CN114630183B - An edge device caching method and evaluation method based on scalability coding - Google Patents

Abstract

The invention discloses an edge equipment caching method and an evaluation method based on scalable coding. The algorithm comprises the following steps: collecting video request information in an area received by the equipment on an edge equipment; the collected video request information is classified according to the request time, the video number and the code rate of the video; counting the request quantity of each code rate of each video of the past time slices including a specified time slice when the specified time slice is finished; calculating the cache cost performance of each video code rate according to the request quantity; and adjusting the video content cached by the device in the next time slice according to the size sequence of the cache price. The invention also provides a method for quantifying the magnitude scores provided by the edge device for caching content for the user in a time period. The algorithm and the evaluation method thereof utilize the advantage of SVC scalable coding on the cache, can effectively improve the cache efficiency of single equipment, and improve the capability of providing video services for users by utilizing the edge cache technology.

Description

An edge device caching method and evaluation method based on scalability coding

Technical field

The invention belongs to the field of computer network technology, and in particular relates to an edge device caching method and evaluation method based on scalability coding.

Background technique

With the increasing number of online video streaming services, video streaming services are becoming the most popular business on the Internet. According to Cisco's predictions, by the end of 2021, video streaming services will account for 82% of all network traffic. However, efficiently delivering video content remains a challenging task due to network heterogeneity in terms of resolution, frame rate, and bit depth among users with different preferences and constraints.

Two important evaluation criteria for video on demand service quality are video response delay and occupied bandwidth. Video response delay directly affects the user's experience when watching videos. The shorter the video response delay, the more comfortable the user experience when watching the video, thereby improving the server quality of the video on demand service. The bandwidth occupied by the video affects the network resources consumed during the transmission of the video. If the video occupies too much bandwidth, it may cause network congestion when a large number of video on demand requests appear in an area, affecting the delay and stability of video transmission. Degrade the viewing experience for all video on demand users in the area.

In order to provide high-quality video on-demand services, current video providers always use content delivery networks (CDNs) to provide on-demand videos. CDN adds a new layer of network architecture to the existing network to publish and transmit the content of the origin site to the edge area closest to the user, so that the user can access the desired content nearby, thus improving the response speed of user access. The basic principle of CDN is to rely on caching servers placed in various places, through global scheduling and content distribution and other functional modules, to deploy the part of the content that users need to the place closest to the user, reducing the user's physical distance when providing video on demand services. The resulting delay transforms an originally inefficient and unreliable network into an efficient and reliable intelligent network, meeting users' higher requirements for content access quality, improving Internet network congestion problems, and improving the response speed for users to access websites. Additionally, the performance of video content caching devices at the edge of the network is comparable to that at mid-network nodes.

However, with the popularity of emerging short video platforms and the increase in user demand for video services, video traffic on the Internet has exploded, making it increasingly difficult for traditional CDN technology to meet users' increasingly demanding demands for video services. , at this time the concept of edge caching was proposed. Traditional networks deploy content (videos, web pages, etc.) in data centers or regional CDN cache servers. There are problems such as long end-to-end delay in content acquisition, limited backhaul bandwidth, and inefficient redundant transmission. They cannot meet the needs of 5G and the future. Application requirements such as low latency. The core idea of mobile edge caching is to sink content to network access units (such as 5G base stations, road base units, etc.) to achieve one-hop nearby content services, which can significantly reduce end-to-end delay and improve network transmission efficiency.

Contents of the invention

Due to the limited capacity of the edge cache, how to make the most of the limited cache space to provide high-quality video services to video on demand users is a problem.

In order to solve the above problems, the present invention discloses an edge device caching method based on scalability coding, SVC-BEC (SVC based on edge caching),

Includes the following steps:

S1. Collect video request information in the area received by the device on an edge device;

S2. The collected video request information is classified according to request time, requested video number, and requested video bit rate;

S3. At the end of a specified time slice, count the number of requests for each video bit rate in the past several time slices including this time slice;

S4. Calculate the cache cost-effectiveness of each video bit rate based on the number of requests;

S5: Adjust the video content cached by the device in the next time slice according to the order of cache cost performance.

The present invention further prefers that step S1 specifically includes the following:

S11, select a normally operating edge cache device and determine its cache space size;

S12, count the video caching services that the device can provide, and design the data structure for collecting request information;

S13, collect data in the actual environment and record the request time.

The present invention further prefers that step S2 specifically includes the following:

S21, classify the collected information according to the request time, requested video number, and requested video bit rate;

S22, classify the collected data into a data structure according to the previous design;

S23: Store the classified data information into the device.

The present invention further prefers that step S3 specifically includes the following:

S31, design the time slice length, and pause the video request response at the end of a time slice;

S32, count the number of requests for each video bit rate in the past several time slices including this time slice;

S33: Classify the statistically completed data according to video type and bit rate.

The present invention further prefers that step S4 specifically includes the following:

S41, based on time limitations, it is believed that the number of video requests indicates the popularity of the video;

S42, extract the number of requests for each bit rate of each video in the file in the past m time slices and sum them up;

S43, divide the popularity by the space occupied by the cached video to obtain the cache cost-effectiveness of the video in the past m time slices;

S45 stores the cache cost-effectiveness of each bit rate of each video into a file.

The present invention further prefers that the S41 is specifically:

S411, retrieve the file containing the video request information;

S412, check whether each video has received requests from users in the past m time periods;

S43, if a bit rate of a video has been requested in the past m time slices, then its popularity is calculated.

The present invention further prefers that step S5 specifically includes the following:

S51, sort the cache cost-effectiveness of all videos requested in the past from high to low;

S52 starts with the most cost-effective video and stores it in the current edge cache device;

S53, during the placement process, if it is found that the current video i has been put into the device, and if the bit rate of the stored video i is higher than the currently processed bit rate, then the current bit rate version will be skipped. If it has been stored The bit rate is lower than the bit rate of the currently processed video i, then the corresponding bit rate of the currently processed video i overwrites the previously stored video i;

S54, repeat this process until the space occupied by the video stored in device D reaches the storage upper limit.

An edge device cache quality assessment method based on scalability coding of the present invention specifically includes the following steps:

S6. Since the peak signal-to-noise ratio (PSNR) is directly proportional to the video bit rate, the logarithm of the video bit rate is used to represent the peak signal-to-noise ratio, which represents the user experience score obtained from the video;

S7. Calculate the sum of user experience scores provided by all videos cached in the device in the past time period;

S8. Divide the sum by the number of user requests in the past time period to normalize, and obtain the cache solution user experience score S of the device in the past time period.

The present invention is further preferred: wherein step S7 specifically includes the following:

S71, for a video code rate, multiply the number of user requests by the PSNR value, which is the logarithm of the code rate, to obtain the video score s;

S72, count the sum of scores s of all types of video bit rates;

S73: Calculate the sum S of scores of the videos requested in the past time slice.

The present invention is further preferred: in step S8, if S is larger, it means that the quality of the video service provided by the device to the user in the past time slice is higher; if S is smaller, it means that the device is in the past time slice. The lower the quality of the video service provided to users

Beneficial effects of the present invention: For video on demand services, by utilizing the advantages of scalable coding in storage and based on the time limitation principle, the cache quality of edge devices can be effectively improved, and can be evaluated through user scores.

Description of drawings

Figure 1 schematically shows the flow chart of the statistical method provided by the embodiment of the present invention;

Figure 2 schematically shows a structure diagram of video request information collection provided by an embodiment of the present invention;

Figure 3 schematically shows a cache content replacement flow chart provided by an embodiment of the present invention.

Detailed ways

The present invention will be further clarified below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. It should be noted that the words "front", "back", "left", "right", "upper" and "lower" used in the following description refer to the directions in the drawings, and the words "inside" and "outside" ” refers to the direction toward or away from the geometric center of a specific part, respectively.

As shown in Figure 1, an embodiment of the present invention provides an edge device caching method based on scalability coding, including:

S1. Collect on an edge device the video request information in the area received by the device. Figure 2 schematically shows the information collection structure diagram. Refer to Figure 2 and the following is a detailed description.

The specific implementation content is:

First, select the edge device, record its cache space size D, deploy probes on the edge device to obtain user request information, use multiple network cards to collect the hardware, and use the switch to shunt the original request traffic to each network card through load balancing technology. Meet information collection needs.

Then, a double-buffer mechanism is implemented on the network card buffer, and the order in which the buffers take effect is controlled through time slices, thus ensuring that the network card is always available. When a network card buffer is full, the worker thread passes the buffer address to the reentrant data collection interface.

Finally, in order to collect request data from all network cards in parallel, multi-threading technology is used. Deploy the DPDK suite on the collection server, and its fast packet processing function can be used for information collection after adjustment; the collection server transfers the user's video request information into the storage area.

S2. Classify the collected video request information according to request time, requested video number, and requested video bit rate;

The specific implementation content is: extract the collected video request information from the storage area; classify the collected information according to the request time, requested video number, and requested video bit rate; count the total number of video content requests in each time slice; according to the previous The design classifies the collected data into data structures.

S3. At the end of a specified time slice, count the popularity of each video and each code rate in the past several time slices including this time slice. The specific implementation steps are described as follows:

First, for a code rate j of a video i, extract the number of times C _ij requested by the user in the past m time slices.

Then, in order to eliminate the difference between time slices with more requests and less requests, the extracted number of requests for a specific video bit rate is divided by the corresponding total number of requests for each time slice Rt, thereby normalizing .

Finally, add the standardized m data to get the popularity of the video bit rate in the past m time slices, that is

Repeat the above operation for all video bit rates that have been requested in the past m time slices.

S4. Calculate the cache cost-effectiveness of each video bit rate based on the number of requests;

Due to the principle of time limitation, videos that were popular in the past will definitely be popular in the future, that is, they will receive more user requests. However, due to the limited cache space of the device, caching too large videos will reduce the total number of cacheable videos. The number of videos causes the overall request hit rate of the device to decrease. Therefore, in order to improve the overall cache quality of the cache device, calculate the cache cost-effectiveness of each video bit rate and select cached videos based on this.

For the past m time slices, locate all video bit rate popularity information files within that time period; read the file, extract the popularity value C _ij ′ of all video bit rates in the file, and divide the popularity value by the video The bit rate v _ij of the version represents the space occupied by the video, thereby obtaining the cache cost-effectiveness W _ij of the video version.

S5: Adjust the video content cached by the device in the next time slice according to the order of cache performance and price.

For all code rate versions of all videos, sort W _ij from large to small, and store it in the edge device starting from the first position. During the placement process, if it is found that the current video i has been placed in the device, if it has been stored If the bit rate of video i is higher than the currently processed bit rate, then the current bit rate version will be skipped. If the already stored bit rate is lower than the bit rate of currently processed video i, then the corresponding version of currently processed video i will be skipped. The code rate covers the previously stored video i, and this process is repeated until the space occupied by the video stored in device D reaches the storage upper limit M. A specific example of this process is shown in Figure 3

The present invention also provides a cache quality evaluation method, including:

S6. Since the peak signal-to-noise ratio PSNR is proportional to the video bit rate, the logarithm of the video bit rate is used to represent the peak signal-to-noise ratio, which represents the experience score the user obtains from the video, that is, it is considered that S _ij =PSNR _ij =lgv _ij The experience score the user gets from a hit requesting the bitrate j version of video i;

S7: Calculate the sum of user experience scores provided by all videos cached in the device in the past time period. That is, for a video bit rate, multiply the number of user requests by the PSNR value, which is the logarithm of the bit rate, to get the video score s; count the sum of the scores s of all types of video bit rates; count the sum of the scores of the videos requested in the past time slice and S;

S8. Divide the sum by the number of user requests in the past time period to normalize, that is Thus, the cache solution user experience score S of the device in the past time period is obtained. If S is larger, it means that the quality of the video service provided by the device to the user in the past time slice is higher; if S is smaller, it means that the quality of the video service provided by the device to the user in the past time slice is lower.

The technical means disclosed in the solution of the present invention are not limited to the technical means disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features.

Claims (7)

1. An edge device caching method based on scalability coding, characterized by: including the following steps: S1. Collect video request information in the area received by the device on an edge device; S2. The collected video request information is classified according to request time, requested video number, and requested video bit rate; S3. At the end of a specified time slice, count the number of requests for each video bit rate in the past several time slices including this time slice; the specific implementation steps are as follows: First, for a type of video i and a code rate j, extract the number of times C _ij it has been requested by the user in the past m time slices; Then, in order to eliminate the difference between time slices with more requests and less requests, the extracted number of requests for a specific video bit rate is divided by the corresponding total number of requests for each time slice Rt, thereby normalizing ; Finally, add the standardized m data to get the popularity of the video bit rate in the past m time slices, that is Repeat the above implementation steps for all video bit rates that have been requested in the past m time slices; S4. Calculate the cache cost-effectiveness of each video bit rate based on the number of requests; details include the following: S41, based on time limitations, it is believed that the number of video requests indicates the popularity of the video; specifically: S411, retrieve the file containing the video request information; S412, check whether each video has received requests from users in the past m time periods; S413, if a bit rate of a video has been requested in the past m time slices, then count its popularity; S42, extract the number of requests for each bit rate of each video in the file in the past m time slices and sum them up; S43, divide the popularity by the space occupied by the cached video to obtain the cache cost-effectiveness of the video in the past m time slices; S45, save the cache cost-effectiveness of each bit rate of each video into a file; Among them, for the past m time slices, locate all video bit rate heat information files within this time period; read the file, extract the heat value C _ij ′ of all video bit rates in the file, and divide the heat value by the The bitrate v _ij of the video version represents the space occupied by the video, thereby obtaining the cache cost-effectiveness W _ij of the video version; S5. Adjust the video content cached by the device in the next time slice according to the order of cache cost performance; specifically including the following: S51, sort the cache cost-effectiveness of all videos requested in the past from high to low; S52 starts with the most cost-effective video and stores it in the current edge cache device; S53, during the placement process, if it is found that the current video i has been put into the device, and if the bit rate of the stored video i is higher than the currently processed bit rate, then the current bit rate version will be skipped. If it has been stored The bit rate is lower than the bit rate of the currently processed video i, then the corresponding bit rate of the currently processed video i overwrites the previously stored video i; S54, repeat this process until the space occupied by the video stored in device D reaches the storage upper limit. 2. An edge device caching method based on scalability coding according to claim 1, characterized in that: The step S1 specifically includes the following: S11, select a normally operating edge cache device and determine its cache space size; S12, count the video caching services that the device can provide, and design the data structure for collecting request information; S13, collect data in the actual environment and record the request time. 3. An edge device caching method based on scalability coding according to claim 2, characterized in that step S2 specifically includes the following: S21, classify the collected information according to the request time, requested video number, and requested video bit rate; S22, classify the collected data into a data structure according to the previous design; S23: Store the classified data information into the device. 4. An edge device caching method based on scalability coding according to claim 1, characterized in that step S3 specifically includes the following: S31, design the time slice length, and pause the video request response at the end of a time slice; S32, count the number of requests for each video bit rate in the past several time slices including this time slice; S33: Classify the statistically completed data according to video type and bit rate. 5. An edge device cache quality assessment method based on scalability coding according to claim 1, characterized in that it specifically includes the following steps: S6. Since the peak signal-to-noise ratio (PSNR) is directly proportional to the video bit rate, the logarithm of the video bit rate is used to represent the peak signal-to-noise ratio, which represents the user experience score obtained from the video; S7. Calculate the sum of user experience scores provided by all videos cached in the device in the past time period; S8. Divide the sum by the number of user requests in the past time period to normalize, and obtain the cache solution user experience score S of the device in the past time period. 6. An edge device cache evaluation method based on scalability coding according to claim 5, characterized in that step S7 specifically includes the following: S71, for a video code rate, multiply the number of user requests by the PSNR value, which is the logarithm of the code rate, to obtain the video score s; S72, count the sum of scores s of all types of video bit rates; S73: Calculate the sum S of scores of the videos requested in the past time slice. 7. An edge device cache evaluation method based on scalability coding according to claim 5, characterized in that: in step S8, if S is larger, it means that the device provided video to the user in the past time slice. The higher the quality of the service; if S is smaller, it means that the quality of the video service provided by the device to users in the past time slice is lower.