Caching and resource scheduling method for adaptive bitrate video in edge networks

By using nonlinear integer planning in edge networks to optimize video cache and resource allocation, combined with user preferences and popularity, the problem of video traffic occupying network traffic in edge computing networks is solved, and the user experience quality and resource utilization efficiency are improved.

CN115720237BInactive Publication Date: 2024-12-06SOUTH CHINA UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211421390.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-14
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art fails to make full use of video caching and resource scheduling in edge computing networks, resulting in video traffic accounting for 82% of network traffic, affecting user experience, and failing to consider user behavior and the preferences of different parts of the video, affecting cache efficiency.

Method used

A method of adaptive bit rate video caching and resource scheduling in edge networks is proposed. Video cache and resource allocation are optimized through nonlinear integer planning problems, combined with user preferences and popularity, and decompose into cache stages and delivery stages, using greedy heuristics and scheduling. 入策略,确保用户体验质量和资源约束。

Benefits of technology

It improves the hit rate and user experience quality of video cache, reduces backhaul traffic, optimizes resource utilization, and improves the network performance of 5G Internet of Things.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115720237B_ABST
    Figure CN115720237B_ABST
Patent Text Reader

Abstract

The present invention discloses a caching and resource scheduling method for edge network adaptive bit rate video, which optimizes the user's QoE under resource constraints. The method takes into account various fine-grained influencing factors in the video service process, ensures the credibility of the solution, and expresses the video caching and video processing problems as a nonlinear integer programming problem. The original problem is decomposed into two sub-problems: the cache placement problem in the caching stage and the resource scheduling problem in the delivery stage. In the caching stage, the nonlinear integer programming problem is simplified, an optimization solver is used, and then a rounding strategy is used to obtain the final result, and segments with appropriate bit rate levels are selected for caching to ensure the feasibility of the solution; in the delivery stage, user request scheduling is performed according to the minimum delay, and bandwidth and power are allocated to users by solving the original problem and data transmission is performed, thereby ensuring the operability of the solution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of edge computing and video services, and in particular to a caching and resource scheduling method for edge network adaptive bit rate video. Background Art

[0002] In today's era, with the development of new mobile applications, the technologies of mobile devices such as smartphones are also developing rapidly. However, when running programs with large computational workloads on mobile devices where storage performance and energy consumption are greatly limited, their resource consumption and computational delay will seriously affect the user experience. One common solution is to offload such applications that require a lot of computing resources to the cloud. However, this will also cause new problems, that is, it will cause greater delays, including delays in computing and communication, and will generate large backhaul traffic on the core network, causing network congestion. How to better solve these problems has become the key to realizing the 5G Internet of Everything. As one of the solutions to the above problems, Multi-access Edge Computing (MEC) has been proposed. Multi-access edge computing is to deploy resources and user services at the edge of the network, such as performing part of the computing work, to reduce the delay of user services and thus improve the quality of user experience.

[0003] Nowadays, the development of mobile applications and the rise of various video software, such as short video applications, Douyin, Bilibili, video communications, etc., have made video traffic occupy a large part of network traffic. According to Cisco data, by 2022, video traffic will account for 82% of network traffic. While these video applications provide sufficient video services, they also increase the traffic burden of the core network, increase service latency, and inevitably affect user experience.

[0004] To address this critical issue, multi-access edge computing (MEC) has become a viable technology that can use edge servers near users for video caching and transmission. However, in recent studies, video caching in edge networks mainly considers the popularity of videos and user preferences for videos, and few studies consider user behavior and user preferences for different parts of the video, which do have an essential impact on caching efficiency. At the same time, due to the limited computing and wireless resources of MEC networks, it is still challenging and urgent to combine MEC with DASH technology and design efficient video caching and resource scheduling schemes to fully utilize MEC networks. Therefore, the present invention attempts to combine these fine-grained factors and propose a new adaptive bitrate video caching and resource scheduling scheme. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and proposes a caching and resource scheduling method for edge network adaptive bit rate video, which optimizes the user's QoE under resource constraints. The scheme takes into account various fine-grained influencing factors in the video service process, ensures the credibility of the scheme, and expresses the video caching and video processing problems as a nonlinear integer programming problem. The original problem is decomposed into two sub-problems: the cache placement problem in the caching stage and the resource scheduling problem in the delivery stage. In the caching stage, the nonlinear integer programming problem is simplified, the optimization solver is used, and then the rounding strategy is used to obtain the final result, and the segments with the appropriate bit rate level are selected for caching to ensure the feasibility of the scheme; in the delivery stage, user request scheduling is performed according to the minimum delay, and bandwidth and power are allocated to users by solving the original problem and data transmission is performed, which ensures the operability of the scheme.

[0006] To achieve the above object, the technical solution provided by the present invention is: a method for caching and resource scheduling of edge network adaptive bit rate video, comprising the following steps:

[0007] 1) Obtaining information: In the offline stage, the current popularity of the video and the user's preference for the video segment are obtained and analyzed based on the video playback volume and user playback record information;

[0008] 2) Video service modeling: Based on the current popularity of the video and the user's preference for video segments, the video service in the edge network is modeled. A mathematical model is established with the user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints. The mathematical model is expressed as a nonlinear integer programming problem.

[0009] 3) Video caching: The nonlinear integer programming problem is simplified by slack variables and solved using an optimization solver. The final result is obtained using a rounding strategy. Based on the result, the edge server caches video segments of various bit rates.

[0010] 4) Service delivery: At each moment, user requests are dispatched to the serviceable base station according to the rule of minimum latency. The edge server allocates bandwidth and power to the user who initiates the request by solving a nonlinear integer programming problem. Finally, based on the cache results and resource allocation results, the edge server serves the user request, that is, transmits video data.

[0011] Furthermore, in step 1), the current popularity of the video and the user's preference for the video segment are obtained and analyzed, and the details are as follows:

[0012] Popularity: The edge server obtains the ratio of the current video’s cumulative playback volume to the total playback volume as the popularity basis Pr v, represents the popularity of the vth video, where v∈{1,2,...,V}, and V represents the number of videos. It is assumed that in most cases, the growth rate of popular videos is higher than that of unpopular videos. In some cases, including a sudden jump in the number of views of a video, or a video has just been released, the current number of views does not match the growth; but this is only a short-term phenomenon. After a period of time, the growth of the video will stabilize and match the number of views;

[0013] User preference: User preference can be obtained from the user's historical records. The ratio of the number of times a single video is played to the total number of times played within a period of time is calculated as the preference of user n for the vth video. and video segment v k Preference where v k represents the kth segment of the vth video, n∈{1,2,...,N}, k∈{1,2,...,K}, N represents the number of users, and K represents the number of video segments.

[0014] Further, the step 2) comprises:

[0015] Confirm model variables: Use To indicate whether the kth segment of the vth video is cached on the mth edge server at the lth bit rate, If it is 0, it means no, if it is 1, it means yes, l∈{1,2,...,L}, m∈{1,2,...,M}, L represents the number of bit rates, M represents the number of edge servers; the combined value of popularity and preference is used as the request probability represents the probability that user n requests the kth segment of the vth video and the lth bitrate level; using the identifier Indicates whether to perform transcoding. This means that if and only if the mth edge server does not cache the video segment v of the lth bit rate k However, when there is a cache with a higher bit rate than the first bit rate, transcoding is performed. Indicates whether the kth segment of the vth video is cached on the mth edge server at the ith bit rate, i = l + 1;

[0016] User experience quality modeling: The user experience quality of each user n is determined by the average video quality AVQ n , Video Bitrate Switching VS n and rebuffer time RT n Calculation: By calculating the total quality of the video segments that the user has watched Divided by the number of video segments watched K n Get the average video quality AVQ n , where BRl Indicates the first bit rate; video bit rate switching VS n The calculation is performed by superimposing the absolute value of the bit rate difference of the historical requested video segments, that is, That is, the absolute value of the difference in bit rate between two adjacent video segments, BR l' represents the l'th bit rate, l'∈{1,2,...,L}; and the rebuffering time RT n The amount of data currently cached by the user device's buffer is BF n (t) determines, where BF n (t) is measured in seconds. When the user buffer is exhausted, BF n (t) = 0, the time between the user pausing the video and the next time the video is played is considered as rebuffering, and the accumulated time is used as the rebuffering time during the user's viewing process; the three are linearly combined to obtain the user experience quality Q, that is, Q = AVQ n -αVS n -βRT n , where α>0, β>0, representing weight parameters;

[0017] Constraint modeling: The bandwidth constraint is the bandwidth B allocated to each user n n The sum cannot exceed the total bandwidth B, ∑ n∈ N B n ≤B; similarly, the power of a single user n is P n The sum cannot exceed the power threshold P, ∑ n∈N P n ≤P; the amount of data cached by the edge server cannot exceed the space size CH m , During the service, the computing resources used for transcoding cannot exceed the PC limit. m , Where S l The data volume of the video with the lth bit rate, i.e., the video data size;

[0018] Energy consumption modeling: Energy consumption E m The calculation includes the energy consumption of the entire service process, namely, cache energy consumption Transcoding energy consumption and transmission energy consumption The energy consumption during the caching process is called cache energy consumption. and the amount of data currently cached The energy consumption of the transcoding process is directly proportional to Similarly, the video data volume S lRegarding transcoding, as the computing power of computers increases, high-bitrate videos can be transcoded into low-bitrate videos through transcoding operations, and then provided to users. The essence is to decode first and then re-encode. According to the hardware configuration of the device, the transcoding energy consumption of the video segment is calculated based on the CPU performance. where ω tc represents the energy consumption factor, S l Indicates the video data size, cp indicates the transcoding intensity, and f cpu Indicates CPU frequency; transmission energy consumption is transmission energy consumption It is related to the amount of video data and network conditions. Where R n Represents the user transmission rate, P n represents the power allocated to user n.

[0019] Further, in step 3), the nonlinear integer programming problem is simplified and an optimization solver is used, and then a rounding strategy is used to obtain the final result. According to the solved cache placement result, the result shows whether a certain video segment should be cached or at what bit rate it should be cached. Then, according to the result, a video segment with a specific bit rate is obtained from the data center and cached on the edge server. The details are as follows:

[0020] A feasible solution is obtained: First, based on the existing popularity and user preference information, a greedy algorithm is used to select video segments with higher preferences from all video segments in turn, and their identifiers are set to 1, that is, The size of the data that has been placed is recorded until the remaining edge servers have insufficient space to place the next video segment. At this time, a feasible result array XL is obtained, which is a feasible solution.

[0021] Get the current optimal solution: Use the feasible solution XL as the result of cache placement to calculate the bandwidth B of each user n at the current moment n and power P n At the same time, based on the feasible solution XL obtained, the identifiers in the original problem are The value range of is relaxed from {0,1} to [0,1], and then the SLSQP optimizer is used to calculate each identifier by solving the original problem The value of makes the user experience quality of the entire cache result optimal and meets the constraints on bandwidth, power, cache space size, computing resource size and energy consumption;

[0022] Obtaining the global optimal solution: For each edge server, the feasible solution XL pre-calculates the bandwidth and power allocation based on the rule of maximizing the user experience quality, and in turn substitutes the allocation result into the original formula to calculate the video cache placement again. Then, the two steps of calculating cache placement, bandwidth and power allocation are iterated multiple times to obtain the optimal solution XU for video cache placement;

[0023] The rounding process to obtain the optimal feasible solution: Based on the optimal solution XU, the identifier in the result The value of is restored to 0 or 1, the process is as follows: by controlling the parameter σ Perform rounding, specifically, for each edge server, for each if set up otherwise, Denote the rounded cache variable set as X and the corresponding user experience quality value as Q X ; If X exceeds the constraints of cache space and computing resources, increase σ to σ+λ|(Q XU -Q XL ) / Q ZU |, where λ is the step size. This is because when σ is increased, the number of variables with a value of 1 decreases, the cache space required for video clips decreases, and the constraint condition can be satisfied. On the contrary, if σ satisfies the constraint, the value of σ is reduced to cache more video clips. If Q X ≥Q XL , X is a feasible result, indicating that X is a better solution than XL. The lower bound of the feasible solution is updated by setting XL←X. Repeat the above rounding process until Q x The improvement in value is less than the threshold ε; note that Q XL The value of increases with the number of iterations, because λ is a constant, Q XU is fixed, when Q XL As the value of increases, the change in σ gradually decreases, so the rounding process can converge.

[0024] Further, in step 4), at each moment, the user request is scheduled to the serviceable base station according to the rule of minimum delay, and the edge server allocates bandwidth and power to the user who initiates the request, so that the average experience quality is the highest and does not exceed the bandwidth and power constraints. According to the cache result and resource allocation result, the edge server serves the user request, that is, transmits the video data. The details are as follows:

[0025] User request scheduling: In dense cellular networks, users may be covered and served by multiple SBSs, so user requests are redirected to appropriate SBSs to pursue high QoE. Specifically, according to the cache results, given a user request and the set of SBSs covering the request, with fixed bandwidth and power, the video transcoding tc and transmission delay tr of each SBS are calculated, and the SBS with the lowest delay is selected to serve the request, where Where cp represents the transcoding intensity, f cpu Indicates CPU frequency, and tr = Data n / Rn , where Data n Indicates the amount of data transmitted, R n Indicates the user transmission rate;

[0026] Bandwidth and power allocation: Based on the request scheduling results, the SLSQP optimizer is used to solve the nonlinear integer programming problem with user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints, and the bandwidth and power currently pre-allocated to the user are calculated to optimize the overall user experience quality.

[0027] Service delivery: Serving user requests based on cache results and resource allocation results, that is, transmitting data. There are four situations in which users initiate requests to edge servers:

[0028] ① The requested video segment with a specific bit rate is cached in the edge server. The video segment can be directly transmitted. Otherwise, the edge server checks whether there is a higher bit rate than the requested bit rate;

[0029] ② If a high bit rate is cached, the edge server can perform the transcoding operation and then send the transcoded version to the user;

[0030] ③ The requested segments are not cached in the edge server, but in the neighbor server. The edge server stores the information of neighbor servers for collaboration; when the information is updated, it will be exchanged through broadcasting; therefore, the edge server can get the video from the neighbor and decide whether to transcode it;

[0031] ④ If all the cooperating edge servers do not cache the video segment requested by the user, the video with a specific bit rate is directly obtained from the cloud.

[0032] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0033] 1. The present invention considers user preferences and edge caching at a finer-grained video segment level in dense cellular networks.

[0034] 2. The present invention models the placement and processing of multi-bitrate video clips as a nonlinear integer programming problem, taking into account the influence of many factors to ensure the credibility of the solution.

[0035] 3. The present invention adopts a divide-and-conquer strategy to decompose the original problem into the cache placement problem in the cache stage and the online request scheduling and resource allocation problem in the delivery stage. Then, an efficient algorithm based on greedy heuristics, relaxation and rounding techniques is designed to solve these two problems, ensuring the feasibility of the solution.

[0036] 4. The present invention uses real data sets to conduct experiments to evaluate the performance comparison between the proposed strategy and the traditional caching strategy. The results show that the proposed strategy achieves better results in terms of hit rate, return traffic and user experience quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a model diagram of the present invention.

[0038] Figure 2 It is a logical flow diagram of the present invention.

[0039] Figure 3 This is an application architecture diagram of the present invention. DETAILED DESCRIPTION

[0040] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0041] This embodiment discloses a method for caching and resource scheduling of adaptive bitrate video in edge networks. Based on video streaming technology, it considers user behavior and user preferences for different video segments, and establishes a model for fine-grained caching and resource scheduling problems in edge networks. We decouple the stated problem into a caching problem and a resource scheduling problem. For the former, we design two effective caching algorithms, including greedy heuristics, relaxation, and rounding nonlinear integer programming techniques. For the latter, we propose an online algorithm to schedule user requests and power allocation. The present invention uses this scheme to improve the user's quality of experience during video services.

[0042] like Figure 1 As shown in Figure 2, we consider the classic framework of video services in dense cellular networks, including the cloud or remote data center, base stations SBS, and users. M represents the set of SBSs and N represents the set of users. Each SBS is attached with an edge server with a storage capacity of CH m , computing resources are PC m SBS provides services to users within its coverage area. According to the actual situation, the coverage area of ​​SBS has a certain overlap to prevent blind spots in the coverage area. At the same time, it is assumed that the cloud stores video clips of all bit rate levels.

[0043] The overall solution process is as follows Figure 2 As shown in the figure, first, in the offline phase, the current popularity of the video and the user's preference for the video segment are obtained. Based on the prior knowledge, including video popularity and user preference, combined with the video service modeling and the proposed caching scheme, the multi-bitrate video segments are cached on multiple servers in the edge network.

[0044] Secondly, in the online service phase, at each moment, user requests are dispatched to the serviceable base station according to the rule of minimum latency. The edge server allocates bandwidth and power to the user who initiates the request based on resources such as bandwidth, so that the average experience quality is the highest and does not exceed the bandwidth and power constraints. Based on the cache results and resource allocation results, the edge server serves the user request, that is, transmits the video data.

[0045] When entering the offline phase again, the cache is updated, that is, the cache placement results are updated according to the changes in the current video popularity and user preferences, including the combination of obtaining prior knowledge and video caching.

[0046] Repeat the above process to provide video services to users.

[0047] like Figure 1-Figure 3 As shown, the specific steps of the solution for the fine-grained cache and resource scheduling problem in the edge network in this embodiment are as follows:

[0048] 1) Obtaining information: In the offline stage, the current popularity of the video and the user's preference for the video segment are obtained and analyzed based on information such as the video playback volume and user playback history.

[0049] Popularity: The edge server obtains the ratio of the current video’s cumulative playback volume to the total playback volume as the popularity basis Pr v , represents the popularity of the vth video, where v∈{1,2,...,V}, and V represents the number of videos, which means that in most cases, the growth rate of videos with higher popularity is higher than that of videos with lower popularity. In some cases, such as when the number of views of a video suddenly jumps, or a video has just been released, the current number of views does not match the growth. But this is only a short-term phenomenon. After a period of time, the growth of the video will stabilize and match the number of views.

[0050] User preference: User preference can be obtained from the user's historical records. The ratio of the number of times a single video is played to the total number of times played within a period of time is calculated as the preference of user n for the vth video. and video segment v k Preference where v k represents the kth segment of the vth video, n∈{1,2,...,N}, k∈{1,2,...,K}, N represents the number of users, and K represents the number of video segments.

[0051] 2) Video service modeling: Based on the current popularity of the video and the user's preference for the video segment, the video service in the edge network is modeled. A mathematical model is established with the user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints. The mathematical model is expressed as a nonlinear integer programming problem. The details are as follows:

[0052] Confirm model variables: Use To indicate whether the kth segment of the vth video is cached on the mth edge server at the lth bit rate, If it is 0, it means no, if it is 1, it means yes, l∈{1,2,...,L}, m∈{1,2,...,M}, L represents the number of bit rates, M represents the number of edge servers; the combined value of popularity and preference is used as the request probability represents the probability that user n requests the kth segment of the vth video and the lth bitrate level; using the identifier Indicates whether to perform transcoding. This means that if and only if the mth edge server does not cache the video segment v of the lth bit rate k However, when there is a cache with a higher bit rate than the first bit rate, transcoding is performed. Indicates whether the kth segment of the vth video is cached on the mth edge server at the ith bit rate, i = l + 1;

[0053] User experience quality modeling: The user experience quality of each user n is determined by the average video quality AVQ n , Video Bitrate Switching VS n and rebuffer time RT n Calculation: By calculating the total quality of the video segments that the user has watched Divided by the number of video segments watched K n Get the average video quality AVQ n , where BR l Indicates the first bit rate; video bit rate switching VS n The calculation is performed by superimposing the absolute value of the bit rate difference of the historical requested video segments, that is, That is, the absolute value of the difference in bit rate between two adjacent video segments, BR l' represents the l'th bit rate, l'∈{1,2,...,L}; and the rebuffering time RT n The amount of data currently cached by the user device's buffer is BF n (t) determines, where BF n (t) is measured in seconds. When the user buffer is exhausted, BF n(t) = 0, the time between the user pausing the video and the next time the video is played is considered as rebuffering, and the accumulated time is used as the rebuffering time during the user's viewing process; the three are linearly combined to obtain the user experience quality Q, that is, Q = AVQ n -αVS n -βRT n , where α>0, β>0, representing weight parameters;

[0054] Constraint modeling: The bandwidth constraint is the bandwidth B allocated to each user n n The sum cannot exceed the total bandwidth B, ∑ n∈ N B n ≤B; similarly, the power of a single user n is P n The sum cannot exceed the power threshold P, ∑ n∈N P n ≤P; the amount of data cached by the edge server cannot exceed the space size CH m , During the service, the computing resources used for transcoding cannot exceed the PC limit. m , Where S l The data volume of the video with the lth bit rate, i.e., the video data size;

[0055] Energy consumption modeling: Energy consumption E m The calculation includes the energy consumption of the entire service process, namely, cache energy consumption Transcoding energy consumption and transmission energy consumption The energy consumption during the caching process is called cache energy consumption. and the amount of data currently cached The energy consumption of the transcoding process is directly proportional to Similarly, the video data volume S l Regarding transcoding, as the computing power of computers increases, high-bitrate videos can be transcoded into low-bitrate videos through transcoding operations, and then provided to users. The essence is to decode first and then re-encode. According to the hardware configuration of the device, the transcoding energy consumption of the video segment is calculated based on the CPU performance. where ω tc represents the energy consumption factor, S l Indicates the video data size, cp indicates the transcoding intensity, and f cpu Indicates CPU frequency; transmission energy consumption is transmission energy consumption It is related to the amount of video data and network conditions. Where R n Represents the user transmission rate, P n represents the power allocated to user n.

[0056] 3) Video caching: Based on prior knowledge, including video popularity and user preferences, combined with the proposed caching scheme, multi-bitrate video segments are cached on multiple servers in the edge network. The edge server calculates the optimal result of cache placement based on the scheme used, which shows whether a certain video segment should be cached or at what bit rate, and then obtains the video segment of a specific bit rate from the data center and caches it on the edge server based on the result. The scheme includes the following steps:

[0057] A feasible solution is obtained: First, based on the existing popularity and user preference information, a greedy algorithm is used to select video segments with higher preferences from all video segments in turn, and their identifiers are set to 1, that is, The size of the data that has been placed is recorded until the remaining edge server space is insufficient to place the next video segment. At this time, a feasible result array XL is obtained, which is a feasible solution.

[0058] Get the current optimal solution: Use the feasible solution XL as the result of cache placement to calculate the bandwidth B of each user n at the current moment n and power P n At the same time, based on the feasible solution XL obtained, the identifiers in the original problem are The value range of is relaxed from {0,1} to [0,1], and then the SLSQP optimizer is used to calculate each identifier by solving the original problem The value of makes the user experience quality of the entire cache result optimal and meets the constraints on bandwidth, power, cache space size, computing resource size and energy consumption.

[0059] Obtaining the global optimal solution: For each edge server, the solution pre-calculates the bandwidth and power allocation based on the rule of maximizing the user experience quality by the feasible solution XL, and in turn substitutes the allocation result into the original formula to calculate the video cache placement again. Then, the two steps of calculating cache placement, bandwidth and power allocation are iterated multiple times to obtain the optimal solution XU for video cache placement.

[0060] The rounding process to obtain the optimal feasible solution: Based on the optimal solution XU, the identifier in the result The value of is restored to 0 or 1, the process is as follows: by controlling the parameter σ Specifically, for each edge server, for each if set up otherwise, Denote the rounded cache variable set as X and the corresponding user experience quality value as Q X If X exceeds the constraints of cache space and computing resources, increase σ to σ+λ|(QXU -Q XL ) / Q ZU | (where λ is the step size). This is because when σ is increased, the number of variables with a value of 1 decreases, the cache space required for video clips decreases, and the constraint condition can be satisfied. On the contrary, if σ satisfies the constraint, the value of σ is reduced to cache more video clips. In this case, if Q X ≥Q XL , X is a feasible result, indicating that X is a better solution than XL. The lower bound of the feasible solution is updated by setting XL←X. Repeat the above rounding process until Q X The improvement in value is less than the threshold ε. Note that Q XL The value of increases with the number of iterations. Since λ is a constant, Q XU is fixed, when Q XL As the value of increases, the change amplitude of σ gradually decreases, so the rounding process can converge.

[0061] 4) In the online service phase, at each moment, user requests are dispatched to the serviceable base station according to the rule of minimum delay. The edge server allocates bandwidth and power to the user who initiates the request based on resources such as bandwidth, so that the average experience quality is the highest and does not exceed the bandwidth and power constraints. Based on the cache results and resource allocation results, the edge server serves the user request, that is, transmits the video data. The details are as follows:

[0062] User request scheduling: In dense cellular networks, users may be covered and served by multiple SBSs, so user requests are redirected to appropriate SBSs to pursue high QoE. Specifically, according to the cache results, given a user request and the set of SBSs covering the request, with fixed bandwidth and power, the video transcoding tc and transmission delay tr of each SBS are calculated, and the SBS with the lowest delay is selected to serve the request, where Where cp represents the transcoding intensity, f cpu Indicates CPU frequency, and tr = Data n / R n , where Data n Indicates the amount of data transmitted, R n Indicates the user transmission rate.

[0063] Bandwidth and power allocation: Based on the request scheduling results, without exceeding the constraints, the SLSQP optimizer is used to solve the nonlinear integer programming problem with user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints. The bandwidth and power currently pre-allocated to the user are calculated to optimize the overall user experience quality.

[0064] Service delivery: Serving user requests based on cache results and resource allocation results, that is, transmitting data. Under this system structure, there are four situations in which users initiate requests to edge servers:

[0065] ① The requested video segment with a specific bit rate is cached in the edge server, and the video segment can be directly transmitted. Otherwise, the edge server checks whether there is a higher bit rate than the requested bit rate.

[0066] ② If a higher bit rate is cached, the edge server can perform the transcoding operation and then send the transcoded version to the user;

[0067] ③ The requested segment is not cached in the edge server, but in the neighbor server. The edge server stores the information of the neighbor server for collaboration. When the information is updated, it will be exchanged through broadcast. Therefore, the edge server can get the video from the neighbor and decide whether to transcode;

[0068] ④ If all the cooperating edge servers do not cache the video segment requested by the user, the video with a specific bit rate is directly obtained from the cloud.

[0069] Cache update: Update the cache placement results according to the changes in the current video popularity and user preferences, that is, obtain the combination of prior knowledge and video caching.

[0070] Cloud / data center: Stores video segments of all bit rate levels. When the edge server updates the cache or does not cache the video segment requested by the user, it transmits the video segment data of a specific bit rate level to the base station.

[0071] Edge node / edge server: calculates pre-caching results based on information such as video segment popularity and user preferences, obtains and caches data from the data center based on the results; receives user requests, calculates bandwidth and power allocation for the current user; transcodes the video and delivers it to the user when the benefit of transcoding is greater than obtaining data from a neighboring server or data center; exchanges information with neighboring servers, and obtains video data from neighboring servers instead of data centers in specific scenarios.

[0072] User: Provides information such as preferences to the edge server; initiates a video segment request with a specific bit rate level to the edge server, and plays and updates the buffer status after receiving the data.

[0073] In summary, the present invention proposes a new adaptive bitrate video caching and resource scheduling scheme, which takes into account the fine-grained factors in cellular network video services. The cloud / data center stores video segments of all bitrate levels, and when the edge server performs cache updates or does not cache the video segment requested by the user, the video segment data of a specific bitrate level is transmitted to the base station; the user provides information such as preferences to the edge server, initiates a video segment request of a specific bitrate level to the edge server at a certain time, and plays it after receiving the data; the edge server calculates the pre-caching result based on information such as the popularity of the video segment and user preferences, obtains data from the data center and caches it based on the result, receives the user request, calculates the bandwidth and power allocation for the current user, performs transcoding operations when necessary, and delivers the data to the user, or exchanges information with the neighbor server, and obtains video data from the neighbor server instead of the data center in a specific scenario.

[0074] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for caching and resource scheduling of adaptive bit rate video in edge networks, characterized in that: The following steps are involved: 1) Obtaining information: In the offline stage, the current popularity of the video and the user's preference for the video segment are obtained and analyzed based on the video playback volume and user playback record information; 2) Video service modeling: Based on the current popularity of the video and the user's preference for video segments, the video service under the edge network is modeled. A mathematical model is established with the user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints. The mathematical model is expressed as a nonlinear integer programming problem. 3) Video caching: The nonlinear integer programming problem is simplified by slack variables and solved using an optimization solver. The final result is obtained using a rounding strategy. Based on the result, the edge server caches video segments of various bit rates. 4) Service delivery: At each moment, user requests are dispatched to the serviceable base station according to the rule of minimum latency. The edge server allocates bandwidth and power to the user who initiates the request by solving a nonlinear integer programming problem. Finally, based on the cache results and resource allocation results, the edge server serves the user request, that is, transmits video data.

2. The method for caching and resource scheduling of edge network adaptive bit rate video according to claim 1, characterized in that: In step 1), the current popularity of the video and the user's preference for the video segment are obtained and analyzed, and the details are as follows: Popularity: The edge server obtains the ratio of the current video’s cumulative playback volume to the total playback volume as the popularity basis Pr v , represents the popularity of the vth video, where v∈{1,2,...,V}, and V represents the number of videos. It is assumed that in most cases, the growth rate of popular videos is higher than that of unpopular videos. In some cases, including a sudden jump in the number of views of a video or a video that has just been released, the current number of views does not match the growth. However, this is only a short-term phenomenon. After a period of time, the growth of the video will stabilize and match the number of views. User preference: User preference can be obtained from the user's historical records. The ratio of the number of times a single video is played to the total number of times played within a period of time is calculated as the preference of user n for the vth video. and video segment v k Preference where v k represents the kth segment of the vth video, n∈{1,2,...,N}, k∈{1,2,...,K}, N represents the number of users, and K represents the number of video segments.

3. The method for caching and resource scheduling of edge network adaptive bit rate video according to claim 2, characterized in that: The step 2) comprises: Confirm model variables: Use To indicate whether the kth segment of the vth video is cached on the mth edge server at the lth bit rate, If it is 0, it means no, if it is 1, it means yes, l∈{1,2,...,L}, m∈{1,2,...,M}, L represents the number of bit rates, M represents the number of edge servers; the combined value of popularity and preference is used as the request probability represents the probability that user n requests the kth segment of the vth video and the lth bitrate level; using the identifier Indicates whether to perform transcoding. This means that if and only if the mth edge server does not cache the video segment v of the lth bit rate k However, when there is a cache with a higher bit rate than the first bit rate, transcoding is performed. Indicates whether the kth segment of the vth video is cached on the mth edge server at the ith bit rate, i = l + 1; User experience quality modeling: The user experience quality of each user n is determined by the average video quality AVQ n , Video Bitrate Switching VS n and rebuffer time RT n Calculation: By calculating the total quality of the video segments that the user has watched Divided by the number of video segments watched K n Get the average video quality AVQ n , where BR l Indicates the first bit rate; video bit rate switching VS n The calculation is performed by superimposing the absolute value of the bit rate difference of the historical requested video segments, that is, That is, the absolute value of the difference in bit rate between two adjacent video segments, BR l' represents the l'th bit rate, l'∈{1,2,...,L}; and the rebuffering time RT n The amount of data currently cached by the user device's buffer is BF n (t) determines, where BF n (t) is measured in seconds. When the user buffer is exhausted, BF n (t) = 0, the time between the user pausing the video and the next time the video is played is considered as rebuffering, and the accumulated time is used as the rebuffering time during the user's viewing process; the three are linearly combined to obtain the user experience quality Q, that is, Q = AVQ n -αVS n -βRT n , where α>0, β>0, representing weight parameters; Constraint modeling: The bandwidth constraint is the bandwidth B allocated to each user n n The sum cannot exceed the total bandwidth B, ∑ n∈N B n ≤B; similarly, the power of a single user n is P n The sum cannot exceed the power threshold P, ∑ n∈N P n ≤P; the amount of data cached by the edge server cannot exceed the space size CH m , During the service, the computing resources used for transcoding cannot exceed the PC limit. m , Where S l The data volume of the video with the lth bit rate, i.e., the video data size; Energy consumption modeling: Energy consumption E m The calculation includes the energy consumption of the entire service process, namely, cache energy consumption Transcoding energy consumption and transmission energy consumption The energy consumption during the caching process is called cache energy consumption. and the amount of data currently cached The energy consumption of the transcoding process is directly proportional to Similarly, the video data volume S l Regarding transcoding, as the computing power of computers increases, high-bitrate videos can be transcoded into low-bitrate videos through transcoding operations, and then provided to users. The essence is to decode first and then re-encode. According to the hardware configuration of the device, the transcoding energy consumption of the video segment is calculated based on the CPU performance. where ω tc represents the energy consumption factor, S l Indicates the video data size, cp indicates the transcoding intensity, and f cpu Indicates CPU frequency; transmission energy consumption is transmission energy consumption It is related to the amount of video data and network conditions. Where R n Represents the user transmission rate, P n represents the power allocated to user n.

4. The method for caching and resource scheduling of edge network adaptive bit rate video according to claim 3, characterized in that: In step 3), the nonlinear integer programming problem is simplified and an optimization solver is used. Then, a rounding strategy is used to obtain the final result. According to the cache placement result obtained, the result shows whether a certain video segment should be cached or at what bit rate. Then, according to the result, a video segment with a specific bit rate is obtained from the data center and cached on the edge server. The details are as follows: A feasible solution is obtained: First, based on the existing popularity and user preference information, a greedy algorithm is used to select video segments with higher preferences from all video segments in turn, and their identifiers are set to 1, that is, The size of the data that has been placed is recorded until the remaining edge servers have insufficient space to place the next video segment. At this time, a feasible result array XL is obtained, which is a feasible solution. Get the current optimal solution: Use the feasible solution XL as the result of cache placement to calculate the bandwidth B of each user n at the current moment n and power P n At the same time, based on the feasible solution XL obtained, the identifiers in the original problem are The value range of is relaxed from {0,1} to [0,1], and then the SLSQP optimizer is used to calculate each identifier by solving the original problem The value of makes the user experience quality of the entire cache result optimal and meets the constraints on bandwidth, power, cache space size, computing resource size and energy consumption; Obtaining the global optimal solution: For each edge server, the feasible solution XL pre-calculates the bandwidth and power allocation based on the rule of maximizing the user experience quality, and in turn substitutes the allocation result into the original formula to calculate the video cache placement again. Then, the two steps of calculating cache placement, bandwidth and power allocation are iterated multiple times to obtain the optimal solution XU for video cache placement; The rounding process to obtain the optimal feasible solution: Based on the optimal solution XU, the identifier in the result The value of is restored to 0 or 1, the process is as follows: by controlling the parameter σ Perform rounding, specifically, for each edge server, for each if set up otherwise, Denote the rounded cache variable set as X and the corresponding user experience quality value as Q X ; If X exceeds the constraints of cache space and computing resources, increase σ to σ+λ|(Q XU -Q XL ) / Q ZU |, where λ is the step size. This is because when σ is increased, the number of variables with a value of 1 decreases, the cache space required for video clips decreases, and the constraint condition can be satisfied. On the contrary, if σ satisfies the constraint, the value of σ is reduced to cache more video clips. If Q X ≥Q XL , X is a feasible result, indicating that X is a better solution than XL. The lower bound of the feasible solution is updated by setting XL←X. Repeat the above rounding process until Q X The improvement in value is less than the threshold ε; note that Q XL The value of increases with the number of iterations, because λ is a constant, Q XU is fixed, when Q XL As the value of increases, the change in σ gradually decreases, so the rounding process can converge.

5. The method for caching and resource scheduling of edge network adaptive bit rate video according to claim 4, characterized in that: In step 4), at each moment, the user request is dispatched to the serviceable base station according to the rule of minimum delay. The edge server allocates bandwidth and power to the user who initiates the request so that the average experience quality is the highest and does not exceed the bandwidth and power constraints. According to the cache result and resource allocation result, the edge server serves the user request, that is, transmits the video data. The details are as follows: User request scheduling: In dense cellular networks, users may be covered and served by multiple SBSs, so user requests are redirected to appropriate SBSs to pursue high QoE. Specifically, according to the cache results, given a user request and the set of SBSs covering the request, with fixed bandwidth and power, the video transcoding tc and transmission delay tr of each SBS are calculated, and the SBS with the lowest delay is selected to serve the request, where Where cp represents the transcoding intensity, f cpu Indicates CPU frequency, and tr = Data n / R n , where Data n Indicates the amount of data transmitted, R n Indicates the user transmission rate; Bandwidth and power allocation: Based on the request scheduling results, the SLSQP optimizer is used to solve the nonlinear integer programming problem with user experience quality as the optimization goal and bandwidth, power, cache space size, computing resources and energy consumption as constraints, and the bandwidth and power currently pre-allocated to the user are calculated to optimize the overall user experience quality. Service delivery: Serving user requests based on cache results and resource allocation results, that is, transmitting data. There are four situations in which users initiate requests to edge servers: ① The requested video segment with a specific bit rate is cached in the edge server. The video segment can be directly transmitted. Otherwise, the edge server checks whether there is a higher bit rate than the requested bit rate; ② If a high bit rate is cached, the edge server can perform the transcoding operation and then send the transcoded version to the user; ③ The requested segments are not cached in the edge server, but in the neighbor server. The edge server stores the information of neighbor servers for collaboration; when the information is updated, it will be exchanged through broadcasting; therefore, the edge server can get the video from the neighbor and decide whether to transcode it; ④ If all the cooperating edge servers do not cache the video segment requested by the user, the video with a specific bit rate is directly obtained from the cloud.