A cache prefetching method for shortening the average response time of video data in named data networks

Abstract

The present invention discloses a cache prefetching method for shortening the average response time of video data in a named data network, comprising: calculating the association relationship of video data by classifying the video data in the network, wherein the association of continuous video is calculated using the position of the video data, and the association of ordinary video data is calculated using the Apriori algorithm, generating a video data association relationship table, and recording the association relationship table in each cache node; a cache prefetching strategy, wherein when a video data request occurs, the name of the video data is obtained, and based on the type of the video data, the video data associated with the video data is found, and multiple prefetched video data are simultaneously sent to the intermediate node for pre-caching. The pre-caching strategy, wherein the CS of the cache node is divided into two levels, namely, a main cache and a sub-cache, wherein the data requested by the consumer is cached in the main cache and the pre-fetched data is cached in the sub-cache.

Description

Technical Field

[0001] The present invention designs a cache prefetching method for shortening the average response time of video data in a named data network, and belongs to the technical field of network information transmission and communication. Background Art

[0002] The internet has become a crucial foundation for modern socioeconomic development, social progress, and technological innovation. However, with ever-increasing demands, especially the rapid growth of mobile internet, a wide variety of applications are emerging. The service quality provided by the current TCP / IP-based internet communication model no longer meets the demands of these new applications. Regarding scalability, the rapid increase in the number of internet users has also led to issues such as a shortage of IPv4 addresses. Regarding security, networks currently face numerous security vulnerabilities, and addressing these issues through continuous patching has led to increasing network bloat. Regarding mobility, with the emergence and adoption of a large number of mobile devices, traditional TCP / IP networks are no longer well-suited for these upper-layer mobile applications. The primary solution to these numerous issues in traditional TCP / IP networks is incremental deployment, which involves continuously adding overlay networks and patching them. While this has improved network utilization and ensured service quality to a certain extent, it cannot fundamentally address the current challenges facing the internet. To provide better services for diverse user services, a fundamental overhaul of the internet is imperative. To address the key issues facing TCP / IP networks, the academic community is optimistic about gradually replacing the TCP / IP protocol with a new network architecture. The introduction of information-centric networks (ICNs) has revolutionized the TCP / IP communication model. ICNs are an information-centric network design and a crucial architectural research topic for future networks. Named Data Networking (NDN), modeled after ICNs, is a network architecture that uses hierarchical naming rules for the content layer. Data matching, forwarding, routing, and caching are all performed using names, making it more suitable for the internet's growing communication needs for data distribution.

[0003] NDN utilizes a completely different resource transmission method from TCP / IP networks. It directly distributes and retrieves content. To power this transmission method, NDN's design prioritizes the ability to deploy large-scale network caching, maximizing bandwidth utilization. NDN employs in-network caching, caching data objects at intermediate nodes along the network. Subsequent requests for the same data are fulfilled at the nearest node and returned directly to the requester. When NDN receives a request, it first checks whether the corresponding data packet exists in the CS. If so, it returns the data directly. The CS acts as a cache. As the CS reaches capacity, it deletes some previously stored information, retaining more useful data. This effectively utilizes cache space, making it more efficient and reliable, and improving cache utilization. NDN's caching strategies are primarily divided into cache decision and cache replacement strategies. When a requester sends a request to a server and receives a response from the server, the server returns the response data to the requester. On the data return path, an appropriate node will be selected to cache the returned data. When the same request is sent, the response data can be obtained directly from the intermediate nodes in the network, without having to access the server every time, thus reducing the pressure on the server. The decision of whether to cache data is usually called a cache decision strategy, also known as cache decision. When the cache space capacity of a routing node reaches the upper limit, some algorithms will be used to determine certain criteria to update the data in the cache, delete some relatively minor information, and leave more space for important or more popular information. The updating and selection of cached data is called a cache replacement strategy. Cache replacement is a very important issue and has always been a hot topic and focus of research.

[0004] NDN routers can cache content, but caching places significant pressure on NDN network nodes. A well-designed and effective cache is essential for making NDN networks practical. Furthermore, current network communications have evolved from a point-to-point model to one dominated by audio and video. According to rough estimates, video traffic currently accounts for 60% of total network traffic and continues to grow. Providing better services for video services will improve network performance and, in turn, provide better service for video users. This patented design implements a video data pre-caching strategy. Based on the relationships between video data, data content that users are likely to request next is pre-cached in intermediate routers. Subsequent user requests are then served at these intermediate nodes, thereby shortening the average response time for user requests. Summary of the Invention

[0005] In order to overcome the defects of existing video data caching in named data networks, the present invention provides a named data network cache prefetching method based on video association relationships, which includes: calculation of video data association relationships, wherein the method is to classify the video data in the network, wherein continuous video uses the position of the video data to calculate the association, and ordinary video data uses the Apriori algorithm to calculate the association of the video data, generate a video data association relationship table, and record the association relationship table in each cache node; cache prefetching strategy, wherein when a video data request occurs, the name of the video data is obtained, and according to the type of video data, the video data that has an association relationship with the video data is found, and multiple video data are prefetched and sent to the intermediate node for pre-caching at the same time. The pre-caching strategy, wherein the CS of the cache node is divided into two levels, namely the main cache and the sub-cache, the data requested by the consumer is cached in the main cache, and the pre-fetched data is cached in the sub-cache. The present invention includes the following steps:

[0006] Step 1: In this invention, the NDN network is divided into multiple regions, and the concepts of edge nodes and regional interconnection nodes are introduced. A request recording module is added to the edge nodes, and an auxiliary prediction module and a request recording module are added to the regional interconnection nodes. The producer structure is modified and divided into three modules: a self-prediction module, an auxiliary prediction module, and a request recording module. The cache of each intermediate cache node in the NDM network is divided into two levels: MainCache and SlaveCache. The data packet structure in the named data network is modified to add video type attributes, prefetch attributes, and association table attributes. The interest packet structure in the named data network is modified to add video type attributes and request recording table attributes.

[0007] Step 2: Divide the video data in the NDN network into three categories: real-time video data, continuous video data, and ordinary video data.

[0008] Step 3: The consumer generates a request Interest packet and sends it to request a specific video data. If the data is cached at the intermediate node, the request data packet is forwarded directly. If the data requested by the consumer is not cached at the intermediate node, the request Interest packet sent by the consumer reaches the producer, and the producer pre-fetches the data based on the requested video data type.

[0009] The specific process of step 3 is as follows:

[0010] Step 3-1: The requested video data type is real-time video data, and no prefetching is performed.

[0011] Step 3-2: When the data is continuous, the self-prediction module pre-fetches n video data that are continuous with the video data according to the name of the video data.

[0012] Step 3-3: When the data is normal, the auxiliary prediction module reads β video data with a high correlation with the requested data from the Data Relation Table.

[0013] Step 4: When the data packet sent by the producer arrives at the intermediate cache node, the node will determine whether to cache the data packet and the specific cache location based on the information in the data packet.

[0014] Step 5: The edge nodes in the NDN network record the consumer's request and periodically send the request record table to the producer and regional connection nodes in the form of interest packets. The producer and regional connection nodes use the Apriori algorithm in the auxiliary prediction module and the request record table to calculate and update the Data Relation Table, and send the updated Data Relation Table to the cache node for caching.

[0015] Step 6: The producer's auxiliary prediction algorithm module uses the Apriori algorithm and the consumer's historical request records to calculate and update the Data Relation Table. The specific process of the Apriori algorithm in the auxiliary prediction module is as follows:

[0016] Step 6-1: Initialize the data

[0017] Step 6-2: Find frequent itemsets and search iteratively layer by layer.

[0018] Step 6-3: Generate strong association rules and generate an association relationship table.

[0019] Step 7: The regional connecting node will summarize the request record tables of two or more regions and use the Apriori algorithm in the auxiliary prediction module to update the Data Relation Table. The updated Data Relation Table is sent to the nodes in each connected region in the form of a data packet. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The request record table structure designed for the present invention.

[0021] Figure 2 The association relationship table structure designed for the present invention.

[0022] Figure 3 This is the modified NDN network cache node structure diagram of the present invention.

[0023] Figure 4 This is a schematic diagram of an NDN network data packet before modification by the present invention.

[0024] Figure 5 This is a schematic diagram of the modified NDN network data packet of the present invention.

[0025] Figure 6 Schematic diagram of an NDN network interest packet before modification by the present invention.

[0026] Figure 7 This is a schematic diagram of the modified NDN network interest packet of the present invention.

[0027] Figure 8 Flowchart for the producer receiving and processing request Interest packets.

[0028] Figure 9 Flowchart of receiving and processing data packets for cache nodes. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The present invention is further described in detail below in conjunction with the drawings:

[0030] Step 1: In this invention, the NDN network is divided into multiple regions, each region is assigned a region ID, and the concepts of edge nodes and regional interconnection nodes are introduced. The edge node is the node directly connected to the consumer. The edge node adds a request record module and records the request record of each consumer in the request record table. Figure 1 The regional connection node is responsible for communication between two or more regions. The regional connection node adds an auxiliary prediction module and a request recording module. The producer structure is modified and divided into three modules, namely the self-prediction module, the auxiliary prediction module and the request recording module. The self-prediction module is used to predict the association relationship of continuous videos; the auxiliary prediction module includes the Apriori algorithm and the association relationship table. Figure 2 The Apriori algorithm uses historical request records to calculate the association between common video data and records the calculation results in the association table. The cache of each intermediate cache node in the NDN network is divided into two levels: Figure 3, namely MainCache and SlaveCache. MainCache is used to cache requested video data, and SlaveCache is used to cache pre-fetched video data. Modify the data packet structure in the named data network, add video type attributes, pre-fetch attributes, and association table attributes, modify the data packet structure in the NDM network, add the Vtype field to record the type of the video data; add the isPre field to record whether the video data is pre-fetched data; add the isDRTable to record whether the data packet is the association table of video data. The default value of this field is 0. The data packet structure before and after the modification is as follows: Figure 4 and Figure 5 As shown in the figure, the structure of the interest packet in the named data network is modified to add the video type attribute and the request record table attribute. The structure of the interest packet in the NDM network is modified to add the Vtype field to record the type of the requested video data; the isRRTable field is added to record whether the interest packet is a synchronous interest packet, that is, the content contained is the request record table. The default value of this field is 0. The structure of the interest packet before and after the modification is shown in the figure. Figure 6 and Figure 7 shown.

[0031] Step 2: Classify the video data in the network into three categories: real-time video, continuous video, and ordinary video. The Vtype field of each video type is different. When the video data is real-time video, Vtype = 0; when the video data is continuous video, Vtype = 1; when the video data is ordinary video, Vtype = 2.

[0032] Step 3: The consumer generates a request interest packet and sends the interest packet to request a certain video data. If the data is cached in the intermediate node, the request data packet is forwarded directly. If the data requested by the consumer is not cached in the intermediate node, the request interest packet sent by the consumer reaches the producer. The producer searches for the corresponding video data according to the name of the interest packet and pre-fetches the video data according to Vtype. Figure 8 .

[0033] The specific process of step 3 is as follows:

[0034] Step 3-1: For real-time video requests, such as live broadcasts, the cache nodes in the named data network do not need to cache the requested data because the requested data will not be requested again after being delivered to the real-time user. Therefore, if the requested video data type is real-time video data, no prefetching of the remaining video data is performed and the requested data packet is sent directly.

[0035] Step 3-2: When the data is continuous, that is, when Vtype is 1, the self-prediction module pre-fetches n video data that are continuous with the video data according to the name of the video data, and sets Vtype of n data packets to 1, isPre to 1, and returns them together with the requested data. The number of n depends on the relevance of the continuous video. That is, a continuous video data set {V1, V2...V k ..V n-1 ,V n}, the data requested by the interest packet is video data V k , where k represents V k The location under the continuous video data directory, V k After reaching the producer, the self-prediction module is started and V is calculated j With V k When the correlation When it is greater than the threshold λ, V will be prefetched j , the formula is as follows, where j>k:

[0036]

[0037] Step 3-3: When the data is normal data, that is, when Vtype is 2, the producer will start the auxiliary prediction module. The auxiliary prediction module will read β video data with a high correlation with the requested data in the Data Relation Table, and set the Vtype of these β data packets to 2 and isPre to 1, and return them together with the requested data.

[0038] Step 4: When the sent data packet reaches the intermediate node Figure 9 The node will first read the isDRTable field in the data packet. If isDRTable is 1, it means that the data packet is an updated association relationship table, and the node will update the DataRelation Table to the latest data; if isDRTable is 0, the node will then read the Vtype field in the data packet. If Vtype = 0, the data packet will not be cached; if Vtype is not 0, the node will read the isPre field in the data packet. If isPre is 0, it means that the data packet is data requested by the consumer, and the data packet will be cached in MainCache; if isPre is 1, it means that the data packet is pre-fetched video data, and the data packet will be cached in SlaveCache.

[0039] Step 5: The edge node in the NDN network records the consumer's request and sends the request record to the producer and regional connection node in the form of an Interest packet in time period T2. The isRRTable of the Interest packet is set to 1.

[0040] Step 6: The producer's auxiliary prediction algorithm module uses the Apriori algorithm and the latest request record to calculate and update the Data Relation Table. The update cycle time is T1. The updated Data Relation Table is sent to the cache node in the region in the form of a data packet. The isDRTable field of the Data Relation Table data packet is set to 1. The specific process of Apriori is as follows:

[0041] Step 6-1: Initialize the data: define a reasonable minimum support size.

[0042] Step 6-2: Find frequent itemsets and iterate the search layer by layer: Scan the request record table, accumulate the counts for each video data point, and collect the items that meet the minimum support. Find the set of frequent 1-item sets, denoted as L1. Then, use L1 to find the set of frequent 2-item sets, L1. Continue this iterative process until no more frequent k-item sets can be found.

[0043] Step 6-3: Generate strong association rules and generate association relationship table: Based on the found L k ,Use non-empty subsets to construct rules, and find the expressions that meet the conditions, i.e. strong association rules, and calculate the confidence degree, which is used as the correlation degree between videos.

[0044] Step 7: The regional connectivity node aggregates the request record tables from multiple regions and updates the Data Relation Table using the Apriori algorithm in the auxiliary prediction module. The update cycle is T3. The updated Data Relation Table is sent as a data packet to each node in the connectivity region. The isDRTable field in the Data Relation Table packet is set to 1. This method ensures the consistency, accuracy, and real-time nature of the video data relation table within a specific timeframe and region.

Claims

1. A cache prefetching method for shortening the average response time of video data in a named data network, characterized in that: The following steps are involved: Step 1: Divide the NDN network into multiple regions and introduce the concepts of edge nodes and regional interconnection nodes; Edge nodes are nodes directly connected to consumers, and regional interconnection nodes are nodes responsible for communication between two or more regions. A request recording module is added to edge nodes, and an auxiliary prediction module and a request recording module are added to regional interconnection nodes. The producer structure is modified and divided into three modules: a self-prediction module, an auxiliary prediction module, and a request recording module. The cache of each intermediate cache node in the NDM network is divided into two levels: MainCache and SlaveCache. MainCache is used to cache requested video data, and SlaveCache is used to cache pre-fetched video data. The data packet structure in the named data network is modified to add video type attributes, pre-fetch attributes, and association relationship table attributes. The interest packet structure in the named data network is modified to add video type attributes and request recording table attributes. Step 2: The video data in the NDN network is divided into three categories: real-time video data, continuous video data, and ordinary video data. The video type attribute Vtype field of the data packets of different video types is different. When the video data is real-time video, Vtype = 0; when the video data is continuous video, Vtype = 1; when the video data is ordinary video, Vtype = 2. Step 3: The consumer generates a request interest packet and sends it to request a certain video data. If the data is cached in the intermediate cache node, the request data packet is directly forwarded. If the data requested by the consumer is not cached in the intermediate node, the request interest packet sent by the consumer reaches the producer, and the producer pre-fetches the data according to the requested video data type. Step 4: When the data packet sent by the producer arrives at the intermediate cache node, the node will first read the isDRTable field in the data packet. If isDRTable is 1, it means that the data packet is an updated data packet, and the node will update the Data Relation Table to the latest data; if isDRTable is 0, the node will then read the Vtype field in the data packet. If Vtype = 0, the data packet will not be cached; if Vtype is not 0, the node will read the isPre field in the data packet. If isPre is 0, it means that the data packet is the data requested by the consumer, and the data packet will be cached to MainCache; if isPre is 1, it means that the data packet is pre-fetched video data, and the data packet will be cached to In SlaveCache; Step 5: The edge nodes in the NDN network record consumer requests and periodically send the request records to the producer and regional connection nodes in the form of interest packets. The producer and regional connection nodes use the Apriori algorithm in the auxiliary prediction module and the request record table to calculate and update the Data Relation Table, and send the updated Data Relation Table to the intermediate cache node for caching. The Apriori algorithm uses historical request records to calculate the association between common video data and records the calculation results in the association table. Step 6: The producer's auxiliary prediction algorithm module uses the Apriori algorithm and the consumer's historical request records to calculate and update the Data Relation Table; Step 7: The regional connecting node aggregates the request record tables of two or more regions and uses the Apriori algorithm in the auxiliary prediction module to update the Data Relation Table. The updated Data Relation Table is sent to the nodes in each connected region in the form of a data packet. In step 3 The method includes the following steps: step 3-1: for real-time video requests, there is no need for intermediate cache nodes of the named data network to cache the requested data, because the requested data will not be requested again after being delivered to the real-time user; if the requested video data type is real-time video data, no prefetching operation is performed on the remaining video data, and the requested data packet is sent directly; step 3-2: when the data is continuous, that is, when the video type attribute Vtype is 1, the self-prediction module prefetches n video data packets that are continuous with the video data according to the name of the video data, sets the video type attribute Vtype of the n data packets to 1, sets the prefetch attribute isPre to 1, and returns them together with the requested data; the number n depends on the relevance of the continuous video. That is, the continuous video data group {V1, V2...Vk..Vn-1, Vn}, the data requested by the interest packet is the video data Vk, where k represents the position of Vk in the continuous video data directory. After Vk reaches the producer and the self-prediction module is started, the correlation between Vj and Vk is calculated. When Vj is greater than the threshold λ, Vj will be prefetched. The formula is as follows, where j>k: Step 3-3: When it is ordinary data, that is, when the video type attribute Vtype is 2, the producer will start the auxiliary prediction module. The auxiliary prediction module will read β video data with a high correlation with the requested data in the association table Data Relation Table, and set the video type attribute Vtype of these β data packets to 2, and the pre-fetch attribute isPre to 1, and return them together with the requested data.

2. The cache prefetching method for shortening the average response time of video data in a named data network according to claim 1, characterized in that: The specific process of the Apriori algorithm in the auxiliary prediction module in step 6 is as follows: Step 6-1: Initialize the data: define the minimum support size; Step 6-2: Find frequent itemsets and iterate the search layer by layer: By scanning the request record table, accumulating the count of each video data, and collecting the items that meet the minimum support, find the set of frequent 1-item sets, recorded as L1; then, use L1 to find the set of frequent 2-item sets L2, and continue iterating in this way until no more frequent k-item sets can be found; Step 6-3: Generate strong association rules and generate an association relationship table: Based on the found Lk, use the non-empty subset to construct rules, and find expressions that meet the conditions, that is, strong association rules, and calculate the confidence, and use the confidence as the correlation between videos.

Images