using the task time consumption estimation map;
(S104) predicting the predicted completion time of all tasks according to the last task completion time and all task consumed time of all current examples;
(S105) increasing and decreasing instances and predicting the task on-time completion condition until the minimum number of instances for enabling the tasks in all the queues to be completed on time is obtained;
(S106) increasing or decreasing the instances according to the minimum number of instances.
Specifically, in step (S101), the storage information of the computing task in the queue includes: 1. task type A, threshold mining G 0 Cluster analysis G 1 Or predictive analysis G 2 (ii) a 2. The quantized values of the time-consuming influence factors of the tasks are mined by the threshold value as the sample size n 0 Cluster analysis as a sample size n 1 And sample dimension N 1 The prediction analysis is the number of predicted steps s 2 (ii) a 3. The latest allowable completion time t of the task p . When a computing task is inserted into the queue, according to t p Performing sequencing insertion to make t in the task queue p The smallest task is at the head of the queue, t p The largest task is at the end of the queue.
Concretely, the steps(S102), using the time T of the ith historical task i Task type A i Factors influencing task time consumption
Figure BDA0003760825160000031
The flow of the mapping of the quantized values to the task time consumption estimation is shown in fig. 2, and comprises the following steps:
(S201) according to the task type A i Classifying the historical task information to form a historical task information set:
Figure BDA0003760825160000032
(S202) separately applying the least squares method to S 0 ,S 1 ,S 2 Multiple linear model of (T = a) 0 n 0 +b 0 ,
T＝a 1 b 1 +b 1 N 1 +c 1 ,T＝a 2 s 2 +b 2 Performing regression to obtain a 0 ,b 0 ,a 1 ,b 1 ,c 1 ,a 2 ,b 2 In the formula, T is
(S203) establishing a task time consumption estimation mapping
Figure BDA0003760825160000033
Wherein
Figure BDA0003760825160000034
Estimation of task time consumption. The specific mapping mode is as follows:
Figure BDA0003760825160000035
specifically, in step (S104), the values of the influencing factors of the currently executed and the tasks in the task queue are substituted
Figure BDA0003760825160000041
Figure BDA0003760825160000042
In the method, the estimated task time consumption of the current execution task is obtained
Figure BDA0003760825160000043
And estimated task time in task queue
Figure BDA0003760825160000044
Specifically, in step (S104), the task completion time set at the jth instance in the current M service instances is
Figure BDA0003760825160000045
The current time is t now Predicting the predicted completion time of all tasks, wherein the prediction process is shown in figure 3, and the specific steps comprise:
(S301) calculating the predicted completion time of the current task for all instances
Figure BDA0003760825160000046
(S302) setting the current task as the head task of the task queue, and recording the estimated time consumption of the task as
Figure BDA0003760825160000047
(S303) finding the example with the minimum predicted completion time of the current task, and recording the predicted completion time of the current task of the example as
Figure BDA0003760825160000048
(S304) updating the predicted completion time
Figure BDA0003760825160000049
Noting the updated time as the predicted completion time of the task being traversed
Figure BDA00037608251600000410
(S305) if the task queue is traversed completely, ending the circulation, otherwise, setting the current taskThe task is the next task in the queue, and the estimated time consumption of the task is recorded as
Figure BDA00037608251600000411
Step (S303) is performed.
Specifically, in step (S105), if there is a task whose predicted completion time is later than the latest permitted completion time in the prediction, the loop execution adds an instance to predict again until the predicted completion times of all tasks in the task queue are earlier than the latest permitted completion time, and the minimum number of instances is taken as the number of instances at the time of loop exit; otherwise, the loop execution reduces one instance to predict again until there is a task whose predicted completion time is later than the latest allowed completion time, and the minimum number of instances is taken as the number of instances at the loop exit plus one.
In the method for computing intensive service elastic stretching based on the task queue described in this embodiment, the tasks are arranged in a queue form, the minimum number of instances within the range of meeting the specified delay requirement is given by combining prediction and adjustment, the computing resource is saved to the greatest extent while the business requirement is met, and the computing intensive service elastic stretching requirement is met.
Although the embodiments of the present application have been described in conjunction with the accompanying drawings, those skilled in the art will be able to make various modifications and variations without departing from the spirit and scope of the application, and such modifications and variations are included in the specific embodiments of the present invention as defined in the appended claims, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of various equivalent modifications and substitutions within the technical scope of the present disclosure, and these modifications and substitutions are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (3)

1. A compute-intensive service elastic scaling method based on a task queue is characterized by comprising the following steps of:
(S101) storing independently executable computing tasks in a task queue, wherein each task is given a latest allowed completion time, and the tasks are sorted according to the latest allowed completion time in the queue;
(S102) recording the historical task time consumption, the task type and the quantized values of the task time consumption influence factors to form a task time consumption estimation mapping;
(S103) estimating task time consumptions of all tasks being executed and in the queue by using the task time consumption estimation map;
(S104) predicting the predicted completion time of all tasks according to the completion time of the last task and the consumed time of all tasks of all current examples;
(S105) increasing and decreasing instances and predicting the task on-time completion condition until the minimum number of instances for enabling the tasks in all the queues to be completed on time is obtained;
(S106) increasing or decreasing the instances according to the minimum number of instances.
2. The method as claimed in claim 1, wherein the task-queue-based elastic scaling method for compute-intensive services is characterized in that the task-consumption estimation mapping in step (S102) is formed by respective multiple linear regression for different types of computation tasks.
3. The method of claim 1, wherein in step (S105), if there is a task whose predicted completion time is later than the latest permitted completion time in the prediction, the loop execution adds an instance to predict again until the predicted completion time of all tasks in the task queue is earlier than the latest permitted completion time, and the minimum number of instances is determined as the number of instances at the time of loop exit; otherwise, the loop execution reduces one instance to predict again until there is a task whose predicted completion time is later than the latest allowed completion time, and the minimum number of instances is taken as the number of instances at the loop exit reduced by one.
CN202210871366.0A 2022-07-22 2022-07-22 Compute intensive service elastic expansion method based on task queue Pending CN115328647A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210871366.0A CN115328647A (en) 2022-07-22 2022-07-22 Compute intensive service elastic expansion method based on task queue

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210871366.0A CN115328647A (en) 2022-07-22 2022-07-22 Compute intensive service elastic expansion method based on task queue

Publications (1)

Publication Number Publication Date
CN115328647A true CN115328647A (en) 2022-11-11

Family

ID=83919746

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210871366.0A Pending CN115328647A (en) 2022-07-22 2022-07-22 Compute intensive service elastic expansion method based on task queue

Country Status (1)

Country Link
CN (1) CN115328647A (en)

Similar Documents

Publication Publication Date Title
JP5946068B2 (en) Computation method, computation apparatus, computer system, and program for evaluating response performance in a computer system capable of operating a plurality of arithmetic processing units on a computation core
CN109324875B (en) Data center server power consumption management and optimization method based on reinforcement learning
US7844968B1 (en) System for predicting earliest completion time and using static priority having initial priority and static urgency for job scheduling
US7752622B1 (en) Method and apparatus for flexible job pre-emption
US7743378B1 (en) Method and apparatus for multi-dimensional priority determination for job scheduling
US20130030785A1 (en) Computer resource utilization modeling for multiple workloads
US20160132359A1 (en) Abnormality detection apparatus, control method, and program
CN110362388B (en) Resource scheduling method and device
US8214836B1 (en) Method and apparatus for job assignment and scheduling using advance reservation, backfilling, and preemption
CN111104211A (en) Task dependency based computation offload method, system, device and medium
US10628214B2 (en) Method for scheduling entity in multicore processor system
CN111046091B (en) Operation method, device and equipment of data exchange system
CN105955809B (en) Thread scheduling method and system
US20140109100A1 (en) Scheduling method and system
CN114579270A (en) Task scheduling method and system based on resource demand prediction
CN113010289A (en) Task scheduling method, device and system
KR20120106089A (en) Method for reducing power consumption of system software using query scheduling of application and apparatus for reducing power consumption using said method
US9983911B2 (en) Analysis controller, analysis control method and computer-readable medium
EP4300305A1 (en) Methods and systems for energy-efficient scheduling of periodic tasks on a group of processing devices
CN115328647A (en) Compute intensive service elastic expansion method based on task queue
Leulseged et al. Probabilistic analysis of multi-processor scheduling of tasks with uncertain parameters
CN116185584A (en) Multi-tenant database resource planning and scheduling method based on deep reinforcement learning
CN111930485B (en) Job scheduling method based on performance expression
US8607245B2 (en) Dynamic processor-set management
Murad et al. Priority Based Fair Scheduling: Enhancing Efficiency in Cloud Job Distribution

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