CN116055567A - Floating license scheduling system and method based on seismic data processing - Google Patents
Floating license scheduling system and method based on seismic data processing Download PDFInfo
- Publication number
- CN116055567A CN116055567A CN202111279582.8A CN202111279582A CN116055567A CN 116055567 A CN116055567 A CN 116055567A CN 202111279582 A CN202111279582 A CN 202111279582A CN 116055567 A CN116055567 A CN 116055567A
- Authority
- CN
- China
- Prior art keywords
- job
- ready queue
- scheduling
- license
- 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
Links
Images
Classifications
-
- 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
- Multi Processors (AREA)
Abstract
The invention provides a floating license scheduling system and method based on seismic data processing, and belongs to the field of seismic data processing. The system comprises: client side: for submitting, monitoring, modifying and deleting jobs; the job server: providing batch processing service; the operation server communicates with the client; the node executor: is a daemon running on each computing node; each node executor is communicated with a job server and a job dispatcher respectively; a job scheduler: scheduling system resources according to the use condition of node resources and the queuing condition of the jobs; the job server and the job dispatcher are respectively communicated with the license server. The invention reasonably schedules the operation of user operation by dynamically monitoring and collecting the states of all nodes in the system and the use conditions of various resources, thereby reducing the length of the waiting queue and achieving the purpose of improving the utilization rate and throughput rate of the cluster system resources.
Description
Technical Field
The invention belongs to the field of seismic data processing, and particularly relates to a floating license scheduling system and method based on seismic data processing.
Background
With the continuous decrease of HPC (high performance computer cluster) threshold, especially the popularization of cluster technology with good cost performance, more and more seismic data software solves the problems of more complex and larger calculation rule by using HPC technology. Seismic data processing software on a high-performance computing platform has become an indispensable technical research means for petroleum exploration, scientific research and production.
However, an unavoidable problem is the high price of commercial software. Even with very powerful computing power, large-scale computation and solution cannot be performed if there is insufficient commercial software license quantity. Large amounts of expensive business software are an important reason for restricting the expansion of computing sizes for enterprises. During seismic data processing, software is often not available.
Authorization means employed by seismic data processing software include persistent, subscription, function-based, floating, trial, and the like. Wherein, the floating license authorization, selling the access rights of the end user to the application program or service according to the maximum number of the concurrent users, will not lock the computing node hardware, and the software running on the HPC platform mostly adopts the authorization mode. Therefore, in order to manage and effectively distribute the software floating license related to the seismic data, the benefit of playing the software license resources is maximized; it is necessary to develop a system for floating software license scheduling to improve the use efficiency of software license resources.
Flexlm is software license management software developed by Macrovision corporation, and is also a license management system widely used in the industry, and its biggest feature is to support floating license management. The license management model may be divided into a server license management model and a no-server license management model according to whether a license server is required in license management. In the server license management model, a license management service is provided by a license management server, and a license file is provided by a software provider; in the serverless license management model, the license file is read directly by the Flexlm-enabled application. The server license management model is largely composed of Flexlm-enabled applications, license management daemons, software provider daemons, license files 4, as well as other optional components (e.g., debug log files, management selection files, etc.). The license management daemon and the software provider process together constitute a license server. The Flexlm license manager lmgrd handles the initial connection with the application and hands over the connection to the corresponding software provider process and is responsible for starting or restarting the software provider process. The software provider process is responsible for managing the number of licenses for Checkout and recording the application using those licenses. The license information is stored in a license file that contains information about the license server and the software provider process. The application program is connected to a program module of a Flexlm client library, which provides the communication function with the license server. Commercial software for processing seismic data such as omega, paradaigm and the like adopts a Flexlm license server to carry out license authorization.
Chinese patent publication CN113312592a discloses a method, apparatus, device and storage medium for scheduling software license, and Flexlm disclosed is software license management software developed by Macrovision company, and is also a license management system widely used in industry, and its biggest feature is to support floating license management. The license management model may be divided into a server license management model and a no-server license management model according to whether a license server is required in license management. In a cluster environment, a user uses seismic data processing software with a software license. Multiple user jobs within the same time period may apply for the same software, which may create a situation where multiple instances of the same software are running in the cluster system at the same time. Running the one software instance must obtain a license, which if lacking may result in a long wait of the user job in the backup queue, thereby increasing the average wait time and turnaround time of the job.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provide a floating license scheduling system and method based on seismic data processing, which achieve the purposes of global load balancing and improving the utilization rate of the licensed resources by dynamically monitoring the state and the utilization rate of the resource objects of a cluster system and performing operation scheduling according to the use condition of various computing resources and the load of the system.
The invention is realized by the following technical scheme:
in a first aspect of the invention, there is provided a floating license scheduling system based on seismic data processing, the system comprising:
client side: for submitting, monitoring, modifying and deleting jobs;
the job server: providing batch processing service; the operation server communicates with the client;
a job scheduler: scheduling system resources according to the use condition of node resources and the queuing condition of the jobs;
the node executor: is a daemon running on each computing node; each node executor is communicated with a job server and a job dispatcher respectively;
the job server and the job dispatcher are respectively communicated with the license server.
The invention further improves that:
the job scheduler adopts a first come first serve scheduling algorithm to schedule, and the method is as follows:
after submitting the job from the client to the job server, the job enters a waiting queue;
each time of scheduling is to select one or more jobs which enter the waiting queue first from the waiting queue, allocate a computing node and a license for the jobs, transmit input data to the computing node, and then place the jobs into a ready queue; the job is run until the license is relinquished after execution or blocking occurs.
The invention further improves that:
the job scheduler adopts a multi-stage feedback queue scheduling algorithm for scheduling, and the method is specifically as follows:
after submitting the job from the client to the job server, the job enters a waiting queue, and the job available with the license is prepared in the waiting queue to obtain a job flow;
setting a plurality of ready queues, giving different priorities to each ready queue, and gradually reducing the priority of the ready queues from the first ready queue to the last ready queue;
when a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; when the workflow execution is rolled, if it can be completed within the time slice, the workflow execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it is not completed after running a time slice in the second ready queue, then put it into the end of the third ready queue, and so on, after a long job flow is sequentially reduced from the first ready queue to the nth ready queue, execute in a time slice round-robin manner in the nth ready queue until the end;
only when the 1 st to (i-1) th ready queues are all empty, scheduling the operation flow in the i th ready queue; if the computing node is serving the workflow in the ith ready queue and a new workflow enters the ready queue with priority higher than that of the ith ready queue, the running workflow is put back to the end of the ith ready queue and a license is allocated to the newly arrived workflow in the high priority ready queue.
Preferably, the higher the priority ready queue, the shorter the time slice.
In a second aspect of the present invention, there is provided a floating license scheduling method based on seismic data processing, the method comprising:
(1) After submitting the job from the client to the job server, the job enters a waiting queue, the job server requests scheduling to the job scheduler, the job scheduler inquires whether the software running the job has a license available, if so, the job server prepares the job with the license available, and a job flow is obtained;
(2) The job scheduler performs job scheduling;
(3) And after the dispatching is finished, the dispatching information is written into the log file.
The invention further improves that:
the preparation operation includes:
splitting the operation flow;
distributing computing nodes;
the input data is transmitted to the compute node.
The invention further improves that:
the operation of step (2) comprises:
setting a plurality of ready queues, giving different priorities to each ready queue, and gradually reducing the priority of the ready queues from the first ready queue to the last ready queue; the higher the priority ready queue, the shorter the time slice;
and placing the operation flow into a ready queue and scheduling.
The operation of placing the workflow in a ready queue and scheduling includes:
when a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle;
when the workflow execution is rolled, if it can be completed within the time slice, the workflow execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it has not been completed after running a time slice in the second ready queue, it is placed at the end of the third ready queue, and so on, when a long job flow is sequentially dropped from the first ready queue to the nth ready queue, execution is performed in a time slice-rotated manner in the nth ready queue until the end.
The invention further improves that: in the scheduling process, only when the 1 st to (i-1) th ready queues are empty, scheduling the operation flow in the i th ready queue;
if the computing node is serving the workflow in the ith ready queue and a new workflow enters the ready queue with priority higher than that of the ith ready queue, the running workflow is put back to the end of the ith ready queue and a license is allocated to the newly arrived workflow in the high priority ready queue.
In a third aspect of the present invention, there is provided a computer-readable storage medium storing at least one program executable by a computer, the at least one program when executed by the computer causing the computer to perform the steps in the above-described floating license scheduling method based on seismic data processing.
Compared with the prior art, the invention has the beneficial effects that:
the invention reasonably schedules the operation of user operation by dynamically monitoring and collecting the states of all nodes in the system and the use conditions of various resources, thereby reducing the length of the waiting queue and achieving the purpose of improving the utilization rate and throughput rate of the cluster system resources.
Drawings
FIG. 1 is a schematic diagram of the composition of the system of the present invention.
Detailed Description
The invention is described in further detail below with reference to the attached drawing figures:
in a cluster environment, a user uses seismic data processing software with a software license. Multiple user jobs within the same time period may apply for the same software, which may create a situation where multiple instances of the same software are running in the cluster system at the same time. Running the one software instance must obtain a license, which if lacking may result in a long wait of the user job in the backup queue, thereby increasing the average wait time and turnaround time of the job.
The invention regards the software license as a system resource, manages with other system resources in a unified way, monitors the state and the utilization rate of the system resources of the cluster dynamically, and carries out job scheduling according to the use condition of various computing resources and the load of the system, thereby achieving the purposes of balancing global load and improving the utilization rate of the licensed resources.
The invention provides a floating license scheduling system based on seismic data processing, which comprises the following embodiments:
[ embodiment one ]
As shown in fig. 1, the system includes:
the system comprises a client, a job server, a node executor and a job scheduler, and specifically comprises the following steps:
client side: for submitting, monitoring, modifying and deleting jobs;
the job server: providing batch processing service, wherein the client and other daemons communicate with the operation server through a network;
the node executor: the node executor is a daemon running on each computing node, each node executor is respectively communicated with the job server and the job scheduler, loads the job and executes the job, feeds back the use condition of the node resource to the job scheduler, and returns the operation result to the job server;
a job scheduler: and scheduling the system resources according to the use condition of the node resources and the queuing condition of the jobs, and selecting proper jobs from the backup queues for scheduling operation.
The 4 parts cooperate to realize the functions of job management, job scheduling, load balancing and the like. The job server and the job dispatcher are respectively communicated with the license server.
The license server in fig. 1 is existing and is provided by a software provider, including a license management daemon and a software provider process, etc., and the main functions are license encryption verification, piracy protection, etc.
The seismic geographic software process in fig. 1 refers to a software main module, and permissions required by some interaction modules can be directly obtained from a license server through a license manager, which is the prior art and is not described herein.
The signals transmitted in fig. 1 are specifically as follows:
the client provides management of scheduling and a user's submission job channel.
1. Two arrows between the client and the job server respectively indicate that the job is submitted and the job completion value is returned, namely, the client sends the submitted job to the job server, and the job server returns the job completion value to the client;
2. two arrows between the job server and the job scheduler indicate, respectively, request scheduling and scheduling cancellation (job is withdrawn from ready queue), i.e. the job server sends request scheduling to job scheduling, and the job scheduler sends scheduling cancellation to the job server.
3. Two arrows between the license server and the scheduler indicate, respectively, acquisition of a license and return of a license, the job scheduler sends the acquisition of a license to the license server, and the license server returns the license to the job scheduler.
4. Two arrows between the job server and the node executor represent job distribution and job reclamation, respectively, i.e., the job server sends job distribution to the node executor, and the node executor sends job reclamation to the job server.
5. Two arrows between the job scheduler and the node executor respectively indicate various license occupation and various license reclamation, namely, the job scheduler sends various license occupation to the node executor, and the node executor sends various license reclamation to the job scheduler.
The scheduling algorithm adopted by the job scheduler for scheduling the system resources according to the use condition of node resources and the queuing condition of jobs can be the following two scheduling algorithms, and the job scheduling algorithm is specified according to the requirements, which scheduling algorithm is adopted, and the method specifically comprises the following steps:
first kind: first come first served scheduling algorithm: after submitting the job from the client to the job server, the job enters a waiting queue, and when the scheduling algorithm is adopted in job scheduling, each scheduling is to select one or more jobs which enter the queue first from the waiting queue, allocate computing nodes and permissions for the job, transmit input data to each computing node, and then place the job into the ready queue. The job does not relinquish permissions until completed or after some event has occurred to block.
The first-come first-serve scheduling algorithm has a certain limitation, has overlong waiting time for short jobs arranged at the back, occupies allowable time in the process of job preparation, and causes allowable waste.
Second kind: multistage feedback queue scheduling algorithm:
(1) After submitting the job from the client to the job server, the job enters a waiting queue, and the job available with the license is prepared in the waiting queue, namely, the following work is performed:
(1) splitting the operation flow;
(2) distributing computing nodes;
(3) transmitting the input data to a computing node;
all the preparation work is realized by adopting the prior art and is not repeated here
(2) Setting a plurality of ready queues, and giving different priorities to each ready queue:
the priority of the first ready queue is highest, the priority of the second ready queue is next, the priority of the rest ready queues is reduced one by one, and so on, and the priority of the last ready queue is lowest. The time (i.e., slice) for which the job flow acquisition permission is given to each ready queue is also different in size, and the time for which the acquisition permission is given for each job flow is smaller in the ready queue having higher priority. Preferably, the time that the latter ready queue takes permission is one time longer than the previous ready queue takes permission, e.g., the time slice of the second ready queue is one time longer than the time slice of the first ready queue, … …, and so on, the time slice of the i+1th ready queue is one time longer than the time slice of the i ready queue.
(3) When a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; when the workflow execution is rolled, if it can be completed within the time slice, the workflow execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it has not completed after running a time slice in the second ready queue, it is placed at the end of the third ready queue in turn, … …, and so on, when a long job flow is dropped from the first ready queue to the nth ready queue in turn, execution is performed in time slice rotation in the nth ready queue until it ends.
(4) Scheduling the workflow in the second ready queue only when the first ready queue is empty; only when the 1 st to (i-1) th ready queues are empty, the workflow in the i-th ready queue is scheduled. If the computing node is servicing the workflow in the ith ready queue and there is a new workflow entering a ready queue having a higher priority than the ith ready queue (i.e., any one of ready queues 1 through (i-1)), then the running workflow is placed back to the end of the ith ready queue and permissions are assigned to the newly arrived workflow in the high priority ready queue.
The multi-stage feedback queue scheduling algorithm optimizes the job preparation stage, so that the permission is not occupied in the job preparation time period, and the waste of the permission is avoided; and the permissible average turnaround time is also improved, the job waiting time is shortened, and the utilization rate of permissible resources and computing resources is improved. And the execution time required by various jobs is not required to be known in advance, so that the requirements of various types of jobs can be met.
The invention also provides a floating license scheduling method based on seismic data processing, and the embodiment of the method comprises the following steps:
[ example two ]
The method comprises the following steps:
(1) After submitting the job from the client to the job server, the job enters a waiting queue (the waiting queue is a queue waiting after submitting the job by the user, at this time, the job has not yet performed any preparation work, that is, has not performed links such as splitting flow, preparation data, etc.), the job server requests scheduling to the job scheduler, the job scheduler inquires whether the software running the job has a license available, if so, the job server performs preparation work on the job having the license available,
the preparation work is specifically as follows:
(1) splitting workflow
(2) Distributing computing nodes
(3) The input data is transmitted to the compute node.
And obtaining the operation flow of the operation after the preparation work is finished.
(2) The job scheduling device performs job scheduling, and the specific operations include:
a, a plurality of ready queues are set (the ready queues are the queues where the job is located after the job has completed the preparation work, and each job can immediately start calculation only by a license), and different priorities are given to each ready queue:
the priority of the first ready queue is highest, the priority of the second ready queue is next, the priority of the rest ready queues is reduced one by one, and so on, and the priority of the last ready queue is lowest. The time (i.e., slice) for which the job flow acquisition permission is given to each ready queue is also different in size, and the time for which the acquisition permission is given for each job flow is smaller in the ready queue having higher priority. Preferably, the time that the latter ready queue takes permission is one time longer than the previous ready queue takes permission, e.g., the time slice of the second ready queue is one time longer than the time slice of the first ready queue, … …, and so on, the time slice of the i+1th ready queue is one time longer than the time slice of the i ready queue. The higher the priority and the shorter the time slice, so that the long operation and the short operation can be simultaneously considered, and the short operation can be completed quickly.
B, putting the operation flow into a ready queue and scheduling, wherein the specific operation comprises the following steps:
when a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle;
when the workflow is rolled to be executed, if the workflow can be completed in the time slice, the workflow is executed until the execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it has not completed after running a time slice in the second ready queue, it is placed at the end of the third ready queue in turn, … …, and so on, when a long job flow is dropped from the first ready queue to the nth ready queue in turn, it is run in a time slice-rotated manner in the nth ready queue. And after the job execution is finished, the dispatching is finished.
C, in the scheduling process, scheduling the operation flow in the second ready queue only when the first ready queue is empty; only when the 1 st to (i-1) th ready queues are empty, the workflow in the i-th ready queue is scheduled. And, if the computing node is servicing the workflow in the ith ready queue and there is a new workflow entering a ready queue having a priority higher than the priority of the ith ready queue (i.e., any one of ready queues 1 through (i-1)), then the running workflow is placed back to the end of the ith ready queue and permissions are assigned to the new workflow in the high priority ready queue.
(3) After the dispatching is finished, the dispatching information (the dispatching information comprises the time of the job occupying license and the time of the job occupying computing resource, so that the charging is convenient) is written into the related log file.
In practical use, a certain proportion may be set, for example, the compute node can execute 20 jobs at most, and schedule 20 jobs from the ready queue at a time, and schedule 20 jobs from the ready queue for execution after the execution of 20 jobs is completed.
Through steps (1) to (3) above, the job scheduler periodically schedules jobs in the wait queue.
The job dispatcher of the present invention invokes the permissions using existing methods, as outlined below: before job scheduling, inquiring whether the software running the job is available with a license currently according to the characteristic key parameters (including the job priority and the modules required by the job) provided by the user job through an API provided by a license server, and not scheduling the software (under the condition that the license is not available or insufficient, if the job is prepared, resources are occupied, so that the job is prepared only after the license meeting the requirement is determined to be available, and the resource waste is avoided); user jobs can only wait in the wait queue. If the job scheduler inquires that the corresponding application software has a license available, checking a license from a license manager through an API (application program interface), and registering a user job using the license; after the normal operation of the job is finished, the license is returned to the license server. The job dispatcher checks whether the job in the ready queue is waiting for the license, and if so, the job is dispatched to run; if the job server finds that the job is abnormally ended, deleting the job and giving the license management server a software license held by the job server. The method comprises the following specific steps:
(1) Calling an lc_init () API to initialize a license server when the system starts to run, and creating a software license job;
(2) The job scheduler queries whether the license is available by calling lc_status () API, and the license server returns the current state of the feature key;
(3) The job scheduler calls lc_checkout () to apply for a license specifying the feature key, and the license server returns an execution result to the job scheduler;
(4) The job scheduler calls lc_checkin () to apply a license specifying a feature key to the license pool;
(5) The job scheduler may call lc_disconn () to disconnect from the license management daemon. For a user job with an error abortion, the job server deletes the job and registers a license held by the job with the license server.
The present invention also provides a computer-readable storage medium, an embodiment of which is as follows:
[ example III ]
The computer-readable storage medium stores at least one program executable by a computer, which when executed by the computer, causes the computer to perform the steps in the floating license scheduling method based on seismic data processing described above.
Seismic data processing software plays an important role in petroleum exploration and becomes an indispensable technical research means for scientific research and production. In the petroleum exploration field, a great deal of specialized software is frequently used, and in the seismic data processing process, the condition that the processing software is insufficient is frequently encountered. With the increase of scientific research production workload and personnel quantity, the demand quantity of scientific research software is continuously increased, and purchasing funds are increased year by year. The invention manages and effectively distributes the software floating license related to the seismic data, and maximizes the benefit of playing the software license resources
Finally, it should be noted that the above-mentioned technical solution is only one embodiment of the present invention, and various modifications and variations can be easily made by those skilled in the art based on the application methods and principles disclosed in the present invention, and are not limited to the methods described in the above-mentioned specific embodiments of the present invention, therefore, the foregoing description is only preferred, and not meant to be limiting.
Claims (10)
1. A floating license scheduling system based on seismic data processing, characterized by: the system comprises:
client side: for submitting, monitoring, modifying and deleting jobs;
the job server: providing batch processing service; the operation server communicates with the client;
a job scheduler: scheduling system resources according to the use condition of node resources and the queuing condition of the jobs;
the node executor: is a daemon running on each computing node; each node executor is communicated with a job server and a job dispatcher respectively;
the job server and the job dispatcher are respectively communicated with the license server.
2. The seismic data processing-based floating license scheduling system of claim 1, wherein: the job scheduler adopts a first come first serve scheduling algorithm to schedule, and the method is as follows:
after submitting the job from the client to the job server, the job enters a waiting queue;
each time of scheduling is to select one or more jobs which enter the waiting queue first from the waiting queue, allocate a computing node and a license for the jobs, transmit input data to the computing node, and then place the jobs into a ready queue; the job is run until the license is relinquished after execution or blocking occurs.
3. The seismic data processing-based floating license scheduling system of claim 1, wherein: the job scheduler adopts a multi-stage feedback queue scheduling algorithm for scheduling, and the method is specifically as follows:
after submitting the job from the client to the job server, the job enters a waiting queue, and the job available with the license is prepared in the waiting queue to obtain a job flow;
setting a plurality of ready queues, giving different priorities to each ready queue, and gradually reducing the priority of the ready queues from the first ready queue to the last ready queue;
when a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; when the workflow execution is rolled, if it can be completed within the time slice, the workflow execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it is not completed after running a time slice in the second ready queue, then put it into the end of the third ready queue, and so on, after a long job flow is sequentially reduced from the first ready queue to the nth ready queue, execute in a time slice round-robin manner in the nth ready queue until the end;
only when the 1 st to (i-1) th ready queues are all empty, scheduling the operation flow in the i th ready queue; if the computing node is serving the workflow in the ith ready queue and a new workflow enters the ready queue with priority higher than that of the ith ready queue, the running workflow is put back to the end of the ith ready queue and a license is allocated to the newly arrived workflow in the high priority ready queue.
4. A floating license scheduling system based on seismic data processing as recited in claim 3, wherein: the higher the priority ready queue, the shorter the time slice.
5. A floating license scheduling method based on seismic data processing is characterized in that: the method comprises the following steps:
(1) After submitting the job from the client to the job server, the job enters a waiting queue, the job server requests scheduling to the job scheduler, the job scheduler inquires whether the software running the job has a license available, if so, the job server prepares the job with the license available, and a job flow is obtained;
(2) The job scheduler performs job scheduling;
(3) And after the dispatching is finished, the dispatching information is written into the log file.
6. The seismic data processing-based floating license scheduling method of claim 5, wherein: the preparation operation includes:
splitting the operation flow;
distributing computing nodes;
the input data is transmitted to the compute node.
7. The seismic data processing-based floating license scheduling method of claim 5, wherein: the operation of step (2) comprises:
setting a plurality of ready queues, giving different priorities to each ready queue, and gradually reducing the priority of the ready queues from the first ready queue to the last ready queue; the higher the priority ready queue, the shorter the time slice;
and placing the operation flow into a ready queue and scheduling.
8. The seismic data processing-based floating license scheduling method of claim 7, wherein: the operation of placing the workflow in a ready queue and scheduling includes:
when a new workflow enters a ready queue, firstly placing the new workflow at the end of a first ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle;
when the workflow execution is rolled, if it can be completed within the time slice, the workflow execution is completed; if the operation process is finished and not finished in one time slice, transferring the operation process to the tail of a second ready queue, and queuing and waiting for scheduling according to a first-come-first-get principle; if it has not been completed after running a time slice in the second ready queue, it is placed at the end of the third ready queue, and so on, when a long job flow is sequentially dropped from the first ready queue to the nth ready queue, execution is performed in a time slice-rotated manner in the nth ready queue until the end.
9. The seismic data processing-based floating license scheduling method of claim 8, wherein: in the scheduling process, only when the 1 st to (i-1) th ready queues are empty, scheduling the operation flow in the i th ready queue;
if the computing node is serving the workflow in the ith ready queue and a new workflow enters the ready queue with priority higher than that of the ith ready queue, the running workflow is put back to the end of the ith ready queue and a license is allocated to the newly arrived workflow in the high priority ready queue.
10. A computer-readable storage medium, characterized by: the computer readable storage medium stores at least one program executable by a computer, which when executed by the computer, causes the computer to perform the steps in the floating license scheduling method based on seismic data processing as claimed in any one of claims 5 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279582.8A CN116055567A (en) | 2021-10-28 | 2021-10-28 | Floating license scheduling system and method based on seismic data processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279582.8A CN116055567A (en) | 2021-10-28 | 2021-10-28 | Floating license scheduling system and method based on seismic data processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116055567A true CN116055567A (en) | 2023-05-02 |
Family
ID=86130021
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111279582.8A Pending CN116055567A (en) | 2021-10-28 | 2021-10-28 | Floating license scheduling system and method based on seismic data processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116055567A (en) |
-
2021
- 2021-10-28 CN CN202111279582.8A patent/CN116055567A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108345501B (en) | Distributed resource scheduling method and system | |
| CN107729139B (en) | Method and device for concurrently acquiring resources | |
| JP4387174B2 (en) | Distributing processes associated with multiple priority groups across multiple resources | |
| CN101501645B (en) | Computer micro-jobs | |
| US7810099B2 (en) | Optimizing workflow execution against a heterogeneous grid computing topology | |
| CN103426072B (en) | The order processing system of a kind of high concurrent competition stock and disposal route thereof | |
| US6591262B1 (en) | Collaborative workload management incorporating work unit attributes in resource allocation | |
| US20060206894A1 (en) | Method of scheduling jobs using database management system for real-time processing | |
| US20080256228A1 (en) | Minimizing complex decisions to allocate additional resources to a job submitted to a grid environment | |
| US5925098A (en) | Apparatus and method for dispatching client method calls within a server computer system | |
| CN102193833A (en) | Efficient utilization of idle resources in a resource manager | |
| JPH0820963B2 (en) | Method for operating computer system and computer system | |
| CN111078436A (en) | Data processing method, device, equipment and storage medium | |
| CN107577534A (en) | A kind of resource regulating method and device | |
| US20020004814A1 (en) | Job distributed processing method and distributed processing system | |
| JPH05216842A (en) | Resources managing device | |
| CN111930485B (en) | Job scheduling method based on performance expression | |
| US9086927B2 (en) | Method and system for processing data for database modification | |
| CN105550025A (en) | Distributed IaaS (Infrastructure as a Service) scheduling method and system | |
| JP2007286703A (en) | Master and slave license management method, master and slave license management method and master and slave license management program | |
| CN116055567A (en) | Floating license scheduling system and method based on seismic data processing | |
| CN116719623A (en) | Job scheduling method, job result processing method and device | |
| CN109446641B (en) | Multi-stage reliability modeling analysis method of cloud computing service system | |
| CN104796494A (en) | Data transmission method for cloud platform | |
| Son et al. | Deadline allocation in a time-constrained workflow |
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 |