NDN congestion control method in heterogeneous network based on topology awareness

Abstract

The invention discloses an NDN congestion control method in a heterogeneous network based on topology awareness, and belongs to the field of computer networks. The method comprises the following steps: 1) an edge server perceives a shared link through an interest packet custom field; 2) the edge server calculates a link score through the interest packet custom field; 3) the edge server performs data packet mapping according to the link score and obtains each path sending window; and 4) the NDN router caches the data packet according to the custom field setting of the data packet, and provides a quick recovery function. According to the method, the user-defined field content of the interest packet is analyzed through the edge server, and the path state information is constructed; and based on the path state information, the edge server binds the data packet to be sent with the path ID, and dynamically allocates a window sequence according to the path quality information. And when the data packet is lost, the user quickly locates a fault path according to the mapping relation to realize quick recovery of the lost data. Compared with the existing congestion control mechanism, the method has the advantages that the content request completion time of the user can be effectively shortened, the path bandwidth utilization rate is remarkably improved while the lower content retransmission proportion is maintained, and the data transmission requirements of large data volume and low time delay are met.

Description

Technical Field

[0001] The present invention belongs to the field of computer networks, and in particular relates to the design of a congestion control mechanism, and specifically to an NDN congestion control method in heterogeneous networks based on topology awareness. Background Art

[0002] With the rapid development of the internet, global network traffic is experiencing explosive growth. Traditional Internet Protocol (IP) networks primarily rely on location-centric data transmission. This approach performed well in the early days of the internet, but has gradually exposed its limitations in the complex and ever-changing modern network environment. Named Data Networking (NDN), on the other hand, enables retrieval and transmission by naming data, improving data reuse and network resource utilization, and enabling efficient dissemination of data content. TCP / IP is currently the most widely used network architecture, while NDN, as an emerging network architecture, is still in the development and exploration stages. It is unrealistic to completely replace TCP / IP with NDN in the short term. A coexisting network architecture, connecting the TCP / IP backbone network and the NDN local area network with edge routers, can ensure that existing TCP / IP services are not impacted, ensuring compatibility and a smooth transition as NDN is gradually implemented.

[0003] Consider the following scenario: the cloud server is located in the IP network, the user is located in the NDN, and the two communicate across the network structure through the edge server. When massive data needs to be transmitted from the IP network to the NDN user, congestion problems in two key areas are particularly prominent: (1) The IP network domain. During the data transmission process from the cloud server to the edge server, the traditional single-path TCP is limited by the bottleneck bandwidth, and the static multi-path solution cannot adapt to dynamic topology changes, resulting in large fluctuations in link utilization and the data flow is easily blocked here. (2) The NDN domain. During the transmission stage from the edge server to the end user, although NDN natively supports multi-path transmission, it lacks an intelligent congestion control mechanism. When multiple users make concurrent requests, the large amount of data leads to increased competition for available resources, which may cause transmission bottlenecks or high latency. These factors work together to make the transmission efficiency of these two areas easily reduced when large amounts of data are transmitted, and may even cause a large amount of packet loss and severe congestion.

[0004] Currently, numerous congestion control algorithms exist for both TCP / IP and NDN network architectures. However, these algorithms are specific to specific network architectures and fail to consider congestion issues in coexisting network architectures. Deploying congestion control algorithms solely for TCP / IP or NDN without considering the coexisting network architecture will inevitably lead to increased network congestion due to the fundamental differences between TCP / IP and NDN.

[0005] Therefore, to solve the congestion problem, the best approach is to propose targeted congestion control algorithms to address the challenges of large data volumes and low-latency transmission in networks where NDN and IP coexist. Summary of the Invention

[0006] Research focuses on the flow of large amounts of data within coexisting networks in coexisting network architectures. To address the challenges of high-volume, low-latency transmission of large amounts of data in NDN and IP coexisting networks, a topology-aware NDN congestion control mechanism for heterogeneous networks was proposed. This mechanism parses custom fields in interest packets via edge servers to construct path status information. Based on this path status information, the edge server binds the data packet to be sent to the path ID and dynamically allocates a window sequence based on path quality information. In the event of data packet loss, users can quickly locate the faulty path based on this mapping, enabling rapid recovery of lost data.

[0007] The present invention is achieved through the following technical solutions:

[0008] A topology-aware NDN congestion control method in a heterogeneous network includes the following steps:

[0009] Step 1: The edge server perceives the shared link through the custom field of the interest packet;

[0010] The Interest packet adds a Sharing Tag field to the existing one. The Sharing Tag field is used to assist the edge server in determining whether there are shared links among multiple paths from the user to the edge server. This field is designed as a 32-bit binary field with an initial value of all 0s.

[0011] If an NDN router has two or more forwardable interfaces and the Sharing Tag field is all 0, it means that the link has forked for the first time. Then the NDN router will modify the shared identifier field in the interest packet. According to the names of the first N contents in its own CS storage space, a hash operation is performed using the CRC32 hash function. Subsequently, all hash results are XORed and the XOR result is placed in the Sharing Tag. The calculation formula of the Sharing Tag is shown in Formula 1, where Name N Indicates the Nth name in the CS storage space. CRC32(Name1) indicates that Name1 is hashed using the CRC32 hash function.

[0012]

[0013] When the edge server receives the interest packet, it checks the shared identification field and infers the network topology.

[0014] Step 2: The edge server calculates the link score through the custom fields of the interest packet;

[0015] The calculation of the path score involves transmission status indicators and time dimension indicators. The transmission status indicator is related to whether packet loss occurred in the previous round of transmission on the path, and the time dimension indicator is related to the transmission time of the interest packet. Specifically, the transmission status indicator is represented by the custom Quality Score field in the interest packet. This field is an integer field with a maximum value of Quality Score. max , the default value is The specific calculation method is shown in Formula 2:

[0016]

[0017] When the interest packet arrives at the edge server, the edge server records the arrival time, and calculates the delay Time of each interest packet based on the Send Time field value carried by the interest packet. The calculation formula is shown in Formula 3:

[0018] Time=Arriving Time-Send Time (Formula 3)

[0019] In the Time of each collected interest packet, determine the maximum time difference Time max and the minimum time difference Time min . For the time dimension indicator, the present invention sets the maximum score to 1-δ and the minimum score to ε. By reasonably selecting the values ​​of δ and ε, the practicality and fairness of the time dimension score Time Score are ensured, and the meaninglessness caused by excessively large or small score values ​​is avoided. The calculation formula for the time dimension score is shown in Formula 4. The link score Score is shown in Formula 5, where α is the score coefficient, 0<α<1.

[0020]

[0021] Assuming that N paths share the same link, the score of this abstract path is share The calculation formula is shown in Formula 6, where Score N Indicates the score of the Nth path:

[0022]

[0023] Step 3: The edge server maps the data packets based on the link scores and derives the sending windows for each path.

[0024] The present invention adjusts the amount of data allocated to each path according to the path quality. In order to indicate the path quality, a scoring threshold Score is set. t, when the path score reaches Score t When , it indicates that the path quality is good and the data packet can be transmitted as expected. For a data packet with sequence number DataSeq, in a network with PathNum non-shared paths, the calculation formula for its sending path sequence number PathSeq is shown in Formula 7. If the path score mapped to the data packet is less than Score t , then the data packet is subjected to secondary path mapping, and the final sending path sequence number PathSeq′ is calculated as shown in Formula 8. Through secondary path mapping, the data packet will be sent to the new path.

[0025] PathSeq=DataSeq%PathNum+1 (Formula 7)

[0026]

[0027] In order to enable the edge server to calculate the sending window of each path relatively accurately and ensure the low time complexity of the calculation process, this paper adopts a shallow neural network to predict and determine the sending window size of each path based on the status score of each path and the length of the sending queue.

[0028] Step 4: The NDN router caches data packets according to the custom field settings of the data packets to provide fast recovery capabilities.

[0029] The present invention adds a custom CS Tag field to the data packet. During multi-path transmission, there are three data packet cache states:

[0030] (1) Priority caching. The CS Tag of the data packet transmitted on the low-scoring link should be the priority caching tag. The CS Tag prefix of the priority caching data packet is "11". The initial caching probability is A. Let the lower limit of the probability change be C and the upper limit of the probability change be D. Then, when the data packet passes through a router, the caching probability is P cache , the router will generate a random number E between 1 and 100, and the next route cache probability P′ cache The formula is shown in Formula 9, where Random(C,D) represents the random number generated in the interval [C,D]. If the router does not cache the data packet, then P′ cache Converted into 7-bit binary, the high-order bits are padded with 0 if the bits are less than 7, and then a "11" prefix is ​​added to indicate the priority cache type; if the router caches the data packet, the CS Tag is set to all 0s, indicating that the data packet has been cached and the next router will not cache it again.

[0031]

[0032] (2) Normal caching. The CS Tag of the data packet transmitted on the high-scoring link should be the normal cache tag. The CS Tag prefix of the normal cache data packet is "10". The initial cache probability is B. Let the lower limit of the probability change be C' and the upper limit of the probability change be D'. Then, when the data packet passes through a router, the cache probability is P cache , the router will generate a random number E in the range of 1-100, and then the next route cache probability P′ cache The formula is shown in Formula 10, where Random(C′,D′) represents the random number generated in the interval [C′,D′]. If the router does not cache the data packet, P′ cache Converted to 7-bit binary, if the number is less than 7, the high-order bits are padded with 0, and then a "10" prefix is ​​added to indicate the normal cache type; if the router caches the data packet, the CS Tag is set to all 0s, indicating that the data packet has been cached and the next router will not cache it again.

[0033]

[0034] (3) Cached. This indicates that the packet has been cached by an NDN router and its CS Tag is all zeros. To fully utilize cache resources and reduce cache redundancy in the network, cached packets are no longer cached repeatedly in NDN routers.

[0035] The present invention designs a CS occupation threshold CS for NDN routers. t and time threshold. When an NDN router is away from the last CS clearing time (timer T cs ) exceeds the threshold, and the remaining CS space of this node is monitored. free When the value is less than the threshold, the NDN router sends a popularity table request to the edge server, satisfying both conditions in Equation 11. Upon receiving the request, the edge server returns the first N records in the popularity table. The NDN router that receives the popularity table forwards it to each connected router. Upon receiving the popularity table, the NDN router checks its cache. If a packet appears in the popularity table, it retains it; otherwise, it is discarded.

[0036]

[0037] The edge server uses the same method as the user to calculate the mapping between data packets and path numbers. When a user loses a data packet, the edge server maps the path number to the packet number and sends a data request to the specified path, allowing for rapid recovery from transmission anomalies caused by missing data.

[0038] Beneficial effects of the present invention: The method of the present invention parses the custom field content of the interest packet through the edge server to construct the path status information. Based on the path status information, the edge server binds the data packet to be sent with the path ID, and dynamically allocates the window sequence according to the path quality information. When a data packet is lost, the user can quickly locate the faulty path according to the mapping relationship to achieve rapid recovery of lost data. The present invention and the existing congestion control mechanism were compared in experiments. Compared with the existing congestion control mechanism, the present invention can effectively shorten the completion time of user content requests, and while maintaining a low content retransmission ratio, it significantly improves the path bandwidth utilization, meeting the data transmission requirements of large data volumes and low latency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the NDN congestion control method in heterogeneous networks based on topology awareness;

[0040] Figure 2 Experimental network topology diagram;

[0041] Figure 3 Shared path identification accuracy topology map;

[0042] Figure 4 User request completion time results under different congestion control mechanisms;

[0043] Figure 5 Results of content retransmission ratio under different congestion control mechanisms;

[0044] Figure 6 Result graph of shared path identification accuracy under different congestion control mechanisms;

[0045] Figure 7 The results of content retransmission ratio changing with sending window under different congestion control mechanisms. DETAILED DESCRIPTION

[0046] A topology-aware NDN congestion control method in a heterogeneous network includes the following steps:

[0047] Step 1: The edge server perceives the shared link through the custom field of the interest packet;

[0048] First, the Interest packet adds a Sharing Tag field to the existing one. The Sharing Tag field is used to assist the edge server in determining whether there are shared links among the multiple paths from the user to the edge server. This field is designed as a 32-bit binary field with an initial value of all 0s.

[0049] If an NDN router has two or more forwardable interfaces and the Sharing Tag field is all 0, it means that the link has forked for the first time. Then the NDN router will modify the shared identifier field in the interest packet. According to the names of the first N contents in its own CS storage space, a hash operation is performed using the CRC32 hash function. Subsequently, all hash results are XORed and the XOR result is placed in the Sharing Tag. The calculation formula of the Sharing Tag is shown in Formula 1, where Name N Indicates the Nth name in the CS storage space. CRC32(Name1) indicates that Name1 is hashed using the CRC32 hash function.

[0050]

[0051] like Figure 1 As shown in Figure 1, Interest packets and data transmitted over the network carry modified network state information. Edge servers analyze Interest packet fields to detect multiple paths and shared links in the network. Based on the status scores of each path, they calculate a score for each path and use this score to determine the maximum data transmission window for each path. To minimize retransmission time in the event of packet loss, edge servers map data to paths when sending data, facilitating subsequent queries to cache specific paths for data.

[0052] Step 2: The edge server calculates the link score through the custom field of the interest packet;

[0053] The calculation of the path score involves transmission status indicators and time dimension indicators. The transmission status indicator is related to whether packet loss occurred in the previous round of transmission on the path, and the time dimension indicator is related to the transmission time of the interest packet. Specifically, the transmission status indicator is represented by the custom Quality Score field in the interest packet. This field is an integer field with a maximum value of Quality Score. max , the default value is The specific calculation method is shown in Formula 2:

[0054]

[0055] When the interest packet arrives at the edge server, the edge server records the arrival time, and calculates the delay Time of each interest packet based on the Send Time field value carried by the interest packet. The calculation formula is shown in Formula 3:

[0056] Time=Arriving Time-Send Time (Formula 3)

[0057] In the Time of each collected interest packet, determine the maximum time difference Time max and the minimum time difference Time min . For the time dimension indicator, the present invention sets the maximum score to 1-δ and the minimum score to ε. By reasonably selecting the values ​​of δ and ε, the practicality and fairness of the time dimension score Time Score are ensured, and the meaninglessness caused by excessively large or small score values ​​is avoided. The calculation formula for the time dimension score is shown in Formula 4. The link score Score is shown in Formula 5, where α is the score coefficient, 0<α<1.

[0058]

[0059] Assuming that N paths share the same link, the score of this abstract path is share The calculation formula is shown in Formula 6, where Score N Indicates the score of the Nth path:

[0060]

[0061] Step 3: The edge server maps the data packets based on the link scores and derives the sending windows for each path.

[0062] The present invention adjusts the amount of data allocated to each path according to the path quality. In order to indicate the path quality, a scoring threshold Score is set. t , when the path score reaches Score t When , it indicates that the path quality is good and the data packet can be transmitted as expected. For a data packet with sequence number DataSeq, in a network with PathNum non-shared paths, the calculation formula for its sending path sequence number PathSeq is shown in Formula 7. If the path score mapped to the data packet is less than Score t , then the data packet is subjected to secondary path mapping, and the final sending path sequence number PathSeq′ is calculated as shown in Formula 8. Through secondary path mapping, the data packet will be sent to the new path.

[0063] PathSeq=DataSeq%PathNum+1 (Formula 7)

[0064]

[0065] In order to enable the edge server to calculate the sending window of each path relatively accurately and ensure the low time complexity of the calculation process, this paper adopts a shallow neural network to predict and determine the sending window size of each path based on the status score of each path and the length of the sending queue.

[0066] Step 4: The NDN router caches data packets according to the custom field settings of the data packets to provide fast recovery capabilities.

[0067] The present invention adds a custom CS Tag field to the data packet. During multi-path transmission, there are three data packet cache states:

[0068] (1) Priority caching. The CS Tag of the data packet transmitted on the low-scoring link should be the priority caching tag. The CS Tag prefix of the priority caching data packet is "11". The initial caching probability is A. Let the lower limit of the probability change be C and the upper limit of the probability change be D. Then, when the data packet passes through a router, the caching probability is P cache , the router will generate a random number E between 1 and 100, and the next route cache probability P′ cache The formula is shown in Formula 9, where Random(C,D) represents the random number generated in the interval [C,D]. If the router does not cache the data packet, then P′ cache Converted into 7-bit binary, the high-order bits are padded with 0 if the bits are less than 7, and then a "11" prefix is ​​added to indicate the priority cache type; if the router caches the data packet, the CS Tag is set to all 0s, indicating that the data packet has been cached and the next router will not cache it again.

[0069]

[0070] (2) Normal caching. The CS Tag of the data packet transmitted on the high-scoring link should be the normal cache tag. The CS Tag prefix of the normal cache data packet is "10". The initial cache probability is B. Let the lower limit of the probability change be C' and the upper limit of the probability change be D'. Then, when the data packet passes through a router, the cache probability is P cache , the router will generate a random number E in the range of 1-100, and then the next route cache probability P′ cache The formula is shown in Formula 10, where Random(C′,D′) represents the random number generated in the interval [C′,D′]. If the router does not cache the data packet, P′ cache Converted to 7-bit binary, if the number is less than 7, the high-order bits are padded with 0, and then a "10" prefix is ​​added to indicate the normal cache type; if the router caches the data packet, the CS Tag is set to all 0s, indicating that the data packet has been cached and the next router will not cache it again.

[0071]

[0072] (3) Cached. This indicates that the packet has been cached by an NDN router and its CS Tag is all zeros. To fully utilize cache resources and reduce cache redundancy in the network, cached packets are no longer cached repeatedly in NDN routers.

[0073] The present invention designs a CS occupation threshold CS for NDN routers. t and time threshold. When an NDN router is away from the last CS clearing time (timer T cs ) exceeds the threshold, and the remaining CS space of this node is monitored. free When the value is less than the threshold, the NDN router sends a popularity table request to the edge server, satisfying both conditions in Equation 11. Upon receiving the request, the edge server returns the first N records in the popularity table. The NDN router that receives the popularity table forwards it to each connected router. Upon receiving the popularity table, the NDN router checks its cache. If a packet appears in the popularity table, it retains it; otherwise, it is discarded.

[0074]

[0075] The edge server uses the same method as the user to calculate the mapping between data packets and path numbers. When a user loses a data packet, the edge server maps the path number to the packet number and sends a data request to the specified path, allowing for rapid recovery from transmission anomalies caused by missing data.

[0076] Comparative Example:

[0077] In order to test the performance of the method of the present invention, the SBAMPS congestion control algorithm and the MPCC congestion control algorithm are used. Figure 2 and Figure 3 The topology shown is compared with NTACC, a topology-aware congestion control method for heterogeneous NDN networks proposed in this paper. Statistics are collected for four aspects: request completion time (RCT), recognition accuracy rate (RAR), content retransmission rate (DRR), and request satisfaction rate (RFR).

[0078] exist Figure 4 In the process, because NTACC tries to improve the network throughput and avoid packet loss as much as possible, large-scale data requests from users can be completed in one request and one response. This significantly reduces the time it takes for NTACC to complete user requests. Figure 5NTACC, with its precise bottleneck link identification capabilities and efficient cache management strategies, enables edge servers to clearly understand NDN status and accurately predict the send window for each interface, thereby improving network throughput while minimizing packet loss. Compared to SBAMPS and MPCC, NTACC reduces the data retransmission rate, eliminating the need for edge servers to retransmit packets.

[0079] exist Figure 6 The NTACC approach optimizes the recognition process by cleverly adding fields to the Interest packet. When an intermediate node detects that two or more interfaces are available for data forwarding, it embeds specific information in the Interest packet field. Given the small size of the Interest packet and its minimal use of network resources, congestion is extremely rare, and the risk of packet loss is significantly reduced. Even with increasing network topology complexity, NTACC maintains its high level of recognition accuracy, demonstrating a significant advantage.

[0080] exist Figure 7 In this study, SBAMPS's ability to detect bottleneck paths was inferior to NTACC's. Because SBAMPS couldn't accurately grasp the real-time status of each path, the neural network struggled to accurately predict the send window for each interface. This limitation caused the actual maximum capacity of some paths to fall below the allocated send window, leading to packet loss and a significant increase in the retransmission rate. In contrast, NTACC maintained high bottleneck identification accuracy even in complex network environments. Thanks to its precise identification capabilities, packet loss was effectively controlled, keeping the retransmission rate low.

[0081] Experimental results show that NTACC exhibits higher robustness and adaptability in a changing network environment, verifying the effectiveness and superiority of NTACC in improving network transmission efficiency and user experience.

Claims

1. A topology-aware NDN congestion control method in a heterogeneous network, involving information exchange between edge servers in a heterogeneous network and user nodes in an NDN, characterized in that: Here are the steps: 1) The edge server perceives the shared link through the custom field of the interest packet; 2) The edge server calculates the link score through the custom fields of the interest packet; 3) The edge server maps data packets based on the link scores and derives the sending windows for each path; 4) NDN routers cache packets based on custom field settings in packets, providing fast recovery capabilities.

2. The NDN congestion control method in a heterogeneous network based on topology awareness according to claim 1 is characterized in that: In the above 1), the specific method is: The Interest packet adds a Sharing Tag field to the original one to assist the edge server in determining whether there is a shared link among multiple paths from the user to the edge server. This field is designed as a 32-bit binary field with an initial value of all 0s. If an NDN router has two or more forwarding interfaces and the Sharing Tag field is all zeros, it indicates that the link has forked for the first time. The NDN router then modifies the Sharing Identifier field in the Interest packet and performs a CRC32 hash operation based on the names of the first N contents in its CS storage space. It then performs an XOR operation on all hash results and places the XOR result in the Sharing Tag. The calculation formula of Sharing Tag is shown in Formula 1, where Name N Indicates the Nth name in the CS storage space. CRC32(Name1) indicates that Name1 is hashed using the CRC32 hash function. When the edge server receives the interest packet, it checks the shared identification field and infers the network topology.

3. The NDN congestion control method in a heterogeneous network based on topology awareness according to claim 1, characterized in that: In the above 2), the specific method is: The calculation of path score involves transmission status indicators and time dimension indicators; the transmission status indicator is related to whether packet loss occurred in the previous round of transmission on the path, and the time dimension indicator is related to the transmission time of the interest packet; specifically, the transmission status indicator is represented by the custom Quality Score field in the interest packet, which is an integer field with a maximum value of Quality Score. max , the default value is The specific calculation method is shown in Formula 2: When the interest packet arrives at the edge server, the edge server records the arrival time, and calculates the delay Time of each interest packet based on the Send Time field value carried by the interest packet. The calculation formula is shown in Formula 3: Time = Arriving Time - Send Time (Formula 3) In the collected Time of each Interest packet, determine the maximum time difference Time max and the minimum time difference Time min For the time dimension indicator, the maximum score is set to 1-δ and the minimum score is set to ε. The calculation formula for the time dimension score is shown in Formula 4. The link score Score is shown in Formula 5, where α is the score coefficient, 0<α<1; Assuming that N paths share the same link, the score of this abstract path is share The calculation formula is shown in Formula 6, where Score N Indicates the score of the Nth path:

4. The NDN congestion control method in a heterogeneous network based on topology awareness according to claim 1, characterized in that: In the above 3), the specific method is: The amount of data allocated to each path is adjusted according to the path quality; in order to represent the path quality, a scoring threshold Score is set t , when the path score reaches Score t When , it means that the path quality is good and the data packet can be transmitted as expected; for the data packet with sequence number DataSeq, in a network with PathNum non-shared paths, the calculation formula of its sending path sequence number PathSeq is shown in Formula 7; if the path score mapped by the data packet is less than Score t , then the data packet is subjected to secondary path mapping, and the final sending path sequence number PathSeq′ is calculated as shown in Formula 8; through secondary path mapping, the data packet will be sent to the new path; PathSeq=DataSeq%PathNum+1 (Formula 7) In order to enable the edge server to calculate the sending window of each path relatively accurately and ensure the low time complexity of the calculation process, this paper adopts a shallow neural network to predict and determine the sending window size of each path based on the status score of each path and the length of the sending queue.

5. The NDN congestion control method in heterogeneous networks based on topology awareness according to claim 1 is characterized in that: In the above 4), the specific method is as follows: a custom CS Tag field is added to the data packet; during multi-path transmission, there are three data packet cache states: (1) Priority caching: The CS Tag of the data packet transmitted on the low-scoring link should be the priority caching tag. The CS Tag prefix of the priority caching data packet is "11". The initial caching probability is A. Let the lower limit of the probability change be C and the upper limit of the probability change be D. Then, when the data packet passes through a router, the caching probability is P cache , the router will generate a random number E between 1 and 100, then the next route cache probability P′ cache The formula is shown in Formula 9, where Random(C,D) represents the generation of a random number in the interval [C,D]. If the router does not cache the data packet, P′ cache Convert to 7-bit binary, fill the high-order bits with 0 if the number is less than 7, and then add the "11" prefix to indicate the priority cache type; If the router has cached the data packet, the CS Tag will be set to all 0s, indicating that the data packet has been cached and the subsequent router will not cache it again. (2) Normal caching: The CS Tag of the data packet transmitted on the high-scoring link should be the normal cache tag. The CS Tag prefix of the normal cache data packet is "10". The initial cache probability is B. Let the lower limit of the probability change be C' and the upper limit of the probability change be D'. Then, when the data packet passes through a router, the cache probability is P cache , the router will generate a random number E in the range of 1-100, and then the next route cache probability P′ cache The formula is shown in Formula 10, where Random(C′,D′) represents the generation of a random number in the interval [C′,D′]. If the router does not cache the data packet, P′ cache Convert to 7-bit binary, fill the high-order bits with 0 if the number is less than 7, and then add a "10" prefix to indicate the normal cache type; If the router has cached the data packet, the CS Tag will be set to all 0s, indicating that the data packet has been cached and the subsequent router will not cache it again. (3) Cached: This indicates that the data packet has been cached by an NDN router and its CS Tag is all zeros. To fully utilize cache resources and reduce cache redundancy in the network, cached data packets are no longer cached repeatedly in NDN routers. The NDN router is designed with a CS occupation threshold CS t and time threshold threshold; When the time interval between the last CS clearing of an NDN router (timer T cs ) exceeds the threshold, and the remaining CS space of this node is monitored. free When it is less than the threshold, the NDN router will send a popularity request table to the edge server, that is, both conditions of Formula 11 are met at the same time; After receiving the request, the edge server returns the first N records in the popularity table. The NDN router that receives the popularity table forwards the table to each connected router. The NDN router that receives the popularity table checks the contents of its cache. If a packet appears in the popularity table, it retains the packet; otherwise, it removes it. The edge server and the user use the same method to calculate the mapping relationship between data packets and path numbers; when the user misses a data packet, the path number is mapped according to the data packet number and a data request is sent to the specified path to quickly recover from the transmission abnormality caused by the data loss.

Images