WO2021102929A1 - 用于视频数据处理的资源分配方法及电子设备 - Google Patents

用于视频数据处理的资源分配方法及电子设备 Download PDF

Info

Publication number
WO2021102929A1
WO2021102929A1 PCT/CN2019/122039 CN2019122039W WO2021102929A1 WO 2021102929 A1 WO2021102929 A1 WO 2021102929A1 CN 2019122039 W CN2019122039 W CN 2019122039W WO 2021102929 A1 WO2021102929 A1 WO 2021102929A1
Authority
WO
WIPO (PCT)
Prior art keywords
video data
target video
user terminal
condition
target
Prior art date
Application number
PCT/CN2019/122039
Other languages
English (en)
French (fr)
Inventor
冯大权
王晨梦
张胜利
Original Assignee
深圳大学
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 深圳大学 filed Critical 深圳大学
Priority to PCT/CN2019/122039 priority Critical patent/WO2021102929A1/zh
Publication of WO2021102929A1 publication Critical patent/WO2021102929A1/zh

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W72/00Local resource management
    • H04W72/04Wireless resource allocation

Definitions

  • This application relates to the field of communication technology, and specifically relates to a resource allocation method and electronic equipment for video data processing.
  • video data transmission is considered an important service for the next generation of mobile communication networks.
  • the transmission of video data usually requires a large amount of radio spectrum resources.
  • efficient wireless transmission is the basis of high-quality video services.
  • Heterogeneous network is one of the important features and key technologies of the next-generation wireless communication network. It can improve the spectrum utilization through the overlapping coverage between the small cell and the macro cell and the dense deployment of the small cell, thereby achieving a substantial increase in network capacity. improve.
  • Video data has an important feature, that is, it can be reused.
  • the video content that is popular with users should be stored in the wireless access network for other users to reuse these video data in the future. This puts a certain degree on the storage capacity of the wireless communication network. Claim. Because a large amount of video content is consumed by a variety of heterogeneous network terminals, including PCs, smart phones, TVs, and tablet computers, these terminal devices require different video data rates, formats, and resolutions. In order to meet different network conditions and match different terminal devices, a video content may be encoded into more than 40 different versions. Due to limited storage space, storing all versions of video content is expensive and impractical.
  • Video transcoding technology Transcoding Technology
  • the transcoding operation of video data usually requires a lot of computing resources. Due to the limitation of size and battery life, mobile terminals cannot provide enough computing resources. This requires the wireless communication network system to be able to provide computing services for users.
  • the storage and replacement cycle of video clips is relatively long, usually from several minutes to several hours, while the frequency spectrum and computing resource allocation cycle involved in transcoding calculation and data transmission is usually in milliseconds. To the second level. Therefore, it is difficult to make video storage decision and resource allocation optimization at the same time.
  • the embodiments of the present application provide a resource allocation method and electronic device for video data processing to solve the resource allocation problem in the video data processing process.
  • embodiments of the present application provide a resource allocation method for video data processing, including:
  • each optimization period is divided into a plurality of consecutive time periods
  • the parameters of the target video data include length and network data flow;
  • the offloading decision It is used to represent the percentage of local transcoding operations performed on the target video data corresponding to each user terminal.
  • the resource allocation method for video data processing divides the processing of resource allocation into two parts to adapt to different update cycles between different optimization variables, that is, whether to store each target in each optimization cycle
  • the video data is determined, and the determined storage result is used as the input for determining the offloading decision, spectrum resource and computing resource allocation at each time period in the optimization cycle, so as to ensure the joint optimization of resource allocation and storage decision at the same time, which improves resource utilization rate.
  • the determining whether to store each of the target video data based on the parameters of each of the target video data includes:
  • the target network data flow Using the parameters of the target video data, the target network data flow, the local storage capacity, and the target network data flow to form a first constraint condition of whether to store each of the target video data;
  • the goal is to maximize the profit of storing all the target video data, and the first constraint condition is used to determine whether to store each of the target video data.
  • the resource allocation method for video data processing aims at maximizing the profit of storing all target video data, and optimizes the storage decision using the first constraint condition, which can ensure that the optimized storage decision meets the constraint Under the condition of the conditions, the storage revenue can be maximized.
  • the determining the target network data traffic caused by storing the target video data based on the length of each target video data includes:
  • the parameters of the target video data, the target network data flow, the local storage capacity, and the target network data flow are used to determine whether to store
  • the first constraint condition of each target video data includes:
  • the first constraint condition and the profit maximization of all the target video data are expressed by the following formula:
  • C11, C12, C13, and C14 respectively represent the first constraint sub-condition, the second constraint sub-condition, the third constraint sub-condition and the fourth constraint sub-condition;
  • ⁇ w represents the income of storing the target video data w;
  • L w represents the length of the target video data w;
  • Y represents the local storage capacity;
  • 1/ ⁇ w represents the retention time of the transmission gap of the target video segment w;
  • r w represents the transmission target video data w
  • ⁇ w represents the user request arrival rate of the target video data w; ⁇ >0;
  • O + [p w ] represents the maximum network capacity allocated to the target video data w;
  • the resource allocation method for video data processing aims at maximizing storage revenue, and uses probability-based network traffic constraints and storage space constraints to ensure that the optimization process can tolerate the uncertainty of optimization parameters and make full use of Storage space improves the robustness of the resource allocation method.
  • the determination of the corresponding offloading decision, spectrum resource and computing resource allocation of each user terminal in each said time period includes:
  • the hit result includes a direct hit, a transcoding hit, and a miss
  • the small cell includes the small base station and the corresponding user terminal ;
  • the total bandwidth of the local available frequency band, the bandwidth of the backhaul link between the small cell and the macro base station, and the spectrum efficiency of the small base station transmitting data to the user terminal it is formed to determine that each user terminal is in each location.
  • the goal is to maximize local revenue
  • the second constraint conditions are used to determine the corresponding offloading decisions, spectrum resources, and computing resource allocations of each user terminal in each of the time periods; wherein, the local revenue is all The sum of the revenue of the spectrum resource and the revenue of the computing resource.
  • the spectrum efficiency of the base station transmitting data to the user terminal forms a second constraint condition that determines the allocation of offloading decisions, spectrum resources, and computing resources corresponding to each user terminal in each time period, including:
  • the second constraint condition and the local profit maximization are expressed by the following formula:
  • k n represents the user terminal k corresponding to the small base station n
  • K n represents the set of all user terminals
  • v n represents the unit price charged by the small cell n
  • F represents all local computing resources
  • ⁇ n represents the unit price of the transmission target video data charged to the user terminal in the small cell n
  • B represents the total bandwidth of the local available frequency band
  • ⁇ n represents the unit price of the leased spectrum of the small cell n
  • ⁇ n represents the unit price of the backhaul link between the leased small cell n and the macro base station
  • p n represents the small base station n to the user terminal k
  • the resource allocation method for video data processing provided by the embodiment of the application optimizes the offloading decision, spectrum resource, and computing resource in the time period based on the storage strategy determined in the optimization period, and partially unloads the strategy, spectrum resource, and computing resource.
  • the resource allocation scheme is jointly modeled as an optimization problem, which can increase the utilization of system storage space and achieve higher system revenue.
  • an electronic device including:
  • a memory and a processor the memory and the processor are in communication connection with each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the first aspect of the present application, or the first The resource allocation method for video data processing described in any one of the embodiments of the aspect.
  • an embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, the computer instructions are used to make the computer execute the first aspect of the present application, or The resource allocation method for video data processing described in any one of the implementations of the first aspect.
  • Fig. 1 is a schematic structural diagram of a communication network system according to an embodiment of the present application
  • Fig. 2 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application
  • Fig. 3 is a schematic diagram of time axis division according to an embodiment of the present application.
  • Fig. 4 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application
  • Fig. 5 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application
  • Fig. 6 is a schematic diagram of data flow corresponding to different hit results according to an embodiment of the present application.
  • Fig. 7 is a structural block diagram of a resource allocation device for video data processing according to an embodiment of the present application.
  • FIG. 8 is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present application.
  • the resource allocation method for video data processing described in the embodiments of this application is based on mobile edge computing (Mobile Edge Computing, referred to as MEC) to ensure simultaneous joint resource allocation and storage decision-making Optimize to improve the utilization of backhaul link resources and storage resources of the converged system in video transmission.
  • MEC Mobile Edge Computing
  • Fig. 1 shows the communication network system described in the embodiment of the present application.
  • the embodiment of the present application adopts a two-layer heterogeneous network (macro cell and small cell) as a wireless transmission scenario.
  • the communication network system It includes an MEC server, a macro base station, and multiple small cells (only two small cells are shown in Figure 1).
  • the macro base station is connected to the MEC server, and multiple small cells are connected to the macro base station to obtain video data from the MEC server through the macro base station.
  • the small cell includes a small base station and multiple user equipment (UE for short). The user equipment in each small cell communicates with the macro base station through a corresponding small base station to obtain video data from the MEC server.
  • UE user equipment
  • the MEC server processes the corresponding video data and sends it to the user terminal after receiving the request for acquiring the video data from the user terminal.
  • video data processing takes video storage and transcoding as an example.
  • the transcoding means that the resolution of the video data stored in the MEC server is inconsistent with the resolution of the requested video data, and the resolution of the stored video data needs to be processed, that is, the Transcoding.
  • video storage decision-making, computing task offloading, and spectrum and computing resource allocation issues are jointly optimized; in order to adapt to different update cycles between different optimization variables, the optimization process of resource allocation is divided into two stages.
  • an embodiment of a resource allocation method for video data processing is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be implemented in a computer system such as a set of computer-executable instructions. Execution, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than here.
  • a resource allocation method for video data processing is provided, which can be used in electronic equipment, such as the MEC server in FIG. 1, and the following description takes the MEC server as an example for detailed description.
  • Fig. 2 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application. As shown in Fig. 2, the process includes the following steps:
  • each optimization period is divided into multiple consecutive time periods.
  • the optimization period and time period involved are explained as follows. As shown in Figure 3, the timeline is divided into end-to-end optimization periods, and the length of each optimization period is T; T is divided into several time periods, and the length of the time period is t.
  • T is a parameter that can be determined by user preferences.
  • T is relatively large, it means that the video segment will be stored by the system for a longer period of time, which can reduce the system overhead of storage replacement, but the update speed of the video data will be lower.
  • T is small, the cost of storage replacement is higher, but this brings a higher video data update speed.
  • T takes several minutes to several hours.
  • Each optimization period T consists of several time periods t, and the value of t is usually a few milliseconds to a few seconds.
  • the resource allocation method in this embodiment is divided into two stages for processing, that is, at the beginning of each optimization period T, the system will perform the first stage of optimization, that is, make video storage decisions. After that, the system temporarily stores the corresponding video data according to the decision result until the next optimization cycle begins. In this optimization cycle, storage decisions will be used as input to the second stage of optimization, namely resource allocation. The second stage of optimization is performed once at the beginning of each time period.
  • each target video data can be encoded into J levels of resolution.
  • the MEC server only stores the highest resolution version of each target video data, then In this case, each requested target video data version may be obtained in one of the following three ways: (1) Direct storage hit: The MEC server stores the requested video segment, and the UE requests the video. The highest resolution version of the clip. (2) Transcoding hit: The MEC server stores the requested video segment, but the UE requests a lower resolution version, so it needs to perform a transcoding operation to complete this video service. (3) Storage miss: The MEC server does not store the requested video clip, and the request is forwarded to the source server on the Internet. In this case, after the video content arrives, the MEC server can decide whether to store the highest resolution version of the content.
  • the parameters of the target video data include length and network data flow.
  • the MEC server After the MEC server obtains multiple target video data, it can obtain the length of each target video data and the corresponding network data flow by analyzing them.
  • the MEC server uses the parameters of each target video data, and combines some of its own parameters (for example, storage capacity, the consumption of target traffic data caused by storing the target video data), etc., or it can also determine whether it is combined with the current network performance Store the acquired target video data.
  • the method for determining whether to store each target video data is not limited in this application, which will be described in detail below.
  • the offloading decision is used to indicate the percentage of local transcoding operation on the target video data corresponding to each user terminal.
  • the MEC server After the MEC server determines whether to store each target video data in each optimization period, it can use the corresponding storage decision at each time period in the optimization period to allocate resources for the user terminal, specifically for the user terminal The corresponding unloading decision, spectrum resource and computing resource allocation in each time period.
  • the MEC server When the user's terminal MEC server sends a video data request and a transcoding hit event occurs, the MEC server needs to transcode the highest resolution version of the requested video to the lower resolution version required by the UE. Or, the highest version of this video clip can be sent to the destination UE through the small cell base station, and the UE performs the transcoding operation by itself.
  • a partial computing offloading mechanism is adopted.
  • part of the requested target video data is transcoded by the MEC, and the remaining part is sent to the UE, where the user terminal performs the transcoding operation.
  • use To represent the calculation and offloading decision corresponding to UE k n.
  • use Represents the proportion of the total bandwidth allocated by the small cell n to the UE k n to transmit the target video data, so Must be established.
  • the MEC server uses the storage result of each target video data to determine the corresponding offloading decision, spectrum resource and computing resource allocation of each user terminal in each time period, it can use the bandwidth resource required to transmit the target video data.
  • the frequency resources are compared with the bandwidth resources and spectrum resources that the system can provide to determine.
  • other methods may be used for determination, etc., and there is no restriction on the specific determination method here.
  • the resource allocation method for video data processing divides the processing of resource allocation into two parts to adapt to different update cycles between different optimization variables, that is, whether to store each target video in each optimization cycle
  • the data is determined, and the determined storage result is used as the input for determining the offloading decision, spectrum resource and computing resource allocation at each time period in the optimization cycle to ensure that resource allocation and storage decision-making are jointly optimized at the same time, and resource utilization is improved.
  • FIG. 4 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application. As shown in Figure 4, the process includes the following steps:
  • each optimization period is divided into multiple consecutive time periods.
  • the parameters of the target video data include length and network data flow.
  • an optimization method is used to determine whether to store each target video data, wherein the optimization constraint condition is determined based on the parameters of the target video data, and the target optimization function is to maximize the profit of storing all target video data.
  • the above S23 includes the following steps:
  • S231 Determine the target network data traffic caused by storing each target video data based on the length of each target video data.
  • the MEC server determines the target network data traffic caused by the target video data, it also combines the user request rate of each target video data to ensure the reliability of the determined target network data traffic.
  • the foregoing S231 may be implemented by adopting the following steps:
  • the MEC assumes that the user request arrival rate of the target video data w obeys the Poisson distribution and uses ⁇ w (Requests/second) to represent the arrival rate.
  • ⁇ w L w ⁇ w (bit/s) as the target network data traffic caused by the transmission of the stored target video data w. Therefore, the network data traffic caused by transmitting all the stored target video data is calculated as
  • S232 Utilize the parameters of the target video data, the target network data flow, the local storage capacity, and the target network data flow to form a first constraint condition of whether to store each target video data.
  • the MEC device After the MEC device obtains the storage target network data flow, it can combine the parameters of each target video data, the target network data flow, and the local storage capacity to form the first constraint condition whether to store each target video data. Specifically, the above S232 includes the following steps:
  • a decision vector for storing the target network data is formed to obtain the first constraint sub-condition.
  • the first constraint sub-condition is used to ensure that the storage decision variable h w is a binary variable, that is, For subsequent optimization processing, the binary variable ⁇ h w ⁇ [0,1] can be relaxed into a continuous variable Then the first constraint sub-condition can be expressed as:
  • the second constraint sub-condition is used to indicate that the sum of the data amount of all the stored target video data does not exceed the storage capacity limit Y of the MEC server. Specifically, the product of the decision vector and the length of the corresponding target network data is used to obtain the data volume of each target network data stored; and then the sum of the data volume of all target network data is calculated to form the second constraint sub-condition. That is, the second constraint sub-condition can be expressed as:
  • the third constraint sub-condition is used to indicate that the network data traffic caused by storing each target video data does not exceed the network data traffic allocated by the MEC server for it, and C (bit/s) is used to represent the network capacity. Further, in order to ensure the fairness of transmission opportunities between different video segments, C w is used to represent the network capacity allocated for transmitting video segments w. C and ⁇ C w ⁇ depend on the bandwidth allocation vector The system can be stable only when the following conditions are met, that is, the third constraint sub-condition is expressed as:
  • the fourth constraint sub-condition is used to indicate that the network data traffic caused by storing all target video data does not exceed the network capacity of the MEC server.
  • the radio spectrum allocation vector s needs to be separated from the third constraint sub-condition and the subsequent fourth constraint sub-condition, and a feedback-based method is used to provide an estimate of the network capacity (used for video data transmission). Assume that the maximum network capacity allocated to the target video data w at time t is The estimated maximum network capacity allocated to the video segment w is calculated as:
  • v ⁇ [0,1] is a constant used to adjust the ratio between the current network state and the previous network state; the superscript [t-1] represents the last evaluation time.
  • the estimated total network capacity is:
  • the long-term mean of the parameter ⁇ w is defined as And it is proposed that the relationship between the accurate request arrival rate and the long-term average is based on the bounded random parameter ⁇ and:
  • ⁇ >0 represents the maximum degree of uncertainty of the parameter ⁇ w
  • is a zero-mean random parameter within the interval [-1, 1]
  • the parameter ⁇ reflects the possible fluctuation of the request arrival rate.
  • This expression means that the actual request arrival rate ⁇ w can not exceed The magnitude of deviation from the estimated value of the request arrival rate Therefore, the possible degree of deviation is actually controlled by the parameter ⁇ . Because a larger ⁇ can bring better robustness, and a smaller ⁇ can bring more reliable resource reservation, the MEC server can adjust this parameter according to the robustness level requirements and historical statistics to achieve robustness. Great balance between performance and resource reservation.
  • the parameter ⁇ is defined as the confidence level, which means that the maximum probability of violating the third constraint sub-condition or the fourth constraint sub-condition is ⁇ .
  • the third constraint sub-condition is further expressed as:
  • the fourth constraint sub-condition can be expressed as:
  • the goal is to maximize the profit of storing all target video data
  • the first constraint condition is used to determine whether to store each target video data
  • the storage vector can be optimized with the goal of maximizing the profit of storing all target video data.
  • ⁇ w is used to represent the revenue of storing target video data w
  • the storage resource overhead of storing target video data w is Therefore, the net income of storing the video clip w is Then, the objective function for maximizing the profit of storing all target video data can be expressed as:
  • the storage vector h can be optimized.
  • the first constraint condition and the profit maximization of all target video data are expressed by the following formula:
  • C11, C12, C13, and C14 respectively represent the first constraint sub-condition, the second constraint sub-condition, the third constraint sub-condition and the fourth constraint sub-condition;
  • ⁇ w represents the income of storing the target video data w;
  • L w represents the length of the target video data w;
  • Y represents the local storage capacity;
  • 1/ ⁇ w represents the retention time of the transmission gap of the target video segment w;
  • r w represents the transmission target video data w
  • ⁇ w represents the user request arrival rate of the target video data w; ⁇ >0;
  • O + [p w ] represents the maximum network capacity allocated to the target video data w;
  • the offloading decision is used to indicate the percentage of local transcoding operation on the target video data corresponding to each user terminal.
  • the resource allocation method for video data processing aims at maximizing the profit of storing all target video data, and optimizes the storage decision using the first constraint condition, which can ensure that the optimized storage decision meets the constraint condition In the case of maximum storage revenue.
  • FIG. 5 is a flowchart of a resource allocation method for video data processing according to an embodiment of the present application. As shown in Figure 5, the process includes the following steps:
  • each optimization period is divided into multiple consecutive time periods.
  • the parameters of the target video data include length and network data flow.
  • S33 Determine whether to store each target video data based on the parameters of each target video data.
  • the offloading decision is used to indicate the percentage of local transcoding operation on the target video data corresponding to each user terminal.
  • the optimization method is also used to optimize the corresponding unloading decision, spectrum resource and computing resource allocation of each user terminal in each time period.
  • the optimization constraints and objective functions will be performed. Detailed description.
  • the above S34 includes the following steps:
  • S341 Based on the storage results of multiple target video data in each optimization period, determine the hit result of the target video data requested by the user.
  • the hit result includes a direct hit, a transcoding hit, and a miss.
  • the MEC server has determined the storage result of the acquired target video data in each optimization cycle. After receiving the user's request for acquiring the target video data, it can judge the hit result based on the user's request.
  • the hit result at this time is a direct hit; when the target video data is stored in the MEC server, but the user requests If it is the low-resolution form of the target video data, the hit result at this time is a transcoding hit; when the target video data is not stored in the MEC server, the hit result at this time is a miss.
  • Figure 6 shows the video data transmission process under different hit results.
  • the hit result is a direct storage hit
  • the user terminal directly obtains the target video data from the MEC server
  • the hit result is a transcoding hit
  • the MEC server transcodes the corresponding target video data and sends it to the corresponding user terminal
  • the hit result is a miss
  • the user terminal obtains the corresponding video data from the source server through the MEC server, the core network, and the Internet.
  • S342 Obtain the total bandwidth of the local available frequency band, the bandwidth of the backhaul link between the small cell and the macro base station, and the spectrum efficiency of the small base station to transmit data to the user terminal.
  • the small cell includes a small base station and a corresponding user terminal.
  • Some local parameters can be stored in the memory of the MEC server in advance, or the MEC server can obtain them from the outside when needed, and so on. Specifically, it is considered that all small cells overlap and use the same spectrum, so there is interference between small cells.
  • There is only downlink transmission that is, the wireless transmission is from the small base station to the UE belonging to it, and the interference is from other small base stations to this UE.
  • the total bandwidth available for the MEC server is B Hz.
  • the backhaul link bandwidth between the macro base station and the MEC server is L bps, and the backhaul link bandwidth between the small cell n and the macro base station is L n bps.
  • S343 Use the hit result of the target video data, the total bandwidth of the local available frequency band, the bandwidth of the backhaul link between the small cell and the macro base station, and the spectrum efficiency of the small base station to transmit data to the user terminal to determine that each user terminal is in each time period.
  • the fifth constraint sub-condition is used to ensure that the computing resources allocated to all user terminals in the entire system do not exceed the total computing resources of the MEC server.
  • the sixth constraint sub-condition is used to ensure that the computing resources allocated to each user terminal are not less than its own computing capacity, otherwise the offloading of computing tasks will be meaningless.
  • It represents the computing capability of UE k n itself
  • F represents all computing resources of the MEC server.
  • the amount of calculation tasks that need to be completed is Specifically, Represents the total number of CPU cycles required to complete the calculation task. The value of depends on the target video data requested by the UE k n and the resolution level it requests.
  • the sixth constraint sub-condition can be expressed as:
  • the inverse of is the MEC server performs computing tasks Time consuming, while The reciprocal of is the time for UE k n to perform this calculation task by itself. This means that the amount of computing resources allocated for each unit of computing task can reflect the time consumption of performing this computing task.
  • the seventh constraint sub-condition is used to indicate that in each small cell, the spectrum allocated to all users does not overlap, so as to meet the requirements of the system model. Then, the seventh constraint sub-condition is expressed as:
  • the eighth constraint sub-condition is used to ensure that the data rate transmitted to each user terminal is not less than the minimum data rate required for the video data it requests
  • the spectrum efficiency of the small base station n transmitting data to the UE k n is:
  • p n is the power density when small base station n transmits to UE k n, and with They are the channel gains from small base station n to UE k n and small base station m to UE k n , m ⁇ n; ⁇ represents the power spectral density of additive white Gaussian noise.
  • the instantaneous data rate at which the small base station n transmits data to UE k n is calculated as:
  • the sum of the data rates transmitted by the small base station n to all UEs it serves cannot exceed its backhaul link bandwidth, so there is The sum of the data rates transmitted by all small base stations in the system cannot exceed the bandwidth of the backhaul link between the macro base station and the MEC server, so there is
  • the ninth constraint condition is due to the bandwidth constraint of the backhaul link, so please combine the above, the ninth constraint sub-condition is expressed as:
  • S344 With the goal of maximizing local revenue, and using the second constraint condition, determine the corresponding offloading decision, spectrum resource, and computing resource allocation of each user terminal in each time period.
  • the local revenue is the sum of the revenue of spectrum resources and the revenue of computing resources.
  • MEC system operators lease spectrum and backhaul link resources from mobile network operators.
  • the unit price of renting the wireless spectrum from the small cell n is ⁇ n per Hz
  • the unit price of renting the backhaul link between the small cell n and the macro cell is ⁇ n per bps.
  • the MEC system operator will charge the UE for transmitting the video data to the UE, and the unit price is defined as ⁇ n per bps (for the UE in the small cell n). Therefore, the net income of the MEC system operator for allocating radio spectrum resources to UE k n is calculated as:
  • the net income of allocating computing resources to UE kn is calculated as:
  • the utility function of the MEC system operator is defined as:
  • u() is a convex function and an increasing function. Because It is always non-negative, and because of the optimality of the problem, this term can always be placed outside the function u(). Because of the increasing and convex nature of the function u(), the Putting it inside the function u() will not affect the optimality of the solution of the problem. Further definition:
  • k n represents the user terminal k corresponding to the small base station n
  • K n represents the set of all user terminals
  • v n represents the unit price charged by the small cell n
  • F represents all local computing resources
  • ⁇ n represents the unit price of the transmission target video data charged to the user terminal in the small cell n
  • B represents the total bandwidth of the local available frequency band
  • ⁇ n represents the unit price of the leased spectrum of the small cell n
  • ⁇ n represents the unit price of the backhaul link between the leased small cell n and the macro base station
  • p n represents the small base station n to the user terminal k
  • ADMM Alternating Direction Method of Multipliers
  • the resource allocation method for video data processing provided in this embodiment optimizes the offloading decision, spectrum resources, and computing resources in the time period based on the storage strategy determined in the optimization period, and partially offloads the strategies, spectrum resources, and computing resources.
  • the allocation scheme is jointly modeled as an optimization problem, which can increase the utilization rate of system storage space and achieve higher system revenue.
  • the resource allocation method for video data processing considers the storage decision-making in video storage and transcoding based on MEC, the joint optimization problem of partial computing offloading decision and resource allocation, and proposes videos with different time scales Storage decision and resource allocation decision plan.
  • a resource allocation device for video data processing is also provided.
  • the device is used to implement the above-mentioned embodiments and preferred implementations, and what has been described will not be repeated.
  • the term "module" can implement a combination of software and/or hardware with predetermined functions.
  • the devices described in the following embodiments are preferably implemented by software, implementation by hardware or a combination of software and hardware is also possible and conceived.
  • This embodiment provides a resource allocation device for video data processing, as shown in FIG. 7, including:
  • the first obtaining module 41 is configured to obtain multiple target video data in each optimization period; wherein, each optimization period is divided into a plurality of consecutive time periods.
  • the second acquisition module 42 is configured to acquire the parameters of each target video data; the parameters of the target video data include length and network data flow.
  • the storage decision module 43 is configured to determine whether to store each target video data based on the parameters of each target video data.
  • the resource allocation module 44 is configured to determine, according to the storage results of the multiple target video data in each optimization period, the corresponding unloading decision, spectrum resource and computing resource allocation of each user terminal in each time period ; Wherein, the offloading decision is used to indicate the percentage of local transcoding operations on the target video data corresponding to each of the user terminals.
  • the resource allocation device for video data processing provided in this embodiment divides the processing of resource allocation into two parts to adapt to different update cycles between different optimization variables, that is, whether to store each target video in each optimization cycle
  • the data is determined, and the determined storage result is used as the input for determining the offloading decision, spectrum resource and computing resource allocation at each time period in the optimization cycle to ensure that resource allocation and storage decision-making are jointly optimized at the same time, and resource utilization is improved.
  • the resource allocation device for video data processing in this embodiment is presented in the form of functional units, where the units refer to ASIC circuits, processors and memories that execute one or more software or fixed programs, and/or other A device that can provide the above-mentioned functions.
  • An embodiment of the present application also provides an electronic device having the resource allocation device for video data processing shown in FIG. 7 above.
  • FIG. 8 is a schematic structural diagram of an electronic device provided by an optional embodiment of the present application.
  • the electronic device may include: at least one processor 51, such as a CPU (Central Processing Unit). Processor), at least one communication interface 53, memory 54, at least one communication bus 52.
  • the communication bus 52 is used to implement connection and communication between these components.
  • the communication interface 53 may include a display screen (Display) and a keyboard (Keyboard), and the optional communication interface 53 may also include a standard wired interface and a wireless interface.
  • the memory 54 may be a high-speed RAM memory (Random Access Memory, volatile random access memory), or a non-volatile memory (non-volatile memory), such as at least one disk memory.
  • the memory 54 may also be at least one storage device located far away from the aforementioned processor 51.
  • the processor 51 may be combined with the device described in FIG. 7, the memory 54 stores an application program, and the processor 51 calls the program code stored in the memory 54 to execute any of the above method steps.
  • the communication bus 52 may be a peripheral component interconnection standard (PCI) bus or an extended industry standard architecture (EISA) bus, etc.
  • PCI peripheral component interconnection standard
  • EISA extended industry standard architecture
  • the communication bus 52 can be divided into an address bus, a data bus, a control bus, and so on. For ease of representation, only one thick line is used in FIG. 8, but it does not mean that there is only one bus or one type of bus.
  • the memory 54 may include a volatile memory (English: volatile memory), such as a random access memory (English: random-access memory, abbreviation: RAM); the memory may also include a non-volatile memory (English: non-volatile memory).
  • memory such as flash memory (English: flash memory), hard disk (English: hard disk drive, abbreviation: HDD) or solid-state hard disk (English: solid-state drive, abbreviation: SSD); memory 54 may also include the above types The combination of memory.
  • the processor 51 may be a central processing unit (English: central processing unit, abbreviation: CPU), a network processor (English: network processor, abbreviation: NP), or a combination of CPU and NP.
  • CPU central processing unit
  • NP network processor
  • the processor 51 may further include a hardware chip.
  • the aforementioned hardware chip may be an application-specific integrated circuit (English: application-specific integrated circuit, abbreviation: ASIC), a programmable logic device (English: programmable logic device, abbreviation: PLD) or a combination thereof.
  • the above-mentioned PLD can be a complex programmable logic device (English: complex programmable logic device, abbreviation: CPLD), field programmable logic gate array (English: field-programmable gate array, abbreviation: FPGA), general array logic (English: generic array) logic, abbreviation: GAL) or any combination thereof.
  • the memory 54 is also used to store program instructions.
  • the processor 51 may call program instructions to implement the resource allocation method for video data processing as shown in the embodiments of FIGS. 1 to 6 of the present application.
  • the embodiment of the present application also provides a non-transitory computer storage medium that stores computer-executable instructions, and the computer-executable instructions can execute the resource allocation for video data processing in any of the foregoing method embodiments method.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (Hard Disk Drive, abbreviation: HDD) or solid-state drive (Solid-State Drive, SSD), etc.; the storage medium may also include a combination of the foregoing types of memories.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)

Abstract

本申请涉及通信技术领域,具体涉及用于视频数据处理的资源分配方法及电子设备,其中方法包括在每个优化周期获取多个目标视频数据;所述每个优化周期划分为多个连续的时间段;获取每个目标视频数据的参数;目标视频数据的参数包括长度以及网络数据流量;基于每个目标视频数据的参数,确定是否存储各个目标视频数据;根据每个优化周期内多个目标视频数据的存储结果,确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。在每个优化周期进行是否存储各个目标视频数据的确定,将所确定的存储结果作为在优化周期内的每个时间段确定卸载决策、频谱资源以及计算资源分配的输入,以保证资源分配与存储决策的同时联合优化。

Description

用于视频数据处理的资源分配方法及电子设备 技术领域
本申请涉通信技术领域,具体涉及用于视频数据处理的资源分配方法及电子设备。
背景技术
因为用户对于视频服务需求的迅猛增长,视频数据传输被认为是下一代移动通信网络的一项重要服务。视频数据的传输通常需要占据大量无线电频谱资源,出于经济和实用角度的考虑,高效的无线传输是优质视频服务的基础。异构网络是下一代无线通信网络的重要特征和关键技术之一,它通过小蜂窝与宏蜂窝之间的重叠覆盖以及小蜂窝的密集部署,可以提高频谱利用率,从而实现网络容量的大幅度提高。
视频数据具有一个重要特征,即可重复利用性。为缓解回程链路负载,节约成本,受用户欢迎的视频内容应该存储于无线接入网内,供其它用户在未来时间里重复利用这些视频数据,这就对无线通信网的存储能力提出了一定要求。由于大量视频内容被多种多样的异构网终端消费,包括PC、智能手机、电视和平板电脑,而这些终端设备需要不同的视频数据速率、格式和分辨率。那么为满足不同的网络状况和匹配不同的终端设备,一个视频内容可能被编码为40多种不同的版本。由于有限的存储空间,存储视频内容的所有版本是很昂贵和不切实际的。因此,为了提高存储效率和节省存储空间,需要使用视频转码技术(Transcoding Technology)来将视频数据转化为不同的版本,以满足不同终端设备的要求。这样理论上只需要在网络中存储一种最高分辨率视频版本。然而,视频数据的转码运算通常需要大量计算资源,由于体积和电池寿命的限制,移动终端难以提供足够多的计算资源,这就要求无线通信网系统有能力为用户提供计算服务。
其中,在视频存储和转码计算服务中,视频片段的存储和替换周期较长,通常为几分钟到数小时,而转码计算和数据传输所涉及的频谱和计算资源分配 的周期通常在毫秒到秒级。因此,视频存储决策和资源分配优化难以同时进行。
发明内容
有鉴于此,本申请实施例提供了一种用于视频数据处理的资源分配方法及电子设备,以解决视频数据处理过程中的资源分配问题。
根据第一方面,本申请实施例提供了一种用于视频数据处理的资源分配方法,包括:
在每个优化周期获取多个目标视频数据;其中,所述每个优化周期划分为多个连续的时间段;
获取每个所述目标视频数据的参数;所述目标视频数据的参数包括长度以及网络数据流量;
基于每个所述目标视频数据的参数,确定是否存储各个所述目标视频数据;
根据每个所述优化周期内所述多个目标视频数据的存储结果,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配;其中,所述卸载决策用于表示本地对各个所述用户终端对应的所述目标视频数据进行转码运算的百分比。
本申请实施例提供的用于视频数据处理的资源分配方法,将资源分配的处理分为两个部分,以适应不同优化变量之间的不同更新周期,即在每个优化周期进行是否存储各个目标视频数据的确定,将所确定的存储结果作为在优化周期内的每个时间段确定卸载决策、频谱资源以及计算资源分配的输入,以保证资源分配与存储决策的同时联合优化,提高了资源利用率。
结合第一方面,在第一方面第一实施方式中,所述基于每个所述目标视频数据的参数,确定是否存储各个所述目标视频数据,包括:
基于每个所述目标视频数据的长度,确定存储各个所述目标视频数据所造成的目标网络数据流量;
利用所述目标视频数据的参数、目标网络数据流量、本地的存储容量以及所述目标网络数据流量,形成是否存储各个所述目标视频数据的第一约束条件;
以存储所有所述目标视频数据的收益最大化为目标,且利用所述第一约束条件确定是否存储各个所述目标视频数据。
本申请实施例提供的用于视频数据处理的资源分配方法,以存储所有目标视频数据的收益最大化为目标,利用第一约束条件对存储决策进行优化,可以保证优化得到的存储决策在满足约束条件的情况下达到存储收益最大化。
结合第一方面第一实施方式,在第一方面第二实施方式中,所述基于每个所述目标视频数据的长度,确定存储所述目标视频数据所造成的目标网络数据流量,包括:
提取各个所述目标视频数据的用户请求到达率;
计算所述用户请求到达率与所述目标视频数据的长度的乘积,得到存储所述目标视频数据所造成的目标网络数据流量。
结合第一方面第一实施方式,在第一方面第三实施方式中,所述利用所述目标视频数据的参数、目标网络数据流量、本地的存储容量以及所述目标网络数据流量,形成是否存储各个所述目标视频数据的第一约束条件,包括:
形成存储所述目标网络数据的决策矢量,以得到第一约束子条件;
计算所述决策矢量与对应的所述目标网络数据的长度的乘积之和,形成第二约束子条件;
计算所述决策矢量与对应的所述目标网络流量的乘积,以形成第三约束子条件;
计算所有所述决策矢量与对应的所述目标网络流量的乘积之和,以形成第四约束子条件。
结合第一方面第三实施方式,在第一方面第四实施方式中,所述第一约束条件以及所述所有所述目标视频数据的收益最大化采用如下公式表示:
Figure PCTCN2019122039-appb-000001
式中,
Figure PCTCN2019122039-appb-000002
1/ι w=L W/r W
Figure PCTCN2019122039-appb-000003
ε>0,γ∈[-1,1];
δ=exp{-Ω 2/2};
其中,C11、C12、C13、C14分别表示所述第一约束子条件、所述第二约束子条件、所述第三约束子条件以及所述第四约束子条件;w∈W=(1,2,...,W)表示每个所述目标视频数据;
Figure PCTCN2019122039-appb-000004
表示对目标视频数据w的存储决定;
Figure PCTCN2019122039-appb-000005
表示h w对应的连续变量;ψ w表示存储目标视频数据w的收益;
Figure PCTCN2019122039-appb-000006
表示存储目标视频数据w的资源开销;L w表示目标视频数据w的长度;Y表示本地的存储容量;1/ι w表示目标视频片段w的传输间隙保持时间;r w表示传输目标视频数据w的速率;λ w表示目标视频数据w的用户请求到达率;ε>0;O +[p w]表示分配给目标视频数据w的最大网络容量;O +[p sum]表示分配给所有目标视频数据的最大网络容量;δ表示违背所述第三约束条件或所述第四约束条件的概率最大值。
本申请实施例提供的用于视频数据处理的资源分配方法,以存储收益最大化为目标,使用基于概率的网络流量约束和存储空间约束,保证优化过程可以容忍优化参数的不确定性并充分利用存储空间,提高了资源分配方法的鲁棒性。
结合第一方面,或第一方面第一实施方式至第三实施方式中任一项,在第一方面第四实施方式中,根据每个所述优化周期内所述多个目标视频数据的存储结果,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配,包括:
基于每个所述优化周期内所述多个目标视频数据的存储结果,确定用户请求所述目标视频数据的命中结果;其中,所述命中结果包括直接命中、转码命中以及未命中;
获取本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率;其中,所述小蜂窝包括所述小基站以及对应的所述用户终端;
利用所述目标视频数据的命中结果、本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率,形成确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配的第二约束条件;
以本地收益最大化为目标,且利用所述第二约束条件,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配;其中,所述本地收益为所述频谱资源的收益与所述计算资源的收益之和。
结合第一方面第五实施方式,在第一方面第六实施方式中,所述利用所述 目标视频数据的命中结果、本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率,形成确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配的第二约束条件,包括:
计算本地为各个所述用户终端分配的计算资源、对应的所述命中结果以及对应的所述卸载决策的乘积之和,形成第五约束子条件;
利用本地为各个所述用户终端分配的计算资源以及所述用户终端自身的计算能力,形成第六约束子条件;
利用本地为各个所述用户终端分配的频谱资源形成第七约束子条件;
计算本地为各个所述用户终端分配的频谱资源、所述本地的可用频带总带宽以及所述频谱效率的乘积,以形成第八约束子条件;
计算所有本地为各个所述用户终端分配的频谱资源、所述本地的可用频带总带宽以及所述频谱效率的乘积之和,以形成第九约束条件。
结合第一方面第六实施方式,在第一方面第七实施方式中,所述第二约束条件以及所述本地收益最大化采用如下公式表示:
Figure PCTCN2019122039-appb-000007
式中,
Figure PCTCN2019122039-appb-000008
Figure PCTCN2019122039-appb-000009
Figure PCTCN2019122039-appb-000010
Figure PCTCN2019122039-appb-000011
其中,k n表示小基站n对应的用户终端k;K n表示所有用户终端的集合;
Figure PCTCN2019122039-appb-000012
表示用户终端k n对应的卸载决策;
Figure PCTCN2019122039-appb-000013
表示本地为用户终端k n分配的计算资源;v n表示小蜂窝n收取费用的单价;F表示本地的全部计算资源;
Figure PCTCN2019122039-appb-000014
表示本地执行的用户终端k n的计算任务;
Figure PCTCN2019122039-appb-000015
表示本地为用户终端k n分配的频谱资源;θ n表示向小蜂窝n中的用户终端所收取的传输目标视频数据的单价;B表示本地的可用频带总带宽;
Figure PCTCN2019122039-appb-000016
表示对应于用户终端k n的频谱效率;υ n表示小蜂窝n租借频谱的单价;η n表示租借小蜂窝n与宏基站之间回程链路的单价;p n表示小基站n向用户终端k n传输时的功率密度;
Figure PCTCN2019122039-appb-000017
Figure PCTCN2019122039-appb-000018
分别表示小基站n到用户终端k n和小基站m到用户终端k n的信道增益,m≠n;σ表示加性高斯白噪声的功率谱密度;
Figure PCTCN2019122039-appb-000019
表示用户终端k n自身的计算能力;
Figure PCTCN2019122039-appb-000020
表示用户终端k n对所请求的目标视频数据所需的最小传输速率;L n表示小蜂窝n与宏基站的回程链路带宽。
本申请实施例提供的用于视频数据处理的资源分配方法,基于优化周期内确定出的存储策略对时间段内的卸载决策、频谱资源以及计算资源进行优化,将部分卸载策略、频谱资源和计算资源分配方案联合建模为一个优化问题,可以增加系统存储空间利用率并实现较高的系统收益。
根据第二方面,本申请实施例还提供了一种电子设备,包括:
存储器和处理器,所述存储器和所述处理器之间互相通信连接,所述存储器中存储有计算机指令,所述处理器通过执行所述计算机指令,从而执行本申请第一方面,或第一方面任一项实施方式中所述的用于视频数据处理的资源分配方法。
根据第三方面,本申请实施例还提供了一种计算机可读存储介质,所述计算机可读存储介质存储有计算机指令,所述计算机指令用于使所述计算机执行本申请第一方面,或第一方面任一项实施方式中所述的用于视频数据处理的资源分配方法。
附图说明
为了更清楚地说明本申请具体实施方式或现有技术中的技术方案,下面将对具体实施方式或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图是本申请的一些实施方式,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1是根据本申请实施例的通信网络系统的结构示意图;
图2是根据本申请实施例的用于视频数据处理的资源分配方法的流程图;
图3是根据本申请实施例的时间轴划分示意图;
图4是根据本申请实施例的用于视频数据处理的资源分配方法的流程图;
图5是根据本申请实施例的用于视频数据处理的资源分配方法的流程图;
图6是根据本申请实施例的不同命中结果对应的数据流向示意图;
图7是根据本申请实施例的用于视频数据处理的资源分配装置的结构框图;
图8是本申请实施例提供的电子设备的硬件结构示意图。
具体实施方式
为使本申请实施例的目的、技术方案和优点更加清楚,下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
需要说明的是,本申请实施例中所述的用于视频数据处理的资源分配方法是基于移动边缘计算(Mobile Edge Computing,简称为MEC)进行的,用以保证资源分配与存储决策的同时联合优化,提高视频传输中融合系统回程链路资源和存储资源的利用率。
图1示出了本申请实施例中所述的通信网络系统,如图1所示,本申请实施例采用双层异构网络(宏蜂窝和小蜂窝)作为无线传输场景,所述通信网络系统包括MEC服务器、宏基站以及多个小蜂窝(图1中仅示出了2个小蜂窝)。 其中,宏基站与MEC服务器连接,多个小蜂窝与宏基站连接,以通过宏基站从MEC服务器中获取视频数据。所述的小蜂窝包括小基站以及多个用户终端(User Equipment,简称为UE),每个小蜂窝中的用户终端通过对应的小基站与宏基站通信,以从MEC服务器中获取视频数据。
其中,MEC服务器在接收到用户终端的获取视频数据的请求之后,对相应的视频数据进行处理后发送给该用户终端。在本申请实施例的描述中视频数据处理以视频存储和转码为例,在进行视频数据的处理过程中就需要考虑资源利用率的问题。其中,所述的转码为MEC服务器中所存储的视频数据的分辨率与用于所请求的视频数据的分辨率不一致,就需要对所存储的视频数据进行分辨率的处理,即为所述的转码。具体地,在本申请实施例中联合优化视频存储决策、计算任务卸载以及频谱和计算资源分配问题;为了适应不同优化变量之间的不同更新周期,将资源分配的优化过程分为两个阶段,分别为目标视频数据存储决策阶段和频谱及计算资源分配阶段。其中,在资源分配的优化过程中,为了容忍优化参数的不确定性,使用鲁棒优化算法进行视频存储决策,以充分利用存储资源。更为具体的描述将在下文中进行详细描述。
根据本申请实施例,提供了一种用于视频数据处理的资源分配方法实施例,需要说明的是,在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行,并且,虽然在流程图中示出了逻辑顺序,但是在某些情况下,可以以不同于此处的顺序执行所示出或描述的步骤。
在本实施例中提供了一种用于视频数据处理的资源分配方法,可用于电子设备,例如图1中的MEC服务器等,在下文中的描述中以MEC服务器为例进行详细描述。图2是根据本申请实施例的用于视频数据处理的资源分配方法的流程图,如图2所示,该流程包括如下步骤:
S11,在每个优化周期获取多个目标视频数据。
其中,所述每个优化周期划分为多个连续的时间段。
在对该步骤进行描述之前,对所涉及到的优化周期以及时间段进行解释如下,如图3所示,时间线划分为首尾相接的优化周期,每个优化周期的长度为 T;优化周期T又划分为若干时间段,时间段的长度为t。
其中,T是一个可以由用户偏好决定的参数。当T比较大的时候,意味着视频片段将被系统存储较长的时间,这可以降低存储替换的系统开销,但视频数据的更新速度会较低。当T较小的时候,存储替换的开销较高,但这带来了较高的视频数据更新速度。通常T取值数分钟到数小时。每个优化周期T由若干时间段t组成,t的取值通常为数毫秒到数秒。
如图3所示,本实施例中的资源分配方法划分为两个阶段进行处理,即在每一个优化周期T开始的时候,系统将执行第一阶段优化,即进行视频存储决策。之后系统按决策结果将相应视频数据暂存下来,直到下一优化周期开始。在本优化周期中,存储决策将作为第二阶段优化即资源分配的输入。第二阶段优化在每个时间段开始时执行一次。
在每个优化周期中网络中存在W个目标视频数据,每一个目标视频数据可以编码为J种级别的分辨率。使用w∈W=(1,2,...,W)来代表每个目标视频数据的内容,在本实施例中,以MEC服务器仅存储每个目标视频数据的最高分辨率版本,那么在这种情况下,每个被请求的目标视频数据版本可能通过如下三种方式中的一种获取:(1)直接存储命中:MEC服务器存储了被请求的视频片段,而UE请求的是该视频片段的最高分辨率版本。(2)转码命中:MEC服务器存储了被请求的视频片段,但UE请求的是较低分辨率的版本,因此需要执行转码运算来完成这次视频服务。(3)存储未命中:MEC服务器没有存储被请求的视频片段,请求被转发至因特网中的源服务器。在此情况下,在视频内容到达后,MEC服务器可以决定是否存储此内容的最高分辨率版本。
S12,获取每个目标视频数据的参数。
其中,所述目标视频数据的参数包括长度以及网络数据流量。
MEC服务器在获取到多个目标视频数据之后,通过对其进行解析就可以得到每个目标视频数据的长度以及对应的网络数据流量。
S13,基于每个目标视频数据的参数,确定是否存储各个目标视频数据。
MEC服务器利用每个目标视频数据的参数,并结合自身的一些参数(例如,存储容量、存储该目标视频数据所带来的目标流量数据的消耗)等等,或者也 可以结合当前网络性能确定是否存储所获取到的目标视频数据。在本申请中对确定是否存储各个目标视频数据的方法并不做任何限制,具体将在下文中进行详细描述。
其中,使用
Figure PCTCN2019122039-appb-000021
来代表对视频片段w的存储决定。具体来讲,如果MEC服务器决定存储此视频片段,设置h w=1;如果决定不存储此视频,设置h w=0。采用作为h={h w} w∈W所获取到的所有目标视频内容的存储决定矢量。
S14,根据每个优化周期内多个目标视频数据的存储结果,确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。
其中,所述卸载决策用于表示本地对各个用户终端对应的目标视频数据进行转码运算的百分比。
MEC服务器在每个优化周期确定对于每个目标视频数据是否存储之后,就可以在该优化周期内的每个时间段利用对应的存储决定,为用户终端进行资源的分配,具体地为用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。
当用户终端性MEC服务器发送视频数据请求且一个转码命中事件发生时,MEC服务器需要将被请求视频的最高分辨率版本转码为UE所需的较低分辨率的版本。或者,此视频片段的最高版本可以通过小蜂窝基站被发送到目的地UE处,由UE自行进行转码运算。为了利用UE和MEC之间平行计算的优势,在本申请实施例中,采用部分计算卸载机制。
在部分计算卸载中,被请求的目标视频数据的一部分由MEC进行转码运算,而剩余部分被发送至UE处,由用户终端进行转码运算。使用
Figure PCTCN2019122039-appb-000022
来代表对应UE k n的计算卸载决定。
具体来讲,
Figure PCTCN2019122039-appb-000023
代表目标视频数据由MEC服务器进行转码运算的百分比,剩余部分发送至UE处执行转码。其中,
Figure PCTCN2019122039-appb-000024
代表全部的转码运算由MEC服务器执行,而
Figure PCTCN2019122039-appb-000025
表示全部转码运算由UE完成。使用
Figure PCTCN2019122039-appb-000026
作为所有UE的计算卸载决策矢量。
使用
Figure PCTCN2019122039-appb-000027
代表小蜂窝n分配给UE k n用于传输目标视频数据的带宽 所占总带宽的比例,因此
Figure PCTCN2019122039-appb-000028
必须成立。使用
Figure PCTCN2019122039-appb-000029
k n∈K n,n∈N作为所有UE的频谱分配矢量。
其中,MEC服务器在利用每个目标视频数据的存储结果确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配时,可以利用传输该目标视频数据所需要的带宽资源、频率资源与该系统所能够提供的带宽资源以及频谱资源进行对比分析进行确定。或者也可以采用其他方式进行确定等等,在此对具体的确定方法并不做任何限制。
本实施例提供的用于视频数据处理的资源分配方法,将资源分配的处理分为两个部分,以适应不同优化变量之间的不同更新周期,即在每个优化周期进行是否存储各个目标视频数据的确定,将所确定的存储结果作为在优化周期内的每个时间段确定卸载决策、频谱资源以及计算资源分配的输入,以保证资源分配与存储决策的同时联合优化,提高了资源利用率。
在本实施例中提供了一种用于视频数据处理的资源分配方法,可用于电子设备,例如服务器等,图4是根据本申请实施例的用于视频数据处理的资源分配方法的流程图,如图4所示,该流程包括如下步骤:
S21,在每个优化周期获取多个目标视频数据。
其中,所述每个优化周期划分为多个连续的时间段。
详细请参见图2所示实施例的S11,在此不再赘述。
S22,获取每个目标视频数据的参数。
其中,所述目标视频数据的参数包括长度以及网络数据流量。
详细请参见图2所示实施例的S12,在此不再赘述。
S23,基于每个目标视频数据的参数,确定是否存储各个目标视频数据。
在本实施例中,采用优化的方式确定是否存储各个目标视频数据,其中,优化的约束条件是基于目标视频数据的参数确定的,目标优化函数为存储所有目标视频数据的收益最大化。具体地,上述S23包括以下步骤:
S231,基于每个目标视频数据的长度,确定存储各个目标视频数据所造成的目标网络数据流量。
MEC服务器在确定目标视频数据所造成的目标网络数据流量时,还结合对 于每个目标视频数据的用户请求达到率,以保证确定出的目标网络数据流量的可靠性。具体地,上述S231可以采用如下步骤实现:
(1)提取各个目标视频数据的用户请求到达率。
MEC设对于目标视频数据w的用户请求到达率服从泊松分布并用λ w(Requests/second)来表示此到达率。
(2)计算用户请求到达率与目标视频数据的长度的乘积,得到存储目标视频数据所造成的目标网络数据流量。
定义ρ w=L wλ w(bit/s)为传输存储的目标视频数据w所造成的目标网络数据流量。因此,传输所有被存储的目标视频数据所造成的网络数据流量计算为
Figure PCTCN2019122039-appb-000030
S232,利用目标视频数据的参数、目标网络数据流量、本地的存储容量以及目标网络数据流量,形成是否存储各个目标视频数据的第一约束条件。
MEC设备在得到存储目标网络数据流量之后,就可以结合各个目标视频数据的参数、目标网络数据流量、本地的存储容量形成是否存储各个目标视频数据的第一约束条件。具体地,上述S232包括以下步骤:
(1)形成存储目标网络数据的决策矢量,以得到第一约束子条件。
第一约束子条件用于保证存储决策变量h w是二进制变量,即,
Figure PCTCN2019122039-appb-000031
为了后续优化的处理,可以将二进制变量{h w}∈[0,1]松弛为连续变量
Figure PCTCN2019122039-appb-000032
那么第一约束子条件就可以表示为:
Figure PCTCN2019122039-appb-000033
(2)计算决策矢量与对应的目标网络数据的长度的乘积之和,形成第二约束子条件。
第二约束子条件用于表明所有被存储的目标视频数据的数据量之和不超过MEC服务器的存储容量限制Y。具体地,利用决策矢量与对应的目标网络数据的长度的乘积,得到存储各个目标网络数据的数据量;再计算所有目标网络数据的数据量之和,以形成第二约束子条件。即,第二约束子条件可以表示为:
Figure PCTCN2019122039-appb-000034
(3)计算决策矢量与对应的目标网络流量的乘积,以形成第三约束子条件。
第三约束子条件用于表示存储各个目标视频数据所引起的网络数据流量不超过MEC服务器为其分配的网络数据流量,使用C(bit/s)来代表网络容量。进一步地,为了保证不同视频片段之间传输机会的公平性,使用C w来代表为传输 视频片段w分配的网络容量。C和{C w}取决于带宽分配矢量
Figure PCTCN2019122039-appb-000035
只有当如下条件满足时,系统才能够稳定,即,第三约束子条件表示为:
Figure PCTCN2019122039-appb-000036
(4)计算所有决策矢量与对应的目标网络流量的乘积之和,以形成第四约束子条件。
第四约束子条件用于表示存储所有目标视频数据所引起的网络数据流量不超过MEC服务器的网络容量。即,
Figure PCTCN2019122039-appb-000037
需要将无线电频谱分配矢量s从第三约束子条件以及后续的第四约束子条件中分离出去,采用一种基于反馈的方法来提供一个对网络容量(用于视频数据传输)的估计值。假设在时刻t为目标视频数据w分配的最大网络容量是
Figure PCTCN2019122039-appb-000038
预计分配给视频片段w的最大网络容量计算为:
Figure PCTCN2019122039-appb-000039
式中v∈[0,1]是一个常量,用于调整当前网络状态和之前网络状态之间的比例;上标[t-1]代表最后评估时间。同理,令
Figure PCTCN2019122039-appb-000040
代表用于传输所有被存储的目标视频数据的最大网络容量。则预计中总网络容量为:
Figure PCTCN2019122039-appb-000041
那么,第三约束子条件可以重新表示为:
Figure PCTCN2019122039-appb-000042
类似地,第四约束子条件可以表示为:
Figure PCTCN2019122039-appb-000043
再进一步地,定义参数λ w的长期均值为
Figure PCTCN2019122039-appb-000044
并且提出准确的请求到达率和长期均值之间的关系建立在有界随机参数γ和之上:
Figure PCTCN2019122039-appb-000045
式中ε>0代表影响参数λ w不确定性的最大程度,而γ是一个取值于区间[-1,1]之内的零均值随机参数;参数γ反映请求到达率的可能波动。这个表达式意味着实际中的请求到达率λ w可按不超过
Figure PCTCN2019122039-appb-000046
的幅度偏离请求到达率的估计 值
Figure PCTCN2019122039-appb-000047
因此,可能的偏离程度实际上是由参数ε控制的。因为较大的ε可以带来较好的鲁棒性,而较小的ε可以带来更可靠的资源预留,MEC服务器可以根据鲁棒等级要求和历史统计数据来调整此参数,以实现鲁棒性能和资源预留程度之间的平衡。
至此,问题符合鲁棒问题的定义。定义参数δ为置信等级,含义是违背第三约束子条件或第四约束子条件的概率最大是δ。那么,第三约束子条件进一步表示为:
Figure PCTCN2019122039-appb-000048
相应地,第四约束子条件可以表示为:
Figure PCTCN2019122039-appb-000049
式中Ω和δ的关系为:δ=exp{-Ω 2/2},另外
Figure PCTCN2019122039-appb-000050
S233,以存储所有目标视频数据的收益最大化为目标,且利用第一约束条件确定是否存储各个目标视频数据。
在上述S232中确定出各个约束条件之后,就可以以存储所有目标视频数据的收益最大化为目标,对存储矢量进行优化了。具体地,使用ψ w来代表存储目标视频数据w的收益,而存储目标视频数据w的存储资源开销为
Figure PCTCN2019122039-appb-000051
因此,存储视频片段w的净收益为
Figure PCTCN2019122039-appb-000052
那么,存储所有目标视频数据的收益最大化的目标函数可以表示为:
Figure PCTCN2019122039-appb-000053
由于二进制变量{h w}∈[0,1]松弛为连续变量
Figure PCTCN2019122039-appb-000054
那么该目标函数可以表示为:
Figure PCTCN2019122039-appb-000055
那么,在确定出第一约束条件,以及目标函数之后,就可以对存储矢量h进行优化,具体地,第一约束条件以及所有目标视频数据的收益最大化采用如下公式表示:
Figure PCTCN2019122039-appb-000056
式中,
Figure PCTCN2019122039-appb-000057
1/ι w=L W/r W
Figure PCTCN2019122039-appb-000058
ε>0,γ∈[-1,1];
δ=exp{-Ω 2/2};
其中,C11、C12、C13、C14分别表示所述第一约束子条件、所述第二约束子条件、所述第三约束子条件以及所述第四约束子条件;w∈W=(1,2,...,W)表示每个所述目标视频数据;
Figure PCTCN2019122039-appb-000059
表示对目标视频数据w的存储决定;
Figure PCTCN2019122039-appb-000060
表示h w对应的连续变量;ψ w表示存储目标视频数据w的收益;
Figure PCTCN2019122039-appb-000061
表示存储目标视频数据w的资源开销;L w表示目标视频数据w的长度;Y表示本地的存储容量;1/ι w表示目标视频片段w的传输间隙保持时间;r w表示传输目标视频数据w的速率;λ w表示目标视频数据w的用户请求到达率;ε>0;O +[p w]表示分配给目标视频数据w的最大网络容量;O +[p sum]表示分配给所有目标视频数据的最大网络容量;δ表示违背所述第三约束条件或所述第四约束条件的概率最大值。
上述问题为凸问题,容易求解。在优化阶段存储决策向量h={h w} w∈W确定后,就可以进入第二阶段的优化。
S24,根据每个优化周期内多个目标视频数据的存储结果,确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。
其中,所述卸载决策用于表示本地对各个用户终端对应的目标视频数据进行转码运算的百分比。
详细请参见图2所示实施例的S14,在此不再赘述。
本实施例提供的用于视频数据处理的资源分配方法,以存储所有目标视频数据的收益最大化为目标,利用第一约束条件对存储决策进行优化,可以保证优化得到的存储决策在满足约束条件的情况下达到存储收益最大化。
在本实施例中提供了一种用于视频数据处理的资源分配方法,可用于电子设备,例如服务器等,图5是根据本申请实施例的用于视频数据处理的资源分 配方法的流程图,如图5所示,该流程包括如下步骤:
S31,在每个优化周期获取多个目标视频数据。
其中,所述每个优化周期划分为多个连续的时间段。
详细请参见图4所示实施例的S21,在此不再赘述。
S32,获取每个目标视频数据的参数。
其中,所述目标视频数据的参数包括长度以及网络数据流量。
详细请参见图4所示实施例的S22,在此不再赘述。
S33,基于每个目标视频数据的参数,确定是否存储各个目标视频数据。
详细请参见图4所示实施例的S23,在此不再赘述。
S34,根据每个优化周期内多个目标视频数据的存储结果,确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。
其中,所述卸载决策用于表示本地对各个用户终端对应的目标视频数据进行转码运算的百分比。
在本实施例中,同样采用优化的方式,对各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配进行优化,在下文中将对优化的各个约束条件以及目标函数进行详细说明。在第一阶段存储决策向量h={h w} w∈W确定后,进入第二阶段优化,对频谱分配向量
Figure PCTCN2019122039-appb-000062
计算卸载决策向量
Figure PCTCN2019122039-appb-000063
和计算资源分配向量
Figure PCTCN2019122039-appb-000064
进行联合优化。具体地,上述S34包括以下步骤:
S341,基于每个优化周期内多个目标视频数据的存储结果,确定用户请求目标视频数据的命中结果。
其中,所述命中结果包括直接命中、转码命中以及未命中。
MEC服务器已经在每个优化周期内确定所获取到的目标视频数据的存储结果,在接收到的用户获取目标视频数据的请求之后,就可以基于用户请求进行命中结果的判断。当MEC服务器中存储有该目标视频数据,且用户请求的是该目标视频数据的最高分辨率形式,则此时的命中结果为直接命中;当MEC服务器中存储有该目标视频数据,但是用户请求的是该目标视频数据的低分辨率形式,则此时的命中结果为转码命中;当MEC服务器中并未存储有该目标 视频数据,则此时的命中结果为未命中。
使用
Figure PCTCN2019122039-appb-000065
来指示UE k n和MEC服务器之间的存储命中情况。令
Figure PCTCN2019122039-appb-000066
代表直接存储命中或是转码命中事件,用
Figure PCTCN2019122039-appb-000067
代表存储未命中事件。同时,引入指示向量
Figure PCTCN2019122039-appb-000068
来区别直接存储命中事件和转码命中事件。使用
Figure PCTCN2019122039-appb-000069
来代表UE k n的直接存储命中,而
Figure PCTCN2019122039-appb-000070
代表转码命中。给定由第一阶段优化得来的存储决策向量
Figure PCTCN2019122039-appb-000071
之后,可以准确地知道每个UE的存储命中情况,即
Figure PCTCN2019122039-appb-000072
Figure PCTCN2019122039-appb-000073
现在是常数向量。
其中,图6示出了不同命中结果下,视频数据的传输过程。当命中结果为直接存储命中时,用户终端直接从MEC服务器中获取目标视频数据;当命中结果为转码命中时,MEC服务器对相应的目标视频数据进行转码后,发送给相应的用户终端;当命中结果为未命中时,用户终端通过MEC服务器、核心网以及英特网从源服务器中获取到相应的视频数据。
S342,获取本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向用户终端传输数据的频谱效率。
其中,所述小蜂窝包括小基站以及对应的用户终端。
本地的一些参数可以事先存储的MEC服务器的内存中,也可以是MEC服务器在需要时从外界获取到等等。具体地,认为所有小蜂窝重叠使用同一段频谱,因此小蜂窝之间存在干扰。只存在下行链路传输,即无线传输是从小基站到从属于它的UE,而干扰是从其它小基站到此UE处。MEC服务器可用的频带总宽度是B Hz。宏基站与MEC服务器之间的回程链路带宽L bps,小蜂窝n与宏基站的回程链路带宽是L n bps。
S343,利用目标视频数据的命中结果、本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向用户终端传输数据的频谱效率,形成确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配的第二约束条件。
具体地,上述S343可以采用如下步骤实现:
(1)计算本地为各个用户终端分配的计算资源、对应的命中结果以及对应的卸载决策的乘积之和,形成第五约束子条件。
第五约束子条件是用于保证在整个系统中分配给所有用户终端的计算资源不超过MEC服务器的计算资源总量。
定义
Figure PCTCN2019122039-appb-000074
为MEC服务器为UE k n分配的服务器计算资源占计算资源总量的百分比。因此有
Figure PCTCN2019122039-appb-000075
使用
Figure PCTCN2019122039-appb-000076
作为对所有UE的计算资源分配矢量。
那么,第五约束子条件可以表示为:
Figure PCTCN2019122039-appb-000077
(2)利用本地为各个用户终端分配的计算资源以及用户终端自身的计算能力,形成第六约束子条件。
第六约束子条件是用于保证分配给每个用户终端的计算资源不小于它自身的计算能力,否则计算任务卸载将失去意义。其中,采用
Figure PCTCN2019122039-appb-000078
表示UE k n自身的计算能力,而F代表MEC服务器的全部计算资源。对于UE k n的转码操作,假定需要完成的计算任务量为
Figure PCTCN2019122039-appb-000079
具体来讲,
Figure PCTCN2019122039-appb-000080
代表完成该计算任务所需要的CPU周期总数。
Figure PCTCN2019122039-appb-000081
的值取决于UE k n所请求的目标视频数据以及它请求的分辨率级别。那么,第六约束子条件可以表示为:
Figure PCTCN2019122039-appb-000082
其中,
Figure PCTCN2019122039-appb-000083
的倒数是MEC服务器执行计算任务
Figure PCTCN2019122039-appb-000084
的时间消耗,而
Figure PCTCN2019122039-appb-000085
的倒数是UE k n自行执行此计算任务的时间。这意味着为每单位计算任务分配的计算资源多少可以反映执行此计算任务的时间消耗。
(3)利用本地为各个用户终端分配的频谱资源形成第七约束子条件。
第七约束子条件用来表示在每个小蜂窝之中,分配给所有用户的频谱不重合,以符合系统模型的要求。那么,第七约束子条件表示为:
Figure PCTCN2019122039-appb-000086
(4)计算本地为各个用户终端分配的频谱资源、本地的可用频带总带宽以及频谱效率的乘积,以形成第八约束子条件。
第八约束子条件用于保证对每个用户终端传输的数据速率不小于它请求的视频数据所需的最低数据速率
Figure PCTCN2019122039-appb-000087
其中,小基站n向UE k n传输数据的频谱效率为:
Figure PCTCN2019122039-appb-000088
式中,p n是小基站n向UE k n传输时的功率密度,而
Figure PCTCN2019122039-appb-000089
Figure PCTCN2019122039-appb-000090
分别是小基站n到UE k n和小基站m到UE k n的信道增益,m≠n;σ代表加性高斯白噪声的功率谱密度。
小基站n向UE k n传输数据的瞬时数据速率计算为:
Figure PCTCN2019122039-appb-000091
小基站n向其所服务的所有UE传输的数据速率之和不能超过其回程链路带宽,因此有
Figure PCTCN2019122039-appb-000092
系统中所有小基站传输的数据速率之和不能超过宏基站与MEC服务器之间的回程链路带宽,因此有
Figure PCTCN2019122039-appb-000093
那么,第八约束子条件表示为:
Figure PCTCN2019122039-appb-000094
(5)计算所有本地为各个用户终端分配的频谱资源、本地的可用频带总带宽以及频谱效率的乘积之和,以形成第九约束子条件。
第九约束条件是因为回程链路的带宽约束,那么请结合上文,第九约束子条件表示为:
Figure PCTCN2019122039-appb-000095
S344,以本地收益最大化为目标,且利用第二约束条件,确定各个用户终端在每个时间段内对应的卸载决策、频谱资源以及计算资源的分配。
其中,所述本地收益为频谱资源的收益与计算资源的收益之和。
(1)频谱资源的净收益
MEC系统运营商向移动网络运营商租借频谱和回程链路资源。从小蜂窝n租借无线频谱的单价是υ nper Hz,而租借小蜂窝n和宏蜂窝之间回程链路的单价是η n per bps。MEC系统运营商会向UE收取传输视频数据到UE的费用,单价定义为θ n per bps(针对小蜂窝n之中的UE)。因此,MEC系统运营商向UE k n分配无线电频谱资源的净收益计算为:
Figure PCTCN2019122039-appb-000096
(2)计算资源的净收益
MEC系统运营商只对分配给每单位计算任务的MEC计算资源和每单位计算任务对应的UE计算能力之间的差值收取费用。而对小蜂窝n收取费用的单位价格是νn。所以对UE kn分配计算资源的净收益计算为:
Figure PCTCN2019122039-appb-000097
只有当一个转码命中事件发生时,MEC系统运营商才会向UE分配计算资源,但是会向所有UE分配频谱资源,因为在全部三种存储命中情况之下都会产生数据传输。因此定义MEC系统运营商的效用函数为:
Figure PCTCN2019122039-appb-000098
式中u()是一个凸函数并且是一个增函数。因为
Figure PCTCN2019122039-appb-000099
总是非负,另外因为问题最优性考虑,这一项总是可以被放在函数u()之外。因为函数u()的增函数和凸函数性质,把
Figure PCTCN2019122039-appb-000100
放在函数u()内部不会影响问题的解的最优性。进一步定义:
Figure PCTCN2019122039-appb-000101
即是说,MEC系统运营商达到最大收益。令
Figure PCTCN2019122039-appb-000102
下面将使用
Figure PCTCN2019122039-appb-000103
作为优化问题的目标函数。
具体地,所述第二约束条件以及本地收益最大化采用如下公式表示:
Figure PCTCN2019122039-appb-000104
式中,
Figure PCTCN2019122039-appb-000105
Figure PCTCN2019122039-appb-000106
Figure PCTCN2019122039-appb-000107
Figure PCTCN2019122039-appb-000108
其中,k n表示小基站n对应的用户终端k;K n表示所有用户终端的集合;
Figure PCTCN2019122039-appb-000109
表示用户终端k n对应的卸载决策;
Figure PCTCN2019122039-appb-000110
表示本地为用户终端k n分配的计算资源;v n表示小蜂窝n收取费用的单价;F表示本地的全部计算资源;
Figure PCTCN2019122039-appb-000111
表示本地执行的用户终端k n的计算任务;
Figure PCTCN2019122039-appb-000112
表示本地为用户终端k n分配的频谱资源;θ n表示向小蜂窝n中的用户终端所收取的传输目标视频数据的单价;B表示本地的可用频带总带宽;
Figure PCTCN2019122039-appb-000113
表示对应于用户终端k n的频谱效率;υ n表示小蜂窝n租借频谱的单价;η n表示租借小蜂窝n与宏基站之间回程链路的单价;p n表示小基站n向用户终端k n传输时的功率密度;
Figure PCTCN2019122039-appb-000114
Figure PCTCN2019122039-appb-000115
分别表示小基站n到用户终端k n和小基站m到用户终端k n的信道增益,m≠n;σ表示加性高斯白噪声的功率谱密度;
Figure PCTCN2019122039-appb-000116
表示用户终端k n自身的计算能力;
Figure PCTCN2019122039-appb-000117
表示用户终端k n对所请求的目标视频数据所需的最小传输速率;L n表示小蜂窝n与宏基站的回程链路带宽。
此问题经过乘积项替换后即成为一个凸优化问题,容易求解。为降低信令开销,使用分布式算法(Alternating Direction Method of Multipliers,简称为ADMM)来求解。
本实施例提供的用于视频数据处理的资源分配方法,基于优化周期内确定出的存储策略对时间段内的卸载决策、频谱资源以及计算资源进行优化,将部分卸载策略、频谱资源和计算资源分配方案联合建模为一个优化问题,可以增加系统存储空间利用率并实现较高的系统收益。
本申请实施例提供的用于视频数据处理的资源分配方法,考虑基于MEC的视频存储与转码中的存储决策、部分计算卸载决策及资源分配的联合优化问 题,提出具有不同的时间尺度的视频存储决策和资源分配决策方案。首先对优化周期的视频存储决策优化。针对视频存储决策的优化过程中用户请求到达的不确定性和传统方案中常数网络流量约束造成的存储空间浪费问题,将存储决策问题建模为一个鲁棒优化问题。该问题以存储收益最大化为目标,使用基于概率的网络流量约束和存储空间约束,保证优化模型可以容忍优化参数的不确定性并充分利用存储空间,提高算法鲁棒性。其次,基于优化周期视频存储最优决策,研究时间段的视频传输频谱资源和转码计算资源分配问题,采用部分计算卸载机制,以最大化系统收益为目标,将部分卸载策略、频谱和计算资源分配方案联合建模为一个优化问题。为降低信令开销并减小计算复杂度,使用一种分布式算法。本申请可以显著增加存储空间利用率并实现较高的系统收益。
在本实施例中还提供了一种用于视频数据处理的资源分配装置,该装置用于实现上述实施例及优选实施方式,已经进行过说明的不再赘述。如以下所使用的,术语“模块”可以实现预定功能的软件和/或硬件的组合。尽管以下实施例所描述的装置较佳地以软件来实现,但是硬件,或者软件和硬件的组合的实现也是可能并被构想的。
本实施例提供一种用于视频数据处理的资源分配装置,如图7所示,包括:
第一获取模块41,用于在每个优化周期获取多个目标视频数据;其中,所述每个优化周期划分为多个连续的时间段。
第二获取模块42,用于获取每个所述目标视频数据的参数;所述目标视频数据的参数包括长度以及网络数据流量。
存储决策模块43,用于基于每个所述目标视频数据的参数,确定是否存储各个所述目标视频数据。
资源分配模块44,用于根据每个所述优化周期内所述多个目标视频数据的存储结果,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配;其中,所述卸载决策用于表示本地对各个所述用户终端对应的所述目标视频数据进行转码运算的百分比。
本实施例提供的用于视频数据处理的资源分配装置,将资源分配的处理分 为两个部分,以适应不同优化变量之间的不同更新周期,即在每个优化周期进行是否存储各个目标视频数据的确定,将所确定的存储结果作为在优化周期内的每个时间段确定卸载决策、频谱资源以及计算资源分配的输入,以保证资源分配与存储决策的同时联合优化,提高了资源利用率。
本实施例中的用于视频数据处理的资源分配装置是以功能单元的形式来呈现,这里的单元是指ASIC电路,执行一个或多个软件或固定程序的处理器和存储器,和/或其他可以提供上述功能的器件。
上述各个模块的更进一步的功能描述与上述对应实施例相同,在此不再赘述。
本申请实施例还提供一种电子设备,具有上述图7所示的用于视频数据处理的资源分配装置。
请参阅图8,图8是本申请可选实施例提供的一种电子设备的结构示意图,如图8所示,该电子设备可以包括:至少一个处理器51,例如CPU(Central Processing Unit,中央处理器),至少一个通信接口53,存储器54,至少一个通信总线52。其中,通信总线52用于实现这些组件之间的连接通信。其中,通信接口53可以包括显示屏(Display)、键盘(Keyboard),可选通信接口53还可以包括标准的有线接口、无线接口。存储器54可以是高速RAM存储器(Random Access Memory,易挥发性随机存取存储器),也可以是非不稳定的存储器(non-volatile memory),例如至少一个磁盘存储器。存储器54可选的还可以是至少一个位于远离前述处理器51的存储装置。其中处理器51可以结合图7所描述的装置,存储器54中存储应用程序,且处理器51调用存储器54中存储的程序代码,以用于执行上述任一方法步骤。
其中,通信总线52可以是外设部件互连标准(peripheral component interconnect,简称PCI)总线或扩展工业标准结构(extended industry standard architecture,简称EISA)总线等。通信总线52可以分为地址总线、数据总线、控制总线等。为便于表示,图8中仅用一条粗线表示,但并不表示仅有一根总线或一种类型的总线。
其中,存储器54可以包括易失性存储器(英文:volatile memory),例如随机存取存储器(英文:random-access memory,缩写:RAM);存储器也可以包括非易失性存储器(英文:non-volatile memory),例如快闪存储器(英文:flash memory),硬盘(英文:hard disk drive,缩写:HDD)或固态硬盘(英文:solid-state drive,缩写:SSD);存储器54还可以包括上述种类的存储器的组合。
其中,处理器51可以是中央处理器(英文:central processing unit,缩写:CPU),网络处理器(英文:network processor,缩写:NP)或者CPU和NP的组合。
其中,处理器51还可以进一步包括硬件芯片。上述硬件芯片可以是专用集成电路(英文:application-specific integrated circuit,缩写:ASIC),可编程逻辑器件(英文:programmable logic device,缩写:PLD)或其组合。上述PLD可以是复杂可编程逻辑器件(英文:complex programmable logic device,缩写:CPLD),现场可编程逻辑门阵列(英文:field-programmable gate array,缩写:FPGA),通用阵列逻辑(英文:generic array logic,缩写:GAL)或其任意组合。
可选地,存储器54还用于存储程序指令。处理器51可以调用程序指令,实现如本申请图1至6实施例中所示的用于视频数据处理的资源分配方法。
本申请实施例还提供了一种非暂态计算机存储介质,所述计算机存储介质存储有计算机可执行指令,该计算机可执行指令可执行上述任意方法实施例中的用于视频数据处理的资源分配方法。其中,所述存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM)、随机存储记忆体(Random Access Memory,RAM)、快闪存储器(Flash Memory)、硬盘(Hard Disk Drive,缩写:HDD)或固态硬盘(Solid-State Drive,SSD)等;所述存储介质还可以包括上述种类的存储器的组合。
虽然结合附图描述了本申请的实施例,但是本领域技术人员可以在不脱离本申请的精神和范围的情况下做出各种修改和变型,这样的修改和变型均落入由所附权利要求所限定的范围之内。

Claims (10)

  1. 一种用于视频数据处理的资源分配方法,其特征在于,包括:
    在每个优化周期获取多个目标视频数据;其中,所述每个优化周期划分为多个连续的时间段;
    获取每个所述目标视频数据的参数;所述目标视频数据的参数包括长度以及网络数据流量;
    基于每个所述目标视频数据的参数,确定是否存储各个所述目标视频数据;
    根据每个所述优化周期内所述多个目标视频数据的存储结果,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配;其中,所述卸载决策用于表示本地对各个所述用户终端对应的所述目标视频数据进行转码运算的百分比。
  2. 根据权利要求1所述的方法,其特征在于,所述基于每个所述目标视频数据的参数,确定是否存储各个所述目标视频数据,包括:
    基于每个所述目标视频数据的长度,确定存储各个所述目标视频数据所造成的目标网络数据流量;
    利用所述目标视频数据的参数、目标网络数据流量、本地的存储容量以及所述目标网络数据流量,形成是否存储各个所述目标视频数据的第一约束条件;
    以存储所有所述目标视频数据的收益最大化为目标,且利用所述第一约束条件确定是否存储各个所述目标视频数据。
  3. 根据权利要求2所述的方法,其特征在于,所述基于每个所述目标视频数据的长度,确定存储所述目标视频数据所造成的目标网络数据流量,包括:
    提取各个所述目标视频数据的用户请求到达率;
    计算所述用户请求到达率与所述目标视频数据的长度的乘积,得到存储所述目标视频数据所造成的目标网络数据流量。
  4. 根据权利要求2所述的方法,其特征在于,所述利用所述目标视频数据的参数、目标网络数据流量、本地的存储容量以及所述目标网络数据流量,形成是否存储各个所述目标视频数据的第一约束条件,包括:
    形成存储所述目标网络数据的决策矢量,以得到第一约束子条件;
    计算所述决策矢量与对应的所述目标网络数据的长度的乘积之和,形成第二约束子条件;
    计算所述决策矢量与对应的所述目标网络流量的乘积,以形成第三约束子条件;
    计算所有所述决策矢量与对应的所述目标网络流量的乘积之和,以形成第四约束子条件。
  5. 根据权利要求4所述的方法,其特征在于,所述第一约束条件以及所述所有所述目标视频数据的收益最大化采用如下公式表示:
    Figure PCTCN2019122039-appb-100001
    Figure PCTCN2019122039-appb-100002
    Figure PCTCN2019122039-appb-100003
    Figure PCTCN2019122039-appb-100004
    Figure PCTCN2019122039-appb-100005
    式中,
    Figure PCTCN2019122039-appb-100006
    1/ι w=L w/r w
    Figure PCTCN2019122039-appb-100007
    ε>0,γ∈[-1,1];
    δ=exp{-Ω 2/2};
    其中,C11、C12、C13、C14分别表示所述第一约束子条件、所述第二约束子条件、所述第三约束子条件以及所述第四约束子条件;w∈W=(1,2,...,W)表示每个所述目标视频数据;h w∈{0,1},
    Figure PCTCN2019122039-appb-100008
    表示对目标视频数据w的存储决定;
    Figure PCTCN2019122039-appb-100009
    表示h w对应的连续变量;ψ w表示存储目标视频数据w的收益;
    Figure PCTCN2019122039-appb-100010
    表示存储目标视频数据w的资源开销;L w表示目标视频数据w的长度;Y表示本地的存储容量;1/ι w表示目标视频片段w的传输间隙保持时间;r w表示传输目标视频数据w的速率;λ w表示目标视频数据w的用户请求到达率;ε>0;O +[p w]表示分配给目标视频数据w的最大网络容量;O +[p sum]表示分配给所有目标视频数据的最大网络容量;δ表示违背所述第三约束条件或所述第四约束条件的概率最大值。
  6. 根据权利要求1-5中任一项所述的方法,其特征在于,根据每个所述优化周期内所述多个目标视频数据的存储结果,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配,包括:
    基于每个所述优化周期内所述多个目标视频数据的存储结果,确定用户请求所述目标视频数据的命中结果;其中,所述命中结果包括直接命中、转码命中以及未命中;
    获取本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率;其中,所述小蜂窝包括所述小基站以及对应的所述用户终端;
    利用所述目标视频数据的命中结果、本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率,形成确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配的第二约束条件;
    以本地收益最大化为目标,且利用所述第二约束条件,确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配;其中,所述本地收益为所述频谱资源的收益与所述计算资源的收益之和。
  7. 根据权利要求6所述的方法,其特征在于,所述利用所述目标视频数据的命中结果、本地的可用频带总带宽、小蜂窝与宏基站的回程链路带宽以及小基站向所述用户终端传输数据的频谱效率,形成确定各个用户终端在每个所述时间段内对应的卸载决策、频谱资源以及计算资源的分配的第二约束条件,包括:
    计算本地为各个所述用户终端分配的计算资源、对应的所述命中结果以及对应的所述卸载决策的乘积之和,形成第五约束子条件;
    利用本地为各个所述用户终端分配的计算资源以及所述用户终端自身的计算能力,形成第六约束子条件;
    利用本地为各个所述用户终端分配的频谱资源形成第七约束子条件;
    计算本地为各个所述用户终端分配的频谱资源、所述本地的可用频带总带宽以及所述频谱效率的乘积,以形成第八约束子条件;
    计算所有本地为各个所述用户终端分配的频谱资源、所述本地的可用频带总带宽以及所述频谱效率的乘积之和,以形成第九约束条件。
  8. 根据权利要求7所述的方法,其特征在于,所述第二约束条件以及所述本地收益最大化采用如下公式表示:
    Figure PCTCN2019122039-appb-100011
    Figure PCTCN2019122039-appb-100012
    Figure PCTCN2019122039-appb-100013
    Figure PCTCN2019122039-appb-100014
    Figure PCTCN2019122039-appb-100015
    Figure PCTCN2019122039-appb-100016
    式中,
    Figure PCTCN2019122039-appb-100017
    Figure PCTCN2019122039-appb-100018
    Figure PCTCN2019122039-appb-100019
    Figure PCTCN2019122039-appb-100020
    其中,k n表示小基站n对应的用户终端k;K n表示所有用户终端的集合;
    Figure PCTCN2019122039-appb-100021
    表示用户终端k n对应的卸载决策;
    Figure PCTCN2019122039-appb-100022
    表示本地为用户终端k n分配的计算资源;v n表示小蜂窝n收取费用的单价;F表示本地的全部计算资源;
    Figure PCTCN2019122039-appb-100023
    表示本地执行的用户终端k n的计算任务;
    Figure PCTCN2019122039-appb-100024
    表示本地为用户终端k n分配的频谱资源;θ n表示向小蜂窝n中的用户终端所收取的传输目标视频数据的单价;B表示本地的可用频带总带宽;
    Figure PCTCN2019122039-appb-100025
    表示对应于用户终端k n的频谱效率;υ n表示小蜂窝n租借频谱的单价;η n表示租借小蜂窝n与宏基站之间回程链路的单价;p n表示小基站n向用户终端k n传输时的功率密度;
    Figure PCTCN2019122039-appb-100026
    Figure PCTCN2019122039-appb-100027
    分别表示小基站n到用户终端k n和小基站m到用户终端k n的信道增益,m≠n;σ表示加性高斯白噪声的功率谱密度;
    Figure PCTCN2019122039-appb-100028
    表示用户终端k n自身的计算能力;
    Figure PCTCN2019122039-appb-100029
    表示用户终端k n对所请求的目标视频数据所需的最小传输速率;L n表示小蜂窝n与宏基站的回程链路带宽。
  9. 一种电子设备,其特征在于,包括:
    存储器和处理器,所述存储器和所述处理器之间互相通信连接,所述存储 器中存储有计算机指令,所述处理器通过执行所述计算机指令,从而执行权利要求1-8中任一项所述的用于视频数据处理的资源分配方法。
  10. 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质存储有计算机指令,所述计算机指令用于使所述计算机执行权利要求1-8中任一项所述的用于视频数据处理的资源分配方法。
PCT/CN2019/122039 2019-11-29 2019-11-29 用于视频数据处理的资源分配方法及电子设备 WO2021102929A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/122039 WO2021102929A1 (zh) 2019-11-29 2019-11-29 用于视频数据处理的资源分配方法及电子设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/122039 WO2021102929A1 (zh) 2019-11-29 2019-11-29 用于视频数据处理的资源分配方法及电子设备

Publications (1)

Publication Number Publication Date
WO2021102929A1 true WO2021102929A1 (zh) 2021-06-03

Family

ID=76129871

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/122039 WO2021102929A1 (zh) 2019-11-29 2019-11-29 用于视频数据处理的资源分配方法及电子设备

Country Status (1)

Country Link
WO (1) WO2021102929A1 (zh)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110248210A (zh) * 2019-05-29 2019-09-17 上海交通大学 视频传输优化方法
CN110312231A (zh) * 2019-06-28 2019-10-08 重庆邮电大学 一种车联网中基于移动边缘计算的内容缓存决策和资源分配联合优化方法

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110248210A (zh) * 2019-05-29 2019-09-17 上海交通大学 视频传输优化方法
CN110312231A (zh) * 2019-06-28 2019-10-08 重庆邮电大学 一种车联网中基于移动边缘计算的内容缓存决策和资源分配联合优化方法

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
SHAO XUN; HASEGAWA GO; KAMIYAMA NORIAKI; LIU ZHI; MASUI HIROSHI; JI YUSHENG: "Joint Optimization of Computing Resources and Data Allocation for Mobile Edge Computing (MEC): An Online Approach.", 2019 28TH INTERNATIONAL CONFERENCE ON COMPUTER COMMUNICATION AND NETWORKS (ICCCN), 26 September 2019 (2019-09-26), pages 1 - 9, XP033620588, DOI: 10.1109/ICCCN.2019.8847046 *
WANG CHENMENG; LIANG CHENGCHAO; YU F RICHARD; CHEN QIANBIN; TANG LUN: "Computation Offloading and Resource Allocation in Wireless Cellular Networks With Mobile Edge Computing", IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, vol. 16, no. 8, 16 May 2017 (2017-05-16), pages 4924 - 4938, XP011658632, ISSN: 1536-1276, DOI: 10.1109/TWC.2017.2703901 *
YANTING WANG, MIN SHENG ,XIJUN WANG ,LIANG WANG, JIANDONG LI: "Mobile-Edge Computing: Partial Computation Offloading Using Dynamic Voltage Scaling.", IEEE TRANSACTIONS ON COMMUNICATIONS, vol. 64, no. 10, 1 October 2016 (2016-10-01), pages 4268 - 4282, XP011625801, ISSN: 0090-6778, DOI: 10.1109/TCOMM.2016.2599530 *

Similar Documents

Publication Publication Date Title
US20180227844A1 (en) Network edge based access network discovery and selection
Manogaran et al. Artificial intelligence and blockchain-assisted offloading approach for data availability maximization in edge nodes
CN111431941A (zh) 一种基于移动边缘计算的实时视频码率自适应方法
US11974147B2 (en) Dynamic cloudlet fog node deployment architecture
Gudkova et al. Service failure and interruption probability analysis for licensed shared access regulatory framework
EP3854063B1 (en) Methods and systems for optimizing processing of application requests
CN111107566B (zh) 一种电力物联网场景中基于协同内容缓存的卸载方法
CN106791887B (zh) 无线网络中视频的分布式缓存和传输优化方法
CN113660303B (zh) 一种端边网云协同的任务卸载方法及系统
Jian et al. Joint computation offloading and resource allocation in C-RAN with MEC based on spectrum efficiency
US9553623B2 (en) Wireless communication device
US10901388B2 (en) Method and system for creating energy demand model
Mazza et al. A partial offloading technique for wireless mobile cloud computing in smart cities
CN111901400A (zh) 一种配备缓存辅助装置的边缘计算网络任务卸载方法
US20200296442A1 (en) Cap-based Client-Network Interaction for Improved Streaming Experience
Akter et al. Q-MAC: QoS and mobility aware optimal resource allocation for dynamic application offloading in mobile cloud computing
US20180152849A1 (en) Enable access point availability prediction
CN111107639B (zh) 用于视频数据处理的资源分配方法及电子设备
El Khatib et al. Optimal proactive resource allocation at the extreme edge
Behravesh et al. ML-driven DASH content pre-fetching in MEC-enabled mobile networks
US11356349B2 (en) Adaptive resource allocation to facilitate device mobility and management of uncertainty in communications
Choi et al. Delay-sensitive and power-efficient quality control of dynamic video streaming using adaptive super-resolution
WO2021102929A1 (zh) 用于视频数据处理的资源分配方法及电子设备
US20230231813A1 (en) Enhanced network with data flow differentiation
Sun et al. Computation offloading with virtual resources management in mobile edge networks

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: 19953744

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 29/09/2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19953744

Country of ref document: EP

Kind code of ref document: A1