WO2019192307A1 - 比例公平调度的实现方法、装置及设备、存储介质 - Google Patents

比例公平调度的实现方法、装置及设备、存储介质 Download PDF

Info

Publication number
WO2019192307A1
WO2019192307A1 PCT/CN2019/078197 CN2019078197W WO2019192307A1 WO 2019192307 A1 WO2019192307 A1 WO 2019192307A1 CN 2019078197 W CN2019078197 W CN 2019078197W WO 2019192307 A1 WO2019192307 A1 WO 2019192307A1
Authority
WO
WIPO (PCT)
Prior art keywords
user terminals
scheduling
user terminal
fairness factor
user
Prior art date
Application number
PCT/CN2019/078197
Other languages
English (en)
French (fr)
Inventor
吴家迪
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2019192307A1 publication Critical patent/WO2019192307A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/04—Wireless resource allocation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/12—Wireless traffic scheduling
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/12—Wireless traffic scheduling
    • H04W72/121—Wireless traffic scheduling for groups of terminals or users
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/50—Allocation or scheduling criteria for wireless resources
    • H04W72/56—Allocation or scheduling criteria for wireless resources based on priority criteria
    • H04W72/566—Allocation or scheduling criteria for wireless resources based on priority criteria of the information or information source or recipient
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/04—Wireless resource allocation
    • H04W72/044—Wireless resource allocation based on the type of the allocated resource
    • H04W72/0446—Resources in time domain, e.g. slots or frames

Definitions

  • the present application relates to the field of communications technologies, but is not limited to the field of communications technologies, and in particular, to a method, an apparatus, and a device for implementing proportional fair scheduling, and a computer readable storage medium.
  • 5G 5th-Generation, 5th generation mobile communication technology
  • NR New Radio
  • the QoS (Quality of Service) priority queue sorting module based on the PF (Proportional Fair) scheduling algorithm will face an extremely large amount of computational demand; in addition, the traffic has time.
  • the location selectivity, in the wilderness and midnight traffic will drop from the peak to the bottom, the PF scheduling algorithm can not be adaptively adjusted.
  • the embodiments of the present application provide a method, an apparatus, and a device for implementing a proportional fair scheduling, and a computer readable storage medium.
  • a method for implementing proportional fair scheduling includes the following steps:
  • the number M of the user terminals is greater than the computing power N, determining a periodic coefficient P; dividing the M user terminals into P shares according to the periodic coefficient P, and the number of terminals of each copy is closed at [0, N]
  • the PF fairness factor of the M user terminals is calculated in the interval, and the PF fairness factor is used for scheduling of the user terminal.
  • a device for implementing a proportional fair scheduling includes a determining module and a sharding module;
  • the determining module is configured to determine a computing capability N of a fairness factor of the user terminal and a number M of user terminals of the QoS priority queue;
  • the sharding module is configured to determine a periodic coefficient P if the number M of the user terminals is greater than the computing capability N, and divide the M user terminals into P shares according to the periodic coefficient P, and the number of terminals in each copy
  • the PF fairness factor of the M user terminals is calculated in a closed interval of [0, N] and used in the P TTIs, wherein the PF fairness factor is used for scheduling of the user terminal.
  • a device for implementing a proportional fair scheduling comprising: a memory, a processor, and a proportional fair scheduling stored on the memory and operable on the processor And a step of implementing the proportional fair scheduling implementation method when the implementation program of the proportional fair scheduling is executed by the processor.
  • a computer readable storage medium on which the implementation program of proportional fair scheduling is stored, and when the implementation program of the proportional fair scheduling is executed by a processor The steps of implementing the above-described proportional fair scheduling implementation method.
  • the calculation amount of each TTI does not exceed the allocated computing power, and at the same time, the original performance of the PF algorithm can be effectively guaranteed.
  • FIG. 1 is a schematic flowchart of a method for implementing proportional fair scheduling according to an embodiment of the present application
  • FIG. 2 is a schematic structural diagram of an apparatus for implementing proportional fair scheduling according to an embodiment of the present application
  • FIG. 3 is another schematic structural diagram of an apparatus for implementing proportional fair scheduling according to an embodiment of the present disclosure
  • FIG. 4 is a schematic structural diagram of an apparatus for implementing proportional fair scheduling according to an embodiment of the present disclosure.
  • an embodiment of the present application provides a method for implementing proportional fair scheduling, where the method includes the following steps:
  • a computing capability N how many user terminals can be supported in a TTI (Transmission Time Interval) for calculating the PF fairness factor.
  • the computing power N is a dependent variable of the current RRC (Radio Resource Control) connection number, and according to different base station scheduling algorithms and hardware implementation levels adopted by different vendors, the computing power N is also under the same RRC connection number. Can be different, the minimum can not be lower than 1.
  • the minimum period coefficient can be calculated.
  • the determining the number N of user terminals of the computing capability N and the quality of service QoS priority queue further includes the following steps:
  • the computing power N and the number M of user terminals of the QoS priority queue are updated.
  • the purpose of the timer is to re-update the computing power N of the current base station and the number M of user terminals included in the current QoS priority queue at a fixed time or under certain trigger conditions.
  • the PF fairness factor of the M user terminals is calculated in the closed interval of the P transmission time interval TTI, wherein the PF fairness factor is used for scheduling of the user terminal.
  • the determined period coefficient P is not less than the minimum period coefficient Pmin, and may be a positive integer multiple of the minimum period coefficient Pmin.
  • the M user terminals are divided into P shares, and the user terminals of each share are not repeated and the number does not exceed N, that is, the number of terminals of each copy is within the closed interval of [0, N].
  • the user terminals of the P shares calculate the PF fairness factor according to the P TTIs, so that each TTI calculates only the PF fairness factor of the N user terminals at most.
  • the M user terminals are divided into P shares, and the number of the user terminals may be the same or different. It is conceivable that some TTIs do not need to calculate the PF fairness factor as the periodic coefficient P increases.
  • the method further includes the following steps:
  • the number M of the user terminals is not greater than the computing capability N, calculating a PF fairness factor of the M user terminals in one TTI, and in the QoS priority queue of the M user terminals Position adjustment.
  • the PF fairness factors of the M user terminals can be directly calculated in one TTI. This embodiment is suitable for low traffic situations such as the wilderness and midnight.
  • the calculating the PF fairness factor of the M user terminals in the P transmission time intervals TTI includes:
  • the PF fairness factor of the user terminal is calculated and updated within a trigger time corresponding to the user terminal, and the user terminal of the share is in the QoS priority queue according to the size of the PF fairness factor The position is adjusted.
  • the t-th TTI that calculates the PF fairness factor of the user terminal is referred to as a triggering time, where 1 ⁇ t ⁇ P, and is an integer; as opposed to the triggering time, the user terminal is not calculated.
  • the remaining (P-1) TTI times of the PF fairness factor are called silence times.
  • the number M of user terminals is 100
  • the computing power N is 20
  • the periodic coefficient P is 5 (where P is the minimum periodic coefficient)
  • 100 user terminals are divided into 5 parts (assuming an average allocation manner), each part includes 20 user terminals, and 5 TTIs are TTI0, TTI1, TTI2, TTI3, and TTI4.
  • TTI0 is the triggering time
  • TTI1-TTI4 is the silent time.
  • the periodic coefficient P is the minimum periodic coefficient.
  • Each TTI has a calculated PF fairness factor.
  • the number M of user terminals is 100
  • the computing power N is 20
  • the periodic coefficient P is 10 (where P is the minimum periodic coefficient) 2 times).
  • 100 user terminals can be divided into 10 shares, and the number of 10 user terminals is (15, 15, 15, 15, 10, 15, 5, 5, 5, 0), and 10 TTIs are TTI0-TTI9. Therefore, when the value of the periodic coefficient P is 10, some TTIs may be allowed to calculate the PF fairness factor.
  • the PF fairness factor of the user terminal needs to be calculated and updated within a triggering time corresponding to one user terminal. Specifically, it can be calculated by the following formula
  • ⁇ and ⁇ are both 1
  • TBsize is the TB (Transport Block) size that the user terminal can support the current channel condition
  • HistoryThroughput is the historical traffic of the PF fairness factor, which refers to the traffic scheduled by the base station. HistoryThroughput is calculated by Equation 2 below:
  • the updated PF fairness factor may be compared with the PF fairness factor of the user terminal in other QoS queues, and the user terminal is adjusted in the QoS priority queue according to the descending ordering manner. The location in .
  • this embodiment only needs to be in the M user terminals.
  • the user terminals are adjusted, and the amount of calculation corresponding to the original is much smaller, and the performance does not deteriorate.
  • the step of adjusting the location of the user terminal in the QoS priority queue according to the size of the PF fairness factor further includes the following steps:
  • the first K user terminals are selected for scheduling (the K values may be different according to each device vendor's own scheduling algorithm, but may not be greater than M), and the scheduling states and historical traffic of the K user terminals are updated. .
  • the method may further include the following steps:
  • the scheduling times of the M user terminals and the maximum consecutive scheduling times are updated.
  • the calculating the PF fairness factor of the M user terminals in the P transmission time intervals TTI includes:
  • the scheduling status of the user terminal is counted in a silent time; if there is a scheduled user terminal, the historical traffic of the scheduled user terminal is updated, wherein the historical traffic is used to calculate the scheduled user terminal.
  • the fairness factor is used to calculate the scheduled user terminal.
  • the scheduling states of the M user terminals are counted in the silent time, and the scheduling state can be maintained by the vector S, and each user terminal corresponds to a vector S.
  • the definition of vector S is as follows:
  • a vector S of length P is generated.
  • An element of the vector S of 1 indicates that the user terminal is scheduled, and a value of 0 indicates that the user terminal is not scheduled.
  • the resulting generated scheduling vector S can be as follows:
  • the historical traffic HistoryThroughput of the user terminal needs to be updated when the user terminal is scheduled, and is not updated when there is no scheduling.
  • the historical traffic HistoryThroughput of the user terminal can be updated in the following manner:
  • (1- ⁇ ) Ns+1 can be calculated by looking up the table in advance, and the calculation amount is relatively small. To this end, it can be approximated that the operational complexity of Equation 3 and Equation 2 is consistent.
  • the method further includes the steps of:
  • the maximum period coefficient Pmax is determined based on the period coefficient P.
  • the implementation manner can be applied to different implementation methods of different vendors, and the maximum periodic coefficient Pmax that can be supported can be estimated in advance, and the available periodic coefficient P and the optimal periodic coefficient P (in all available periodic coefficients P) The period corresponding to the case where the amount of calculation is the smallest is selected, which facilitates the overall design planning of each device manufacturer and designs the periodic configuration in different user scenarios.
  • the number M of user terminals, and the determined periodic coefficient P, according to the Hoeffding inequality 38 to 50 iterations can be performed to ensure the reliability of the statistical result (the sample error is less than 2 ⁇ ).
  • Each iteration operation runs tens of thousands of simulation calculations of the PF fairness factor.
  • the expected values of the scheduling number error ((maximum scheduling times - minimum scheduling times) ⁇ average scheduling times)
  • M user terminals are calculated respectively.
  • the expected value of the maximum number of consecutive schedules, and the average of the total calculated amount is calculated.
  • the calculation of the overall calculation amount is based on one multiplication method, one division is equal to four multiplications for conversion, and the addition and subtraction are not counted in the form of statistics.
  • the expected value of the scheduling number error is within 2% to 8%
  • the expected value of the maximum continuous scheduling number does not exceed the continuous scheduling number of the PF algorithm itself 100 to 200
  • the currently selected periodic coefficient P is considered Available to meet the performance requirements of the PF algorithm.
  • the period coefficient P is increased by the minimum period coefficient Pmin as the step size, and the above process is repeated until the above two conditions are not satisfied, thereby obtaining the maximum period coefficient Pmax. Further, all available period coefficients P and an optimum period coefficient P can be determined, wherein the optimum period coefficient P is a period corresponding to the case where the calculation amount is the smallest among all available period coefficients P.
  • the performance gradually decreases, so the performance is better when the maximum number of UEs scheduled per TTI is much smaller than M.
  • the maximum number of UEs that can be scheduled by each TTI is much smaller than the number of UEs M in the QoS priority queue.
  • the maximum supported period is 100 TTI, which is balanced by the scheduling error, the calculation amount, and the maximum continuous scheduling number.
  • the optional P value also includes all values between 20 and 80. The balance of the calculation amount exceeds 80%, and the error of the scheduling number is within 2%, which not only effectively reduces the computational cost, but also reliably guarantees the original performance of the PF algorithm.
  • the method for implementing the proportional fair scheduling in the embodiment of the present application uses a distributed computing manner to periodically update the PF fairness factor of each user terminal; so that the calculation amount of each TTI does not exceed The computing power of the allocation can effectively guarantee the original performance of the PF algorithm.
  • the embodiment of the present application provides a device for implementing fair proportional scheduling, the device includes a determining module 21 and a segmentation module 22;
  • the determining module 21 is configured to determine a computing capability N of the fairness factor of the user terminal and a number M of user terminals of the QoS priority queue.
  • a computing capability N how many user terminals can be supported in a TTI (Transmission Time Interval) for calculating the PF fairness factor.
  • the computing power N is a dependent variable of the current RRC (Radio Resource Control) connection number, and according to different base station scheduling algorithms and hardware implementation levels adopted by different vendors, the computing power N is also under the same RRC connection number. Can be different, the minimum can not be lower than 1.
  • the minimum period coefficient can be calculated.
  • the device further includes a detection module 24;
  • the detecting module 24 is configured to detect whether the timer overflows; if the timer overflows, update the computing capability N and the number M of user terminals of the QoS priority queue.
  • the purpose of the timer is to re-update the computing power N of the current base station and the number M of user terminals included in the current QoS priority queue at a fixed time or under certain trigger conditions.
  • the segmentation module 22 is configured to determine a periodic coefficient P if the number M of the user terminals is greater than the computing capability N, and divide the M user terminals into P shares according to the periodic coefficient P, and each terminal The number is in a closed interval of [0, N], and the PF fairness factor of the M user terminals is calculated in P TTIs, wherein the PF fairness factor is used for scheduling of the user terminal.
  • the determined period coefficient P is not less than the minimum period coefficient Pmin, and may be a positive integer multiple of the minimum period coefficient Pmin.
  • the M user terminals are divided into P shares, and the user terminals of each share are not repeated and the number does not exceed N, that is, the number of terminals of each copy is within the closed interval of [0, N].
  • the user terminals of the P shares calculate the PF fairness factor according to the rotation of the P TTIs, so that each TTI calculates only the PF fairness factor of the N user terminals at most.
  • the M user terminals are divided into P shares, and the number of the user terminals may be the same or different. It is conceivable that some TTIs do not need to calculate the PF fairness factor as the periodic coefficient P increases.
  • the device further includes a computing module 28;
  • the calculating module 28 is configured to calculate a PF fairness factor of the M user terminals in one TTI, if the number M of the user terminals is not greater than the computing capability N, and the M user terminals are in the TTI The position in the QoS priority queue is adjusted.
  • the PF fairness factors of the M user terminals can be directly calculated in one TTI. This embodiment is suitable for low traffic situations such as the wilderness and midnight.
  • the device further includes a trigger time processing module 23;
  • the trigger time processing module 23 is configured to calculate and update the PF fairness factor of the user terminal in the trigger time corresponding to the user terminal for any one of the user terminals, and the share of the PF fairness factor according to the size of the PF fairness factor The location of the user terminal in the QoS priority queue is adjusted.
  • the PF fairness factor of the user terminal needs to be calculated and updated within a triggering time corresponding to one user terminal. Specifically, it can be calculated by the following formula
  • ⁇ and ⁇ are both 1
  • TBsize is the TB (Transport Block) size that the user terminal can support the current channel condition
  • HistoryThroughput is the historical traffic of the PF fairness factor, which refers to the traffic scheduled by the base station. HistoryThroughput is calculated by Equation 2 below:
  • the updated PF fairness factor may be compared with the PF fairness factor of the user terminal in other QoS queues, and the user terminal is adjusted in the QoS priority queue according to the descending ordering manner. The location in .
  • this embodiment only needs to be in the M user terminals.
  • the user terminals are adjusted, and the amount of calculation corresponding to the original is much smaller, and the performance does not deteriorate.
  • the apparatus further includes a scheduling module 26 and an update module 27;
  • the scheduling module 26 is configured to select the top K user terminals for scheduling according to the order of priority (the K values may be different according to each device vendor's own scheduling algorithm, but may not be greater than M), and update the K The scheduling status and historical traffic of the user terminal;
  • the update module 27 is configured to update the scheduling times of the M user terminals and the maximum consecutive scheduling times according to the scheduling states of the M user terminals.
  • the device further includes a silent time processing module 25;
  • the silent time processing module 25 is configured to count the scheduling status of the user terminal in a silent time; if there is a scheduled user terminal, update the historical traffic of the scheduled user terminal, where the historical traffic A fairness factor for calculating the scheduled user terminal.
  • the scheduling states of the M user terminals are counted in the silent time, and the scheduling state can be maintained by the vector S, and each user terminal corresponds to a vector S.
  • the definition of vector S is as follows:
  • a vector S of length P is generated.
  • An element of the vector S of 1 indicates that the user terminal is scheduled, and a value of 0 indicates that the user terminal is not scheduled.
  • the resulting generated scheduling vector S can be as follows:
  • the historical traffic HistoryThroughput of the user terminal needs to be updated when the user terminal is scheduled, and is not updated when there is no scheduling.
  • the historical traffic HistoryThroughput of the user terminal can be updated in the following manner:
  • (1- ⁇ ) Ns+1 can be calculated by looking up the table in advance, and the calculation amount is relatively small. To this end, it can be approximated that the operational complexity of Equation 3 and Equation 2 is consistent.
  • the device further includes an evaluation module 29;
  • the evaluation module 29 is configured to determine a maximum period coefficient Pmax according to the period coefficient P.
  • the implementation manner can be applied to different implementation methods of different vendors, and the maximum periodic coefficient Pmax that can be supported can be estimated in advance, and the available periodic coefficient P and the optimal periodic coefficient P (in all available periodic coefficients P) The period corresponding to the case where the amount of calculation is the smallest is selected, which facilitates the overall design planning of each device manufacturer and designs the periodic configuration in different user scenarios.
  • the number of user terminals M and the periodic coefficient P, according to the Hoeffding inequality, 38 to 50 iterations can be performed to ensure the reliability of the statistical results (sample error is less than 2 ⁇ ), each time
  • the iterative operation runs tens of thousands of simulation calculations of the PF fairness factor.
  • the expected values of the scheduling number error ((maximum scheduling times - minimum scheduling times) ⁇ average scheduling times)
  • M user terminals are calculated respectively.
  • the expected value of the maximum number of consecutive schedules, and the average of the total calculated amount is calculated.
  • the calculation of the overall calculation amount is based on one multiplication method, one division is equal to four multiplications for conversion, and the addition and subtraction are not counted in the form of statistics.
  • the expected value of the scheduling number error is within 2% to 8%
  • the expected value of the maximum continuous scheduling number does not exceed the continuous scheduling number of the PF algorithm itself 100 to 200
  • the currently selected periodic coefficient P is considered Available to meet the performance requirements of the PF algorithm.
  • the period coefficient P is increased by the minimum period coefficient Pmin as the step size, and the above process is repeated until the above two conditions are not satisfied, thereby obtaining the maximum period coefficient Pmax. Further, all available period coefficients P and an optimum period coefficient P can be determined, wherein the optimum period coefficient P is a period corresponding to the case where the calculation amount is the smallest among all available period coefficients P.
  • the performance gradually decreases, so the performance is better when the maximum number of UEs scheduled per TTI is much smaller than M.
  • the maximum number of UEs that can be scheduled by each TTI is much smaller than the number of UEs M in the QoS priority queue.
  • the maximum supported period is 100 TTI, which is balanced by the scheduling error, the calculation amount, and the maximum continuous scheduling number.
  • the optional P value also includes all values between 20 and 80. The balance of the calculation amount exceeds 80%, and the error of the scheduling number is within 2%, which not only effectively reduces the computational cost, but also reliably guarantees the original performance of the PF algorithm.
  • the apparatus for implementing the proportional fair scheduling according to the embodiment of the present application periodically updates the PF fairness factor of each user terminal according to the number of user terminals and the computing capability, so that the calculation amount of each TTI does not exceed The computing power of the allocation can effectively guarantee the original performance of the PF algorithm.
  • an embodiment of the present application provides a device for implementing a proportional fair scheduling, where the device includes: a memory 31, a processor 32, and is stored on the memory 31 and executable on the processor 32.
  • the implementation program of the proportional fair scheduling when the implementation program of the proportional fair scheduling is executed by the processor 32, is used to implement the steps of the implementation method of the proportional fair scheduling described below:
  • the number M of the user terminals is greater than the computing power N, determining a periodic coefficient P; dividing the M user terminals into P shares according to the periodic coefficient P, and the number of terminals of each copy is closed at [0, N]
  • the PF fairness factor of the M user terminals is calculated in the interval, and the PF fairness factor is used for scheduling of the user terminal.
  • the PF fairness factor of the user terminal is calculated and updated within a trigger time corresponding to the user terminal, and the user terminal of the share is in the QoS priority queue according to the size of the PF fairness factor Position adjustment;
  • the scheduling status of the user terminal is counted in a silent time; if there is a scheduled user terminal, the historical traffic of the scheduled user terminal is updated, wherein the historical traffic is used to calculate the scheduled user terminal.
  • the fairness factor is used to calculate the scheduled user terminal.
  • the first K user terminals are selected for scheduling, and the scheduling states and historical traffic of the K user terminals are updated.
  • the number M of the user terminals is not greater than the computing capability N, calculating a PF fairness factor of the M user terminals in one TTI, and in the QoS priority queue of the M user terminals Position adjustment.
  • the maximum period coefficient Pmax is determined based on the period coefficient P.
  • the computing power N and the number M of user terminals of the QoS priority queue are updated.
  • the device for implementing the proportional fair scheduling according to the embodiment of the present application periodically updates the PF fairness factor of each user terminal according to the number of user terminals and the computing capability, so that the calculation amount of each TTI does not exceed The computing power of the allocation can effectively guarantee the original performance of the PF algorithm.
  • the embodiment of the present application provides a computer readable storage medium, where the implementation program of the proportional fair scheduling is stored on the computer readable storage medium, and the implementation program of the proportional fair scheduling is implemented by the processor to implement any of the foregoing embodiments.
  • the computer readable storage medium of the embodiment of the present application periodically updates the PF fairness factor of each user terminal according to the number of user terminals and the computing capability, so that the calculation amount of each TTI does not exceed the allocation.
  • the computing power can effectively guarantee the original performance of the PF algorithm.
  • the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and can also be implemented by hardware, but in many cases, the former is A better implementation.
  • the technical solution of the present application which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM, disk,
  • the optical disc includes a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the methods described in various embodiments of the present application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请公开一种比例公平调度的实现方法、装置及设备、计算机可读存储介质，该方法包括步骤：确定对用户终端公平因子的计算能力N和QoS优先级队列的用户终端的数目M；若用户终端的数目M大于所述计算能力N，则确定周期系数P；根据周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个传输时间间隔内计算M个用户终端的PF公平因子，其中所述PF公平因子用于用户终端的调度。

Description

比例公平调度的实现方法、装置及设备、存储介质
相关申请的交叉引用
本申请基于申请号为201810295075.5、申请日为2018年04月04日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请涉及通信技术领域但不限于通信技术领域，尤其涉及一种比例公平调度的实现方法、装置及设备、计算机可读存储介质。
背景技术
5G(5th-Generation，第五代移动通信技术)NR(New Radio，新无线电)作为新一代无线通信技术，在其技术指标中要求在2020年商用时，5G NR的连接密度要达到10 6devices/km 2。
面对如此大的连接密度，基于PF(Proportional Fair，比例公平)调度算法的QoS(Quality of Service，服务质量)优先级队列排序模块将面临无比庞大的计算量需求；另外话务量还具有时间、地点的选择性，在旷野和半夜话务量会从峰值跌落至谷底，PF调度算法并不能自适应地进行调整。
发明内容
有鉴于此，本申请实施例提供一种比例公平调度的实现方法、装置及设备、计算机可读存储介质。
本申请实施例解决上述技术问题所采用的技术方案如下：
根据本申请实施例的一个方面，提供的一种比例公平调度的实现方法， 所述方法包括步骤：
确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M；
若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
根据本申请实施例的另一个方面，提供的一种比例公平调度的实现装置，所述装置包括确定模块以及切分模块；
所述确定模块，用于确定对用户终端公平因子的计算能力N和QoS优先级队列的用户终端的数目M；
所述切分模块，用于若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
根据本申请实施例的另一个方面，提供的一种比例公平调度的实现设备，所述设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的比例公平调度的实现程序，所述比例公平调度的实现程序被所述处理器执行时实现上述的比例公平调度的实现方法的步骤。
根据本申请实施例的另一个方面，提供的一种计算机可读存储介质，所述计算机可读存储介质上存储有比例公平调度的实现程序，所述比例公平调度的实现程序被处理器执行时实现上述的比例公平调度的实现方法的步骤。
本申请实施例的比例公平调度的实现方法、装置及设备、计算机可读存储介质，根据用户终端的数目和计算能力，采用分布式计算的方式，周期性地更新各个用户终端的PF因子；使得每个TTI的计算量都不超过分配的计算能力，同时又能有效地保证PF算法的原有性能。
附图说明
图1为本申请实施例提供的比例公平调度的实现方法流程示意图；
图2为本申请实施例提供的比例公平调度的实现装置结构示意图；
图3为本申请实施例提供的比例公平调度的实现装置另一结构示意图；
图4为本申请实施例提供的比例公平调度的实现设备结构示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
为了使本申请所要解决的技术问题、技术方案及有益效果更加清楚、明白，以下结合附图和实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
如图1所示，本申请实施例提供一种比例公平调度的实现方法，所述方法包括步骤：
S11、确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M。
在本实施例中，在一个TTI(Transmission Time Interval，传输时间间隔)内能够支持多少个用户终端进行PF公平因子的计算，称作计算能力N。计算能力N是关于当前RRC(Radio Resource Control，无线资源控制)连接数的因变量，并且根据不同厂商采用的不同的基站调度算法和硬件实现水平，在相同的RRC连接数下，计算能力N也是可以不同的，最小不能低于1。
在本实施例中，当计算能力N和用户终端的数目M确定之后，即可计算得到最小周期系数
Figure PCTCN2019078197-appb-000001
在一种实施方式中，所述确定计算能力N和服务质量QoS优先级队列的用户终端的数目M之前还包括步骤：
检测定时器是否溢出；
若所述定时器溢出，则更新所述计算能力N和所述QoS优先级队列的用户终端的数目M。
在该实施方式中，定时器的目的是为了在固定的时间或者某些触发条件下重新更新当前基站的计算能力N、当前的QoS优先级队列包含的用户终端数目M。
S12、若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
在本实施例中，确定的周期系数P不小于最小周期系数Pmin，可以为最小周期系数Pmin的正整数倍。
在本实施例中，将M个用户终端分成P份，每份的用户终端之间不重复且数目不超过N，即每一份的终端数目在[0，N]的闭区间内。P份的用户终端按照P个TTI计算PF公平因子，这样每个TTI最多只计算N个用户终端的PF公平因子。需要说明的是，将M个用户终端分成P份，每份的用户终端的数目可以相同也可以不相同。可以想象得到的，随着周期系数P的增大，某些TTI是不需要计算PF公平因子的。
在一种实施方式中，所述确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M之后还包括步骤：
若所述用户终端的数目M不大于所述计算能力N，则在一个TTI内计算所述M个用户终端的PF公平因子，并对所述M个用户终端在所述QoS优先级队列中的位置进行调整。
在该实施方式中，如果用户终端的数目M不大于计算能力N，可直接在一个TTI内计算M个用户终端的PF公平因子。该实施方式适用于在旷野和半夜等低话务情景下。
在一种实施方式中，所述在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子包括：
对于任一份用户终端，在该份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子，并根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整。
针对某一份用户终端，将计算该份用户终端PF公平因子的第t个TTI称为触发时刻，其中1≤t≤P，且为整数；与触发时刻相对的，将没有计算该份用户终端PF公平因子的其余(P-1)个TTI时间称为静默时间。
作为示例地，假设用户终端的数目M为100，计算能力N为20，周期系数P为5(此时P为最小周期系数
Figure PCTCN2019078197-appb-000002
则100个用户终端分成5份，(假设为平均分配的方式)每份包括20个用户终端，5个TTI为TTI0、TTI1、TTI2、TTI3、TTI4。假设第一份用户终端在TTI0内有计算PF公平因子，在TTI1-TTI4内没有计算PF公平因子，对于第一份用户终端来说，TTI0为触发时刻，TTI1-TTI4则为静默时间。从以上可以看出，周期系数P为最小周期系数
Figure PCTCN2019078197-appb-000003
每一个TTI都有计算PF公平因子。
作为另一示例地，假设用户终端的数目M为100，计算能力N为20，周期系数P为10(此时P为最小周期系数
Figure PCTCN2019078197-appb-000004
的2倍)。则100个用户终端可分成10份，10份用户终端的数量分别为(15、15、15、15、10、15、5、5、5、0)，10个TTI为TTI0-TTI9。因此，在周期系数P的取值为10时，可以允许某些TTI是不需要计算PF公平因子的。
在该实施方式中，需要在一份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子。具体地，可通过以下公式进行计算
Figure PCTCN2019078197-appb-000005
公式1中的α，β均为1，TBsize为用户终端当前信道状况所能够支持的TB(Transport Block，传输块)大小，HistoryThroughput为PF公平因子的历史流量，指的是基站调度出去的流量。HistoryThroughput通过以下的公式2进行计算：
Figure PCTCN2019078197-appb-000006
公式2中的TBSize_CH为用户终端实际调度的TB块大小；θ是平滑因子，默认为1/128；Coef是上下行配比系数，作为示例地上下行配比可以为7:3，即Coef＝0.7。
在该实施方式中，在得到更新后的PF公平因子，可将更新后的PF公平因子与其他QoS队列中的用户终端的PF公平因子进行比较，根据降序排列方式调整用户终端在QoS优先级队列中的位置。
可以想象得到的，随着M的增大，相对于对M个用户终端整体进行排序，本实施例只需要对M个用户终端中的
Figure PCTCN2019078197-appb-000007
个用户终端进行调整，相对应原先的计算量会小很多，并且性能不会恶化。
可选地，所述根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整之后还包括步骤：
按照优先级的高低顺序，选择前K个用户终端进行调度(根据各个设备商自己的调度算法，K值可以不同，但是不能大于M)，并更新所述K个用户终端的调度状态和历史流量。
可选地，更新所述K个用户终端的调度状态和历史流量之后还可包括步骤：
根据所述M个用户终端的调度状态，更新所述M个用户终端的调度次数以及最大连续调度次数。
在另一种实施方式中，所述在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子包括：
在静默时间内统计该份用户终端的调度状态；若存在被调度的用户终端，则更新所述被调度的用户终端的历史流量，其中，所述历史流量用于计算所述被调度的用户终端的公平因子。
在该实施方式中，在静默时间内统计M个用户终端的调度状态，调度状态可通过向量S来进行维护，每个用户终端对应一个向量S。向量S的定义如下：
P个TTI内(包括计算PF公平因子的触发时刻)，生成一个长度为P的向量S，向量S中的元素为1表示用户终端被调度，为0表示用户终端没 有被调度。最终生成的调度向量S可如下所示：
S＝[1110......0101]。
在该实施方式中，用户终端的历史流量HistoryThroughput需要在用户终端被调度的时候更新，而没有调度时则不更新。
作为示例地，可通过以下方式更新用户终端的历史流量HistoryThroughput：
记录前一次触发时刻s的历史流量，记为His(s)，s＝20*frameNum+TTI,如果是刚开机，那么His(s)＝His(0)＝0。
相对于s’时刻，经过连续N*s个TTI不调度后(N*s<P-1)，第N*s+1个TTI被调度，则根据公式(2)得到His(s’+N*s+1)为：
His(s'+Ns+1)＝(1-θ) Ns+1×His(s')+θ×TBSize_CH×Coef，公式3
其中(1-θ) Ns+1的运算可以通过提前定义好的折算表格，查表进行运算，计算量相对较小。为此可近似认为公式3与公式2的运算复杂度是一致的。
在一种实施方式中，所述若所述用户终端的数目M大于所述计算能力N，则确定周期系数P之后，还包括步骤：
根据所述周期系数P，确定最大周期系数Pmax。
具体地，根据所述计算能力N、所述用户终端的数目M以及确定的周期系数P，进行迭代运算；
计算调度次数误差的期望值以及最大连续调度次数的期望值；
若所述调度次数误差的期望值和所述最大连续调度次数的期望值满足预设值，则以最小周期系数Pmin为步长增大所述周期系数P，并重复前述步骤直到获得最大周期系数Pmax；其中最小周期系数
Figure PCTCN2019078197-appb-000008
该实施方式可以适用于不同厂商的差异性实施方法，可以进行性能摸底提前评估出可以支持的最大周期系数Pmax，进而评估出可用周期系数P和最优周期系数P(在所有可用周期系数P中选择计算量最小的情况对应的周期)，便于各个设备厂商整体设计规划，设计不同的用户场景下的周期配 置。
为了更好地说明该实施方式的迭代运算，以下进行示例说明：
首先根据计算能力N、用户终端的数目M以及确定的周期系数P，依据霍夫丁(Hoeffding)不等式，进行38～50次迭代运算即可保证统计结果的可靠性(样本误差小于2‰)，每次迭代运算运行上万次PF公平因子的仿真计算。
然后根据统计得到的M个用户终端各自的总调度次数和最大连续调度次数的结果，分别计算调度次数误差的期望值((最大调度次数－最小调度次数)÷平均调度次数)、M个用户终端的最大连续调度次数的期望值、以及整体计算量的平均值。其中对整体计算量统计是以1个乘法为基本单位，一个除法等于4个乘法进行折算，加减法不计的形式进行统计。
如果满足以下两个条件：1)调度次数误差的期望值在2％～8％以内、2)最大连续调度次数的期望值不超过PF算法本身连续调度次数100～200，则认为当前选择的周期系数P可用，满足PF算法的性能要求。
最后以最小周期系数Pmin为步长增大周期系数P，重复上述过程直至上述两个条件都不满足，从而得到最大周期系数Pmax。进而可确定所有可用周期系数P、最优周期系数P，其中最优周期系数P是在所有可用周期系数P中选择计算量最小的情况对应的周期。
请查看以下表格所示，表中给出了N＝2，M＝10每个TTI调度的用户终端数目不同时在不同周期系数P下的性能参数。
Figure PCTCN2019078197-appb-000009
Figure PCTCN2019078197-appb-000010
从表中可以看出，不同的调度状态下(每个TTI最大调度UE数目)，随着周期系数P的上升，调度次数误差降低或者先降后升，计算量减小，但是最大连续调度次数增多。
随着每个TTI最大调度UE数目接近M，性能逐渐下降，所以当每个TTI调度的最大UE数目远小于M时性能越好。然而在5G NR的应用场景中，绝大部分情况每个TTI所能调度的最大UE数目都远小于QoS优先级队列中的UE数目M。
在计算能力N＝2和调度UE数目M＝10，每个TTI最大调度UE数目为2的条件下，最大支持的周期为100TTI，权衡于调度次数误差、计算量和最大连续调度次数这三个关键指标，认为P＝50最优。当然在所述设置的环境下，可选的P值还包含20～80之间的所有值。计算量的结余都超过80％，调度次数误差在2％以内，既有效降低了计算量开销，又可靠的保障了PF算法的原有性能。
本申请实施例的比例公平调度的实现方法，根据用户终端的数目和计算能力，采用分布式计算的方式，周期性地更新各个用户终端的PF公平因子；使得每个TTI的计算量都不超过分配的计算能力，同时又能有效地保 证PF算法的原有性能。
如图2所示，本申请实施例提供一种比例公平调度的实现装置，所述装置包括确定模块21以及切分模块22；
所述确定模块21，配置为确定对用户终端公平因子的计算能力N和QoS优先级队列的用户终端的数目M。
在本实施例中，在一个TTI(Transmission Time Interval，传输时间间隔)内能够支持多少个用户终端进行PF公平因子的计算，称作计算能力N。计算能力N是关于当前RRC(Radio Resource Control，无线资源控制)连接数的因变量，并且根据不同厂商采用的不同的基站调度算法和硬件实现水平，在相同的RRC连接数下，计算能力N也是可以不同的，最小不能低于1。
在本实施例中，当计算能力N和用户终端的数目M确定之后，即可计算得到最小周期系数
Figure PCTCN2019078197-appb-000011
请参考图3所示，在一种实施方式中，所述装置还包括检测模块24；
所述检测模块24，配置为检测定时器是否溢出；若所述定时器溢出，则更新所述计算能力N和所述QoS优先级队列的用户终端的数目M。
在该实施方式中，定时器的目的是为了在固定的时间或者某些触发条件下重新更新当前基站的计算能力N、当前的QoS优先级队列包含的用户终端数目M。
所述切分模块22，配置为若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
在本实施例中，确定的周期系数P不小于最小周期系数Pmin，可以为最小周期系数Pmin的正整数倍。
在本实施例中，将M个用户终端分成P份，每份的用户终端之间不重 复且数目不超过N，即每一份的终端数目在[0，N]的闭区间内。P份的用户终端按照P个TTI的轮流计算PF公平因子，这样每个TTI最多只计算N个用户终端的PF公平因子。需要说明的是，将M个用户终端分成P份，每份的用户终端的数目可以相同也可以不相同。可以想象得到的，随着周期系数P的增大，某些TTI是不需要计算PF公平因子的。
请参考图3所示，在一种实施方式中，所述装置还包括计算模块28；
所述计算模块28，配置为若所述用户终端的数目M不大于所述计算能力N，则在一个TTI内计算所述M个用户终端的PF公平因子，并对所述M个用户终端在所述QoS优先级队列中的位置进行调整。
在该实施方式中，如果用户终端的数目M不大于计算能力N，可直接在一个TTI内计算M个用户终端的PF公平因子。该实施方式适用于在旷野和半夜等低话务情景下。
请参考图3所示，在一种实施方式中，所述装置还包括触发时刻处理模块23；
所述触发时刻处理模块23，配置为对于任一份用户终端，在该份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子，并根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整。
在该实施方式中，需要在一份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子。具体地，可通过以下公式进行计算
Figure PCTCN2019078197-appb-000012
公式1中的α，β均为1，TBsize为用户终端当前信道状况所能够支持的TB(Transport Block，传输块)大小，HistoryThroughput为PF公平因子的历史流量，指的是基站调度出去的流量。HistoryThroughput通过以下的公式2进行计算：
Figure PCTCN2019078197-appb-000013
公式2中的TBSize_CH为用户终端实际调度的TB块大小；θ是平滑因子，默认为1/128；Coef是上下行配比系数，作为示例地上下行配比可以为7:3，即Coef＝0.7。
在该实施方式中，在得到更新后的PF公平因子，可将更新后的PF公平因子与其他QoS队列中的用户终端的PF公平因子进行比较，根据降序排列方式调整用户终端在QoS优先级队列中的位置。
可以想象得到的，随着M的增大，相对于对M个用户终端整体进行排序，本实施例只需要对M个用户终端中的
Figure PCTCN2019078197-appb-000014
个用户终端进行调整，相对应原先的计算量会小很多，并且性能不会恶化。
请参考图3所示，可选地，所述装置还包括调度模块26和更新模块27；
所述调度模块26，用于按照优先级的高低顺序，选择前K个用户终端进行调度(根据各个设备商自己的调度算法，K值可以不同，但是不能大于M)，并更新所述K个用户终端的调度状态和历史流量；
所述更新模块27，用于根据所述M个用户终端的调度状态，更新所述M个用户终端的调度次数以及最大连续调度次数。
请参考图3所示，在一种实施方式中，所述装置还包括静默时间处理模块25；
所述静默时间处理模块25，用于在静默时间内统计该份用户终端的调度状态；若存在被调度的用户终端，则更新所述被调度的用户终端的历史流量，其中，所述历史流量用于计算所述被调度的用户终端的公平因子。
在该实施方式中，在静默时间内统计所述M个用户终端的调度状态，调度状态可通过向量S来进行维护，每个用户终端对应一个向量S。向量S的定义如下：
P个TTI内(包括计算PF公平因子的触发时刻)，生成一个长度为P的向量S，向量S中的元素为1表示用户终端被调度，为0表示用户终端没有被调度。最终生成的调度向量S可如下所示：
S＝[1110......0101]。
在该实施方式中，用户终端的历史流量HistoryThroughput需要在用户 终端被调度的时候更新，而没有调度时则不更新。
作为示例地，可通过以下方式更新用户终端的历史流量HistoryThroughput：
记录前一次触发时刻s的历史流量，记为His(s)，s＝20*frameNum+TTI,如果是刚开机，那么His(s)＝His(0)＝0。
相对于s’时刻，经过连续N*s个TTI不调度后(N*s<P-1)，第N*s+1个TTI被调度，则根据公式(2)得到His(s’+N*s+1)为：
His(s'+Ns+1)＝(1-θ) Ns+1×His(s')+θ×TBSize_CH×Coef，公式3
其中(1-θ) Ns+1的运算可以通过提前定义好的折算表格，查表进行运算，计算量相对较小。为此可近似认为公式3与公式2的运算复杂度是一致的。
请再参考图3所示，在一种实施方式中，所述装置还包括评估模块29；
所述评估模块29，配置为根据所述周期系数P，确定最大周期系数Pmax。
具体地，根据所述计算能力N、所述用户终端的数目M以及所述周期系数P，进行迭代运算；计算调度次数误差的期望值以及最大连续调度次数的期望值；若所述调度次数误差的期望值和所述最大连续调度次数的期望值满足预设值，则以最小周期系数Pmin为步长增大所述周期系数P，并重复前述步骤直到获得最大周期系数Pmax；其中最小周期系数
Figure PCTCN2019078197-appb-000015
该实施方式可以适用于不同厂商的差异性实施方法，可以进行性能摸底提前评估出可以支持的最大周期系数Pmax，进而评估出可用周期系数P和最优周期系数P(在所有可用周期系数P中选择计算量最小的情况对应的周期)，便于各个设备厂商整体设计规划，设计不同的用户场景下的周期配置。
为了更好地说明该实施方式的迭代运算，以下进行示例说明：
首先根据计算能力N、用户终端的数目M以及周期系数P，依据霍夫 丁(Hoeffding)不等式，进行38～50次迭代运算即可保证统计结果的可靠性(样本误差小于2‰)，每次迭代运算运行上万次PF公平因子的仿真计算。
然后根据统计得到的M个用户终端各自的总调度次数和最大连续调度次数的结果，分别计算调度次数误差的期望值((最大调度次数－最小调度次数)÷平均调度次数)、M个用户终端的最大连续调度次数的期望值、以及整体计算量的平均值。其中对整体计算量统计是以1个乘法为基本单位，一个除法等于4个乘法进行折算，加减法不计的形式进行统计。
如果满足以下两个条件：1)调度次数误差的期望值在2％～8％以内、2)最大连续调度次数的期望值不超过PF算法本身连续调度次数100～200，则认为当前选择的周期系数P可用，满足PF算法的性能要求。
最后以最小周期系数Pmin为步长增大周期系数P，重复上述过程直至上述两个条件都不满足，从而得到最大周期系数Pmax。进而可确定所有可用周期系数P、最优周期系数P，其中最优周期系数P是在所有可用周期系数P中选择计算量最小的情况对应的周期。
请查看以下表格所示，表中给出了N＝2，M＝10每个TTI调度的用户终端数目不同时在不同周期系数P下的性能参数。
Figure PCTCN2019078197-appb-000016
Figure PCTCN2019078197-appb-000017
从表中可以看出，不同的调度状态下(每个TTI最大调度UE数目)，随着周期系数P的上升，调度次数误差降低或者先降后升，计算量减小，但是最大连续调度次数增多。
随着每个TTI最大调度UE数目接近M，性能逐渐下降，所以当每个TTI调度的最大UE数目远小于M时性能越好。然而在5G NR的应用场景中，绝大部分情况每个TTI所能调度的最大UE数目都远小于QoS优先级队列中的UE数目M。
在计算能力N＝2和调度UE数目M＝10，每个TTI最大调度UE数目为2的条件下，最大支持的周期为100TTI，权衡于调度次数误差、计算量和最大连续调度次数这三个关键指标，认为P＝50最优。当然在所述设置的环境下，可选的P值还包含20～80之间的所有值。计算量的结余都超过80％，调度次数误差在2％以内，既有效降低了计算量开销，又可靠的保障了PF算法的原有性能。
本申请实施例的比例公平调度的实现装置，根据用户终端的数目和计算能力，采用分布式计算的方式，周期性地更新各个用户终端的PF公平因子；使得每个TTI的计算量都不超过分配的计算能力，同时又能有效地保证PF算法的原有性能。
如图4所示，本申请实施例提供一种比例公平调度的实现设备，所述 设备包括：存储器31、处理器32及存储在所述存储器31上并可在所述处理器32上运行的比例公平调度的实现程序，所述比例公平调度的实现程序被所述处理器32执行时，用于实现以下所述的比例公平调度的实现方法的步骤：
确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M；
若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
所述比例公平调度的实现程序被所述处理器32执行时，还用于实现以下所述的比例公平调度的实现方法的步骤：
对于任一份用户终端，在该份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子，并根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整；
在静默时间内统计该份用户终端的调度状态；若存在被调度的用户终端，则更新所述被调度的用户终端的历史流量，其中，所述历史流量用于计算所述被调度的用户终端的公平因子。
所述比例公平调度的实现程序被所述处理器32执行时，还用于实现以下所述的比例公平调度的实现方法的步骤：
按照优先级的高低顺序，选择前K个用户终端进行调度，并更新所述K个用户终端的调度状态和历史流量。
所述比例公平调度的实现程序被所述处理器32执行时，还用于实现以下所述的比例公平调度的实现方法的步骤：
若所述用户终端的数目M不大于所述计算能力N，则在一个TTI内计算所述M个用户终端的PF公平因子，并对所述M个用户终端在所述QoS优先级队列中的位置进行调整。
所述比例公平调度的实现程序被所述处理器32执行时，还用于实现以 下所述的比例公平调度的实现方法的步骤：
根据所述周期系数P，确定最大周期系数Pmax。
所述比例公平调度的实现程序被所述处理器32执行时，还用于实现以下所述的比例公平调度的实现方法的步骤：
检测定时器是否溢出；
若所述定时器溢出，则更新所述计算能力N和所述QoS优先级队列的用户终端的数目M。
本申请实施例的比例公平调度的实现设备，根据用户终端的数目和计算能力，采用分布式计算的方式，周期性地更新各个用户终端的PF公平因子；使得每个TTI的计算量都不超过分配的计算能力，同时又能有效地保证PF算法的原有性能。
本申请实施例提供一种计算机可读存储介质，所述计算机可读存储介质上存储有比例公平调度的实现程序，所述比例公平调度的实现程序被处理器执行时实现前述任意实施例所述的比例公平调度的实现方法的步骤。
本申请实施例的计算机可读存储介质，根据用户终端的数目和计算能力，采用分布式计算的方式，周期性地更新各个用户终端的PF公平因子；使得每个TTI的计算量都不超过分配的计算能力，同时又能有效地保证PF算法的原有性能。
需要说明的是，上述装置实施例与方法实施例属于同一构思，其具体实现过程详见方法实施例，且方法实施例中的技术特征在装置实施例中均对应适用，这里不再赘述。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件来实现，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如 ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
以上参照附图说明了本申请的优选实施例，并非因此局限本申请的权利范围。本领域技术人员不脱离本申请的范围和实质，可以有多种变型方案实现本申请，比如作为一个实施例的特征可用于另一实施例而得到又一实施例。凡在运用本申请的技术构思之内所作的任何修改、等同替换和改进，均应在本申请的权利范围之内。

Claims (10)

  1. 一种比例公平调度的实现方法，所述方法包括步骤：
    确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M；
    若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；
    根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
  2. 根据权利要求1所述的方法，其中，所述在P个传输时间间隔TTI内计算所述M个用户终端的PF公平因子包括：
    对于任一份用户终端，在该份用户终端对应的触发时刻内计算并更新该份用户终端的PF公平因子，并根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整；
    在静默时间内统计该份用户终端的调度状态；若存在被调度的用户终端，则更新所述被调度的用户终端的历史流量，其中，所述历史流量用于计算所述被调度的用户终端的公平因子。
  3. 根据权利要求2所述的方法，其中，所述根据PF公平因子的大小对该份的用户终端在所述QoS优先级队列中的位置进行调整之后，还包括步骤：
    按照优先级的高低顺序，选择前K个用户终端进行调度，并更新所述K个用户终端的调度状态和历史流量。
  4. 根据权利要求1所述的方法，其中，所述确定对用户终端公平因子的计算能力N和服务质量QoS优先级队列的用户终端的数目M之后还包括步骤：
    若所述用户终端的数目M不大于所述计算能力N，则在一个TTI内计算所述M个用户终端的PF公平因子，并对所述M个用户终端在所述QoS优先级队列中的位置进行调整。
  5. 根据权利要求1所述的方法，其中，所述若所述用户终端的数目M大于所述计算能力N，则确定周期系数P之后，还包括步骤：
    根据所述周期系数P，确定最大周期系数Pmax。
  6. 根据权利要求1所述的方法，其中，所述确定计算能力N和服务质量QoS优先级队列的用户终端的数目M之前还包括步骤：
    检测定时器是否溢出；
    若所述定时器溢出，则更新所述计算能力N和所述QoS优先级队列的用户终端的数目M。
  7. 一种比例公平调度的实现装置，所述装置包括确定模块以及切分模块；
    所述确定模块，配置为确定对用户终端公平因子的计算能力N和QoS优先级队列的用户终端的数目M；
    所述切分模块，配置为若所述用户终端的数目M大于所述计算能力N，则确定周期系数P；根据所述周期系数P将M个用户终端分成P份，每一份的终端数目在[0，N]的闭区间内，并在P个TTI内计算所述M个用户终端的PF公平因子，其中，所述PF公平因子用于用户终端的调度。
  8. 根据权利要求7所述的装置，其中，所述装置还包括计算模块；
    所述计算模块，配置为若所述用户终端的数目M不大于所述计算能力N，则在一个TTI内计算所述M个用户终端的PF公平因子，并对所述M个用户终端在所述QoS优先级队列中的位置进行调整。
  9. 一种比例公平调度的实现设备，其中，所述设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的比例公平调度的实现程序，所述比例公平调度的实现程序被所述处理器执行时实现如权利要求1至6中任一项所述的比例公平调度的实现方法的步骤。
  10. 一种计算机可读存储介质，其中，所述计算机可读存储介质上存储有比例公平调度的实现程序，所述比例公平调度的实现程序被处理器执行时实现如权利要求1至6中任一项所述的比例公平调度的实现方法的步骤。
PCT/CN2019/078197 2018-04-04 2019-03-14 比例公平调度的实现方法、装置及设备、存储介质 WO2019192307A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810295075.5 2018-04-04
CN201810295075.5A CN110351863B (zh) 2018-04-04 2018-04-04 比例公平调度的实现方法、装置及设备、存储介质

Publications (1)

Publication Number Publication Date
WO2019192307A1 true WO2019192307A1 (zh) 2019-10-10

Family

ID=68100113

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/078197 WO2019192307A1 (zh) 2018-04-04 2019-03-14 比例公平调度的实现方法、装置及设备、存储介质

Country Status (2)

Country Link
CN (1) CN110351863B (zh)
WO (1) WO2019192307A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060285490A1 (en) * 2005-06-20 2006-12-21 Kadaba Srinivas R Method and apparatus for quality-of-service based admission control using a virtual scheduler
CN101651518A (zh) * 2009-05-22 2010-02-17 中山大学 一种hsdpa***的快速分组调度方法
CN102083221A (zh) * 2009-11-27 2011-06-01 ***通信集团北京有限公司 一种基于hsdpa的资源调度方法及其装置
CN102448119A (zh) * 2010-10-14 2012-05-09 中兴通讯股份有限公司 分组调度方法及装置
CN105517179A (zh) * 2015-12-25 2016-04-20 上海华为技术有限公司 一种无线资源调度的方法以及调度器

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20100022989A (ko) * 2007-06-27 2010-03-03 닛본 덴끼 가부시끼가이샤 다지점 접속 장치, 신호 분석 및 제어 장치와, 그 방법 및 기록 매체
CN101600231A (zh) * 2009-06-30 2009-12-09 中兴通讯股份有限公司 在分组调度中确定用户设备的优先级的方法和***
CN101765123B (zh) * 2009-12-11 2012-05-23 哈尔滨工业大学 多用户mimo-ofdm上行链路资源分配方法
WO2012030521A1 (en) * 2010-09-03 2012-03-08 Spidercloud Wireless, Inc. Method and device for conditional access
CN105744634A (zh) * 2014-12-08 2016-07-06 中兴通讯股份有限公司 图样开关模式下的用户调度方法及基站
CN104955154A (zh) * 2015-06-08 2015-09-30 电子科技大学 基于用户业务量满意度的lte上行比例公平资源调度方法
US9854536B2 (en) * 2015-08-03 2017-12-26 Cisco Technology, Inc. User equipment power level selection for downlink transmissions

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060285490A1 (en) * 2005-06-20 2006-12-21 Kadaba Srinivas R Method and apparatus for quality-of-service based admission control using a virtual scheduler
CN101651518A (zh) * 2009-05-22 2010-02-17 中山大学 一种hsdpa***的快速分组调度方法
CN102083221A (zh) * 2009-11-27 2011-06-01 ***通信集团北京有限公司 一种基于hsdpa的资源调度方法及其装置
CN102448119A (zh) * 2010-10-14 2012-05-09 中兴通讯股份有限公司 分组调度方法及装置
CN105517179A (zh) * 2015-12-25 2016-04-20 上海华为技术有限公司 一种无线资源调度的方法以及调度器

Also Published As

Publication number Publication date
CN110351863B (zh) 2022-03-25
CN110351863A (zh) 2019-10-18

Similar Documents

Publication Publication Date Title
KR101539214B1 (ko) 보장된 비트 레이트 및/또는 최대 비트 레이트 서비스들에 대한 스케줄링 및 수락 제어 방법
EP2530988B1 (en) A scheduling concept
CN110266551A (zh) 一种带宽预测方法、装置、设备及存储介质
JP5413631B2 (ja) 規定のサービス品質を満たすための負荷推定
US9369925B2 (en) System and methods to achieve optimum efficiency-jain fairness in wireless systems
WO2013109888A1 (en) Systems and method for uplink resource allocation
US20200374721A1 (en) Streaming Media Quality of Experience Prediction for Network Slice Selection in 5G Networks
US11729657B2 (en) Apparatus and method for communication network
Kumar et al. A delay-optimal packet scheduler for M2M uplink
US20080175152A1 (en) Service differentiating and overload indication for downlink
CN110780986B (zh) 一种基于移动边缘计算的物联网任务调度方法及***
CN113840333B (zh) 电网资源分配的方法、装置、电子设备及存储介质
EP3338499A1 (en) Decreasing free-riding data traffic in uplink scheduling
US9281960B2 (en) Estimating processing workloads
WO2019192307A1 (zh) 比例公平调度的实现方法、装置及设备、存储介质
Mardani et al. Robust resource allocation scheme under channel uncertainties for LTE-A systems
Mandelli et al. Power optimization and throughput enhancement in 6G networks by delay-aware resource leverage
US10405209B2 (en) Blue-printing interference for LTE access in unlicensed spectrum
US10356799B2 (en) Scheduling method and apparatus of wireless communication system
US10582529B2 (en) Speculative scheduling in mobile networks
US9980278B2 (en) Weighted scheduling metric for user equipment in wireless communication system
WO2023157148A1 (ja) スケジューリング制御装置、スケジューリング制御システム、スケジューリング制御方法、及びプログラム
Ahmed et al. Analytical model of two level scheduling algorithm for WiMAX networks
US11582161B2 (en) Enhanced network with data flow differentiation
Shanmugam et al. Wireless downloading delay under proportional fair scheduling with coupled service and requests: An approximated analysis

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19782197

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 04/02/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19782197

Country of ref document: EP

Kind code of ref document: A1