An efficient edge caching method based on NDN

Abstract

The application discloses a high-efficiency edge caching method based on NDN and belongs to the technical field of edge caching in computer networks. The method relates to a scenario that a user uploads data to an edge server, and constructs an edge caching routing table and a whole-network caching cost table. When the user wants to upload data to the edge server, firstly, an edge server is determined as a starting point of caching according to the routing cost table; and on the basis, the edge server selects an edge server in the network as a terminal point of caching according to the edge caching routing table and taking the edge server itself as the starting point. Through one-time bidirectional communication between the terminal point edge server and the user, the starting point edge server and the terminal point edge server can both cache the content uploaded by the user, so that distributed caching of the data content is realized. The method is suitable for the fields of network communication and the like, can effectively improve caching efficiency and reduce the response time when the user requests data from the edge server.

Description

TECHNICAL FIELD

[0001] The present application relates to an efficient edge caching method based on NDN, belonging to the technical field of edge caching in computer networks. BACKGROUND

[0002] Edge caching refers to caching data to the network edge close to users or data sources to provide edge intelligent services for users on the spot, meeting the key needs of digital industries in agile connection, real-time business, data optimization, application intelligence, security and privacy protection, etc. Due to the existence of edge servers, it is not necessary to transmit all data to the data center, which can effectively save network bandwidth; at the same time, the widely distributed edge servers can realize distributed caching of data and improve the fault tolerance of the communication system.

[0003] At present, the implementation of edge caching is as follows: when a user uploads or downloads data to an edge server, the IP address of the edge server needs to be obtained through a center server first, when the user sends a request for uploading or downloading data, the network transmits the uploading data or downloading data request to the edge server, the edge server sends a request response, the user and the edge server establish an end-to-end connection, and the user uploads or downloads data to the edge server. At present, edge caching has the following deficiencies in selecting cache nodes and distributed caching data:

[0004] (1) In the selection of cache nodes, when a user uploads data, it is usually necessary to rely on a center server to select a suitable edge server for data caching, and when a user downloads data, it is also necessary to rely on a center server to locate an edge server containing cached data. The mapping relationship of the edge server is managed by the center server, and the terminal user only analyzes the cached content to be uploaded or downloaded, determines the edge server address corresponding to the cached content, and sends a request for obtaining data or uploading data. This solution not only causes additional network overhead and increases service response delay, but also faces single point failure.

[0005] (2) In the aspect of distributed caching data, since edge caching often requires the same data to be backed up in multiple edge servers, when the data in a certain edge server is damaged, it can be recovered through data backup in other edge servers. When a user uploads data to multiple edge servers to realize distributed caching, a separate communication link needs to be established each time data is uploaded, which greatly increases the network overhead of the communication system and reduces the caching efficiency.

[0006] The root of the above-mentioned deficiencies is that the current edge cache adopts a network architecture based on TCP / IP, uses an IP address-based addressing mode and an end-to-end data pushing manner, i.e., a host-centered communication mode. This host-centered and location-addressed communication mode has low communication efficiency and cannot support complex edge cache service requirements, and cannot efficiently implement cache node selection and distributed cache of data.

[0007] As a future network architecture, the named data network (NDN) realizes a fundamental change from host-centered to content-centered by naming data instead of naming hosts. Due to its powerful forwarding plane, NDN has the potential to provide more efficient network support for edge computing than TCP / IP. First, NDN solves the problem of application name to IP address mapping by directly using application names in network layer forwarding, which simplifies the content discovery process and no longer requires a centralized entity. Second, its routing and forwarding are more intelligent than IP, and NDN can directly forward cache tasks to appropriate edge servers through naming and forwarding information. Third, NDN has application layer semantic information at the network layer, which can support dynamic deployment of services in a distributed manner. Fourth, in terms of caching, compared with the relatively centralized control of caching in IP networks, NDN caching is more conducive to saving network resources, improving user request response speed, and reducing the load of content source servers.

[0008] The in-network caching function of NDN mainly depends on the content cache table (CS) in the network node. In NDN, each node is equipped with a certain size of cache cost, and the CS caches all data packets transmitted via the node. When a node receives an interest packet request, it first searches in the CS, and if there is a corresponding data packet, the hit data packet is directly transmitted back to the requester. If the corresponding content cannot be found in the CS, further forwarding processing is performed according to the forwarding strategy, etc.

[0009] This mechanism of directly forwarding cache tasks to appropriate network nodes, with the network node cache data in the content cache table, can be used in the field of edge caching to achieve efficient cache node selection and distributed cache data. SUMMARY

[0010] In view of the technical defects of TCP / IP edge caching in the aspects of cache node selection and cache data sharing, i.e. "relying on a central server to select a proper edge server to cache data" and "large communication overhead in distributed storage of data", the main purpose of the present application is to provide an efficient edge caching method based on NDN, which takes into account the scenario of user uploading data to an edge server, and constructs an edge caching routing table and a network-wide cache cost table. When a user wants to upload data to an edge server, a routing cost table is used to determine an edge server as a starting point of caching; on this basis, the edge server also selects an edge server in the network as an end point of caching according to the edge caching routing table and taking itself as the starting point. Through one-time two-way communication between the end point edge server and the user, the starting point edge server and the end point edge server can both cache the content uploaded by the user, so as to realize distributed caching of data content. The present application can effectively improve the caching efficiency and reduce the response time when a user requests data from an edge server.

[0011] The purpose of the present application is realized by the following technical solutions:

[0012] The present application discloses an efficient edge caching method based on NDN, comprising the following steps:

[0013] Step 1, setting a network center server and notifying the whole network, specifically comprising the following sub-steps:

[0014] Step 1.1, determining the position of the center server in the network;

[0015] The center server functions to maintain the global topology of the network, maintain the cache cost table of the edge servers in the network, and has the routing function in the process of communication;

[0016] Step 1.2, the center server broadcasts an authentication interest packet to all edge servers and routers in the whole network, all nodes receiving the authentication interest packet add a routing entry to the center server in their forwarding information table, and return a data packet as an acknowledgement, so as to ensure that all nodes in the network except the center server are reachable to the center server;

[0017] Step 2, all nodes in the whole network establish a local link state database and send a link state interest packet to the center server, and the edge servers cache the link state interest packet to the center server. Specifically comprising the following sub-steps:

[0018] Step 2.1, each two adjacent nodes in the network send a hello interest packet to each other once, so as to determine the names of adjacent nodes, link states and communication overheads in the network.

[0019] The nodes include routers and edge servers;

[0020] The adjacent nodes are two nodes which do not need to pass through another router or edge server for communication;

[0021] The content in the hello interest packet is "the name of the node, the routing cost of sending data to the node and the adjacent segment identification";

[0022] Step 2.2. Each communication node in the network records the name of all adjacent nodes of the node, the routing cost of the corresponding link and the adjacent segment identification of the link according to all received hello interest packets, and forms a link state database of the node;

[0023] The content of the link state database includes the name of all adjacent nodes of the node, the routing cost of the corresponding link and the adjacent segment identification;

[0024] Step 2.3. All nodes in the network send link state interest packets to the center server; and the edge server sends a cache state interest packet to the center server;

[0025] The content of the link state interest packet is "the name of the node, the name of all adjacent routers of the node, the routing cost of the corresponding link and the adjacent segment identification", that is, all information of the link state database of the local router; and the content of the cache state interest packet is the cache space available to the edge server.

[0026] Step 3. The center server establishes a full-network topology graph and a full-network cache capacity table, sends the full-network topology graph to all routers and edge servers in the network, and distributes the full-network cache cost table to all edge servers. Specifically, the following sub-steps are included:

[0027] Step 3.1. The center server aggregates all received link state interest packets to establish a full-network topology graph;

[0028] The content of the full-network topology graph includes "the name of all nodes in the network, the routing cost of the corresponding link and the adjacent segment identification";

[0029] Step 3.2. The center server aggregates all received cache state interest packets to maintain a full-network cache cost table;

[0030] The full-network cache cost table is obtained by transforming the available cache capacity of each edge server in the network. The cache cost is negatively correlated with the cache capacity, that is, the larger the cache capacity, the smaller the cache cost.

[0031] Step 3.3. The center server encapsulates a topology data packet with the content of "the full-network topology graph" and a cache cost data packet with the content of "the full-network cache cost table". The encapsulated topology data packet is broadcasted to all routers and edge servers in the network, and the encapsulated cache cost data packet is sent to all edge servers.

[0032] Step 3.4 After receiving the topology data packet, the full-network router and the edge server obtain the full-network topology graph and save it. After receiving the cache cost data packet, the full-network edge server obtains the full-network cache cost table and saves it;

[0033] The steps 3.3 and 3.4 are used to ensure that all routers and edge servers in the network can obtain the full-network topology graph, and the edge servers can obtain the full-network cache cost table;

[0034] Step 4, each router maintains its forwarding information table according to the full-network topology graph, and the edge server establishes an edge cache routing table. The specific steps include the following sub-steps:

[0035] Step 4.1 The router calculates its forwarding information table according to the full-network topology graph and link routing cost using the Dijkstra optimal path algorithm. The forwarding information table includes the name of the target router or edge server, the next interface, and the routing cost C r .

[0036] Step 4.2 All edge servers maintain an edge cache routing table for subsequent cache node selection.

[0037] The "edge cache routing table" includes "target edge computing server name, all paths to the edge computing server, corresponding communication cost, and adjacent segment identifier corresponding to the outgoing interface";

[0038] The communication cost C c is obtained by calculating the routing cost C r and the cache cost C s and C c =C r +C s ; the adjacent segment identifier corresponding to the outgoing interface is obtained from the link state database of the node;

[0039] There are edge servers in the network that can cache data, so there are multiple forwarding interfaces in the edge cache routing table. The routing cost of each edge cache server forwarding interface must be declared;

[0040] The routing cost C r is calculated according to the node connection and link routing cost in the full-network topology graph; the cache cost C s is obtained by adding the cache costs of all nodes on the routing path according to the full-network cache cost table calculated by the center server; the edge cache routing table includes all routing entries starting from the edge server.

[0041] Step 5, the user sends a cache announcement interest packet to the edge server, the interest packet is forwarded in the network via the router, reaches the edge server according to the shortest path algorithm. The first arriving edge server combines the communication cost in the edge cache routing table and the hop number to the target node to select the most suitable cache end edge server, and adds the adjacent segment identifier to the interest packet, forwards the interest packet to the end edge server according to the specific path, and the end edge server returns a cache response data packet. Specifically, the following sub-steps are included:

[0042] Step 5.1 The user sends a cache announcement interest packet, and the router receives the interest packet and queries according to the forwarding information table, and forwards the cache announcement interest packet to the edge server with the smallest communication cost.

[0043] The "cache announcement interest packet" is used to send an announcement to the edge server that the user wants to upload data. The content of the "cache announcement interest packet" includes the size, name and node name of the parameter file to be uploaded.

[0044] Step 5.2 After the edge server receives the cache announcement interest packet, it combines the communication cost in the edge cache routing table to select the target node with the lowest communication cost of three hops or more as the end edge server, and adds the adjacent segment identifier information in the local network topology diagram to the interest packet. The edge server forwards the cache announcement interest packet to the end edge server along the specific path according to the adjacent segment identifier information.

[0045] The edge server can also be called the cache start edge server. The target node corresponding to the routing entry selected by the edge server can also be called the end edge server.

[0046] The number of hops refers to the number of edge servers the user passes through to reach the target edge server. The more the number of hops, the more edge servers can cache data in one communication, and the higher the cache efficiency.

[0047] Step 5.3 The end edge server receives the interest packet and feeds back a "cache announcement data packet" for receiving confirmation, indicating that the end edge server has established a connection with the user.

[0048] Step 6, the end edge server sends a specific number of cache task interest packets to the user according to the "file size to be uploaded" information in the cache announcement interest packet, distinguishes different users according to the "user node name", and records the "name of the parameter file to be uploaded" in the PIT table. The user returns the cache task data packet along the reverse path of the interest packet, and the edge servers in the middle cache the task data packet based on the LCE strategy and update their own cache according to the LRU strategy;

[0049] The ''cache task interest package'' is used to obtain the data of the user, and the user returns the ''cache content data package'' to the terminal edge server along the reverse path forwarded by the cache task interest package;

[0050] The LCE (Leave Copy Everywhere) strategy refers to caching a copy of the content at each node in the data return path.

[0051] The LRU (Least Recently Used) strategy refers to a cache replacement strategy using the most recently cached content as a reference, and the least recently used object in the cache queue is deleted from the cache to make room for the newly arrived content.

[0052] Step 7, the edge server sends a cache state interest package to the center server, the center server updates the global cache cost table and sends a cache cost data package to each edge server node, and the edge server updates the edge cache routing table. The efficient utilization of the edge cache based on NDN is realized, the pressure of the center server and the network overhead of the communication system are reduced, the cache efficiency is improved, and the response time of the user to the edge server for requesting data is reduced.

[0053] The ''cache state interest package'' includes the available cache capacity of the edge server.

[0054] Step 8, according to the content cached by the edge server in step 6, the data acquisition delay and cache hit rate can be verified, whether the performance requirements of low delay and high cache hit rate in the edge cache field can be realized can be detected. In addition, this scheme can be applied to complex networks with limited network bandwidth and high dynamic topology, and the actual problems of low cache efficiency and long server response time can be solved.

[0055] Advantages:

[0056] The disclosed efficient edge cache method based on NDN has the following advantages compared with the existing edge cache method based on TCP / IP:

[0057] 1. Compared with the proxy edge cache solution based on TCP / IP, in the selection of cache nodes, when the user uploads data, the intelligent forwarding plane of NDN forwards the cache announcement interest package sent by the user to the edge server with the lowest routing cost, and then the edge server selects the cache node according to the edge cache routing table, without the need of the center server to select the cache node, thereby reducing the pressure of the center server and the network overhead of the communication system;

[0058] 2. Using the cache mechanism of NDN, only one bidirectional communication between the cache end edge server and the user can enable the edge nodes through which the data packet passes to cache the content source, realizing distributed cache of the user uploaded content and improving the cache efficiency of the edge server when caching data;

[0059] 3. Using the adjacent segment identification information of the link, adopting the segment routing forwarding mode, ensuring that the cache announcement interest packet is forwarded according to a specific path, ensuring that the selection of the cache node is only determined by the edge server with the lowest routing cost to the user, and avoiding the ambiguity of the selection of the cache node by different edge servers;

[0060] 4. When the user uploads large data, the data packet may be sent in fragments. Since the cache announcement interest packet contains the 'to-be-uploaded data size' information, and the edge server sends the cache task interest packet and already knows the total size of the data packet to be returned, therefore, under the condition that the size of each data packet is known, the number of interest packets can be controlled by calculation, and the balance between the number of interest packets and data packets in the system is realized. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 The application discloses a session model of edge cache based on NDN; the session model comprises the following steps: the edge server sends available cache space to the center server, the center server returns a whole network cache cost table, the user sends a cache announcement interest packet, the edge server selects and queries an edge cache routing table and selects an optimal server to cache user data, and the edge server acquires user uploaded data.

[0062] Figure 2 The application discloses a network topology and center server setting of an efficient edge cache method based on NDN, and a cache task interest packet and data packet flow direction diagram.

[0063] Figure 3 The application discloses a Dijkstra optimal path algorithm calculation flowchart used in step four in the establishment of the edge cache routing table based on NDN.

[0064] Figure 4 The application discloses a flowchart of an efficient edge cache method based on NDN. DETAILED DESCRIPTION

[0065] The application will be described in detail below with reference to the drawings and embodiments, and the technical problems solved by the application and the beneficial effects will be discussed. It should be pointed out that the described embodiments are intended to facilitate the understanding of the application and do not limit the application in any way.

[0066] The embodiment is based on an NDN network. After a user and an edge server establish a connection, the user sends a cache notification interest packet to a nearest edge server. The edge server selects a target node with the lowest communication cost as a terminal edge cache server according to an edge cache routing table. The terminal edge cache server sends a cache task interest packet to the user. The user returns a cache result data packet and caches the packet at a passing edge cache server. In the embodiment, a network topology, node settings and data flow are shown in Figure 1 A communication session model of uploading data of the user to the edge server is shown in Figure 2 In the embodiment, the names of the nodes in the network are “nodex (x = 1, 2, 3…)”, the nodes include routers and edge servers, the routers and the edge servers have routing functions, the user is user, and the central server is server. The content to be uploaded by the user is denoted as “videox (x = 1, 2, 3…)”. The embodiment discloses an efficient edge cache method based on NDN, which specifically includes the following steps.

[0067] Step I, setting a network central server and notifying the whole network, specifically including the following sub-steps.

[0068] Step I.1, selecting a central server in a network central area. In the embodiment, the node server is set as the network central server. The network topology and the setting of the central server are shown in Figure 1

[0069] Step I.2, the central server broadcasts an authentication interest packet with a name of “ / Server”. The content of the authentication interest packet is “the node server is the network central server”. All the routers and edge servers receiving the authentication interest packet add a routing entry with a name of “ / Server” in a forwarding information table, and the corresponding forwarding interface is the interface receiving the authentication interest packet.

[0070] Step II, establishing a local link state database by each router and edge server, and notifying the central server of the link state and cache cost, specifically including the following sub-steps.

[0071] Step II.1, each two adjacent edge servers or routers in the network send a hello interest packet to each other once. The name of the hello interest packet is “ / Hello / nodex / router” or “ / Hello / nodex / server” (nodex is the node name, x = 1, 2, 3…). Router represents that the current node is a router, and server represents that the current node is an edge server. The content of the hello interest packet is “the node name information, the routing cost of sending data to the node and the adjacent segment identifier of the link”.

[0072] ​Step II.2 Each router and edge server records all adjacent node names of the current node, corresponding link routing cost and adjacent segment identification of the link according to all received hello interest packets, to form a link state database of the edge server; in this embodiment, the edge server node1 receives hello interest packets from routers node2 and node4, and the link state database list of the router node1 is shown in Table 1.

[0073] Table 1 Link state database list of the router node1

[0074] Adjacent node name Forwarding interface Route cost of link node2 interface 1 100 node4 interface 2 110

[0075] Step II.3 All routers and edge servers except the center server in the network send link state interest packets to the center server, and the edge server also needs to send cache state interest packets to the center server, the name of the link state interest packet is “ / Server / nodex / linkstate”, and the name of the cache state interest packet is “ / Server / nodex / capacity” (nodex is the name of the current node, x = 1, 2, 3…), the content of the link state interest packet includes “the name of the current node, all information of the link state database of the current node”; the content of the cache state interest packet includes “the available cache capacity of the edge server”; according to the route entry named “ / Server” added in the forwarding information table of each router, the link state interest packet and the cache state interest packet are forwarded to the center server.

[0076] Step III, the center server establishes a full-network topology graph and distributes it to all nodes in the network, and establishes a full-network cache cost table and distributes it to all edge servers, which specifically includes the following sub-steps:

[0077] Step III.1 The center server aggregates all received link state interest packets to establish a full-network topology graph; in this embodiment, the data information of the full-network topology graph is shown in Table 2.

[0078] Table 2 Data information table of the full-network topology graph

[0079] Node name Adjacent node name Route cost of link node1 node2 110 node1 node4 100 node2 node3 100 node2 node5 120 node3 node6 100 node3 server 300 node4 node5 130 node4 node7 100 node5 node6 120 node5 node8 100 node6 node9 110 node6 server 320 node7 node8 100 node8 node9 130 node9 server 310

[0080] Step III.2 The center server aggregates all received cache state interest packets and establishes a full-network cache cost table; in this embodiment, the data information of the full-network cache cost table is shown in Table 3.

[0081] Table 3 Full-network cache cost table

[0082]

[0083]

[0084] Step III.3 The central server broadcasts a topology data packet with the name of " / NET" to all nodes in the network, and a cache cost data packet with the name of " / STORE" to all edge servers. The content of the topology data packet is "the whole network topology graph", and the content of the capacity data packet is "the whole network cache cost table";

[0085] Step III.4 After receiving the topology data packet, each router obtains and saves the whole network topology graph. After receiving the topology data packet and the capacity data packet, each edge server obtains and saves the whole network topology graph and the whole network cache cost table;

[0086] Step IV. Each router maintains its forwarding information table according to the whole network topology graph, and each edge server establishes an edge cache routing table according to the whole network topology graph and the whole network cache cost table;

[0087] Step IV.1 Each router calculates its forwarding information table by using the Dijkstra optimal path algorithm according to the whole network topology graph and the link routing cost. The forwarding information table includes the name of the target router or edge server, the next interface, and the routing cost C r .

[0088] The forwarding information table of the router node1 is shown in Table 4.

[0089] Table 4 Forwarding information table of the router node1

[0090] Destination node Forwarding interface Routing cost C r ]]> node2 interface 1 110 node3 interface 1 210 node4 interface 2 100 node5 interface 1 230 node6 interface 1 310 node7 interface 2 200 node8 interface 2 300 node9 interface 1 420 server interface 1 610

[0091] Step IV.2 Each edge server establishes an edge cache routing table according to the whole network topology graph and the whole network cache cost table.

[0092] Specifically, all edge servers in the network establish an edge cache routing table in the edge server. The edge cache routing table includes the name of the target edge computing server, all paths to the edge computing server, the corresponding communication cost, and the adjacent segment identifier corresponding to the outgoing interface. The node name is obtained from the whole network topology graph, the communication cost is calculated by the edge server according to the whole network topology graph and the whole network cache cost table, and the adjacent segment identifier is obtained from the local link state database. The edge cache routing table established by the edge server node5 in this embodiment is shown in Table 5.

[0093] Table 5 Edge cache routing table of the edge server node5

[0094]

[0095] Step V, the user sends a cache announcement interest packet to the edge server, the naming instance of the interest packet is " / Producer / user / video". The interest packet reaches the edge server with the minimum communication cost, selects the most suitable destination edge server in combination with the communication cost in the edge cache routing table and the hop count to the target node, adds the adjacent segment identifier to the interest packet, forwards the interest packet to the target edge server according to the specific path, and the target edge server replies to the cache response data packet.

[0096] Step V.1 The user sends a cache announcement interest packet, which is forwarded to the router with a routing cost C r The lowest edge server.

[0097] Specifically, when the user sends a cache announcement interest packet, the user accesses the network from the router node1, and the node1 learns from the forwarding information table that the routing cost C r to the edge server node5 is the lowest, and therefore the cache announcement interest packet is first forwarded to the node5.

[0098] Step V.2 The edge server selects the optimal cache routing path according to the edge cache routing table, and adds the adjacent segment identifier to the interest packet, and the interest packet is forwarded in the network based on the adjacent segment identifier according to the determined path. Specifically,

[0099] The edge server selects the edge server with the lowest communication cost of three hops or more as the cache target node according to the communication cost and the hop count in the edge cache routing table.

[0100] In this embodiment, the target node node9 is the target node with the minimum communication cost of three hops or more from the edge server node5, and the specific path is node5→node8→node9. Therefore, the edge server node5 adds the adjacent segment identifiers 803 and 902 in front of the cache announcement interest packet, so that the cache announcement interest packet is forwarded along the path node5→node8→node9 to the edge server node9, and the edge server node9 sends a cache response data packet to the user as feedback.

[0101] Step VI, edge server node9 sends cache task interest packet to edge server node5 along node9→node8→node5, the naming instance of cache task interest packet is " / video". Edge server node5 forwards a certain number of cache task interest packets to the user along node5→node4→node1 according to "file node name" user, "file size" 32MB in the cache announcement interest packet and the effective data payload of the data packet in the communication system. The user returns the cache task data packet along node1→node4→node5→node8→node9. Edge servers node5, node8 and node9 cache the task data packet based on the LCE cache strategy and update the available cache space of the edge server according to the LRU strategy.

[0102] The number of cache task interest packets sent by the edge server can be calculated from the "size of the data file to be uploaded" in the cache announcement interest packet. Assuming that the size of the data file uploaded by the user is 32MB and the effective data payload of each data packet is 1MB, the edge server will send 32 cache task interest packets to obtain the data to be uploaded by the user.

[0103] Step VII, edge servers node5, node8 and node9 send cache status interest packets to the center server, the center server updates the global cache cost table and sends the cache cost data packet to each edge server node, and the edge server updates the edge cache routing table of itself.

[0104] Thus, the introduction and use process of the edge cache scheme based on NDN is completed.

[0105] The system session model involved in the embodiment is shown in Figure 2 which includes that the edge server obtains the global cache cost table, the user sends the cache passing interest packet, the edge server queries the edge cache routing table and determines the terminal edge cache server, the terminal edge cache server sends the cache task interest packet request, the user returns the task data packet to be cached, the edge servers on the way cache the data packet and send the cache status interest packet to the center server to update the available cache space of itself.

[0106] The various naming instances of interest packets used in the embodiment are shown in Table 6; the various naming instances of data packets are shown in Table 7;

[0107] Table 6 Various naming instances of interest packets

[0108]

[0109]

[0110] Table 7. Various data packet naming examples

[0111] Packet type Sender Receiver Named instance Topology packet Central server Network node / NET Cache cost packet Central server Edge server / STORE Cache task packet User End edge server / video

[0112] The embodiment describes an NDN-based edge cache session model and a use method, which on the one hand improves the time consumption and network resource waste problem caused by the current edge cache needing to rely on a central server to select an edge cache server; on the other hand, the method only needs a one-way communication between the cache end edge server and the user to enable all edge servers through which the data packet passes to cache the content source, realizes distributed cache of the user uploaded content, and improves the cache efficiency when the edge server caches data. In addition, through distributed cache, when other users request resources from the edge server, the response efficiency of the user can be effectively improved.

[0113] The method designs an edge cache routing table and a network-wide cache cost table. When a user wants to upload data to an edge server, an edge server will be determined as the starting point of cache according to the routing cost. On this basis, the edge server will select an edge server in the network as the cache end point according to the communication cost and the number of hops based on the edge cache routing table and the network-wide cache cost table. The embodiment realizes the user uploading data to the edge server and the distributed cache of the user uploaded data by the end edge server sending a cache task interest packet to the user, the user returning a cache task data packet and caching the data packet in the edge server, and finally realizes the user uploading data to the edge server and the distributed cache of the user uploaded data by the end edge server sending a cache task interest packet to the user, the user returning a cache task data packet and caching the data packet in the edge server; the size of the data to be uploaded is defined through the cache advertisement interest packet, and then the number of cache task interest packets sent by the edge server when obtaining the user data is controlled, realizing the balance of the number of interest packets and data packets in the network; the edge cache routing table is established by the edge server, realizing the selection of the most suitable edge server from multiple edge cache servers to cache data. The system can reduce the data acquisition delay of distributed cache service, and is particularly suitable for high-difficulty and algorithm-complexity edge cache schemes.

[0114] The above specific description further details the purpose, technical scheme and beneficial effects of the application. It should be understood that the above description is only a specific embodiment of the application and is not used to limit the protection scope of the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. An efficient edge caching method based on NDN, characterized in that: It comprises the following steps, Step 1, setting a network center server and announcing the whole network; Step 2, all nodes in the whole network establish a local link state database and send a link state interest package to the center server, and the edge server sends a state interest package to the center server; Step 3, the center server establishes a whole network topology structure diagram and a whole network cache capacity table, sends the whole network topology structure diagram to all routers and edge servers in the network, and distributes the whole network cache cost table to all edge servers; Step 4, each router maintains its own forwarding information table according to the whole network topology structure diagram, and the edge server establishes an edge cache routing table; Step 5, a user sends a cache announcement interest package to the edge server, the interest package is forwarded in the network through the router, reaches the edge server according to the shortest path algorithm, the first arriving edge server selects the most suitable cache end edge server according to the communication cost in the edge cache routing table and the hop number to the target node, adds an adjacent segment identifier to the interest package, forwards the interest package to the end edge server according to a specific path, and the end edge server returns a cache response data package; Step 6, the end edge server sends a specific number of cache task interest packages to the user according to the "to-be-uploaded file size" information in the cache announcement interest package, distinguishes different users according to "user node name", and records "to-be-uploaded parameter file name" in the PIT table; the user returns the cache task data package along the reverse path of the interest package, and the edge servers passing through in the middle cache the task data package based on the LCE strategy, and update the cache of themselves according to the LRU strategy; Step 7, the edge server sends a cache state interest package to the center server, the center server updates the whole network cache cost table and sends the cache cost data package to each edge server node, and the edge server updates the edge cache routing table; the efficient utilization of the edge cache based on NDN is realized, the pressure of the center server and the network overhead of the communication system are reduced, the cache efficiency is improved, and the response time of the user to request data from the edge server is reduced.

2. The method of claim 1, wherein: It also comprises step 8, verifying the data acquisition delay and cache hit rate according to the content cached by the edge server in step 6, and detecting whether the performance requirements of low delay and high cache hit rate in the edge cache field can be realized. 3.The method of claim 1, wherein: It is applied to a complex network with limited network bandwidth and high dynamic topology, and solves the technical problems of low cache efficiency and long server response time.

4. The efficient edge caching method based on NDN of claim 1, 2 or 3, characterized in that: The implementation method of step 1 is, Step 1.1, determining the position of the center server in the network; The function of the center server is to maintain the global topology of the network, maintain the cache cost table of the edge server in the network, and have the function of routing in the process of communication; Step 1.2, the center server broadcasts an authentication interest package to all edge servers and routers in the whole network, all nodes receiving the authentication interest package add a routing entry to the center server in the forwarding information table, and return a data package as confirmation, so that all nodes in the network except the center server are reachable to the center server.

5. The method of claim 4, wherein: The implementation method of step 2 is, Step 2.1 Each two adjacent nodes in the network send a hello interest packet to each other to determine the names of the adjacent nodes, the link status and the communication cost in the network; The nodes include routers and edge servers; The adjacent nodes are two nodes that do not need to pass through another router or edge server for communication; The content in the hello interest packet is "the name of the node, the routing cost for sending data to the node and the adjacent segment identifier"; Step 2.2 Each communication node in the network records the names of all adjacent nodes of the node, the routing cost of the corresponding link and the adjacent segment identifier of the link according to all the received hello interest packets to form a link status database of the node; The content of the link status database includes the names of all adjacent nodes of the node, the routing cost of the corresponding link and the adjacent segment identifier; Step 2.3 All nodes in the network send a link status interest packet to the center server, and the edge server sends a cache status interest packet to the center server; The content of the link status interest packet is "the name of the node, the names of all adjacent routers of the node, the routing cost of the corresponding link and the adjacent segment identifier", that is, all information of the link status database of the local router; and the content of the cache status interest packet is the available cache space of the edge server.

6. The method of claim 5, wherein: The implementation method of step 3 is as follows: Step 3.1 The center server aggregates all the received link status interest packets to establish a network topology diagram; The content of the network topology diagram includes "the names of all nodes in the network, the routing cost of the corresponding link and the adjacent segment identifier"; Step 3.2 The center server aggregates all the received cache status interest packets to maintain a network cache cost table; The network cache cost table is obtained by transforming the available cache capacities of the edge servers in the network; the cache cost is negatively correlated with the cache capacity, that is, the larger the cache capacity, the smaller the cache cost; Step 3.3 The center server encapsulates a topology data packet with the content of "the network topology diagram" and a cache cost data packet with the content of "the network cache cost table"; and broadcasts the encapsulated topology data packet to all routers and edge servers in the network, and sends the encapsulated cache cost data packet to all edge servers; Step 3.4 After receiving the topology data packet, all routers and edge servers in the network obtain and save the network topology diagram; and after receiving the cache cost data packet, all edge servers obtain and save the network cache cost table; Steps 3.3 and 3.4 are used to ensure that all routers and edge servers in the network can obtain the network topology diagram, and the edge servers can obtain the network cache cost table.

7. The method of claim 6, wherein: The implementation method of step 4 is as follows: Step 4. The step 4.1 router calculates its own forwarding information table according to the whole network topology graph and link routing cost, using Dijkstra optimal path algorithm, and the forwarding information table contains the name of target router or edge server, next interface and routing cost C r ; Step 4.2 All edge servers maintain an edge cache routing table for subsequent cache node selection; The "edge cache routing table" includes "the name of the target edge computing server, all paths to the edge computing server, the corresponding communication cost and the adjacent segment identifier corresponding to the outgoing interface"; The communication cost C c By calculating the routing cost C r And the cache cost C s And C c =C r +C s ; the adjacent segment identifier corresponding to the out-interface is obtained from the link state database of the node; There are edge servers in the network that can cache data, so there are multiple forwarding interfaces in the edge cache routing table, and the forwarding interface corresponding to each edge cache server must declare its routing cost; The computing routing cost C r , which is calculated according to the connection of nodes and the link routing cost in the whole network topology diagram; the cache cost C s , which is obtained by superimposing the cache costs of all nodes in the routing path according to the whole network cache cost table calculated by the center server; and the edge cache routing table including all routing entries with the edge server as a starting point.

8. The method of claim 7, wherein: The implementation method of step 5 is, Step 5.1 The user sends a cache announcement interest packet, and the router receives the interest packet and queries according to the forwarding information table, and forwards the cache announcement interest packet to the edge server with the smallest communication overhead; The "cache announcement interest packet" is used to send an announcement to the edge server that the user wants to upload data; The content of the "cache announcement interest packet" includes the size, name and node name of the parameter file to be uploaded; Step 5.2 After the edge server receives the cache announcement interest packet, it selects the target node with the lowest communication cost of three hops or more in the edge cache routing table of the edge server itself as the end edge server, and adds the adjacent segment identification information in the local network topology graph to the interest packet; The edge server forwards the cache announcement interest packet to the end edge server along the specific path according to the adjacent segment identification information; The edge server can also be called the cache starting edge server; The target node corresponding to the routing entry selected by the edge server can also be called the end edge server; The number of hops refers to the number of edge servers the user passes through to reach the target edge server; The more the number of hops, the more edge servers that can cache data in one communication, and the higher the cache efficiency; Step 5.3 After the end edge server receives the interest packet, it feeds back a "cache announcement data packet" as a receiving confirmation, indicating that the end edge server has established a connection with the user.

9. The efficient edge caching method based on NDN of claim 8, wherein: The "cache task interest packet" is used to obtain the user's data, and the user returns the "cache content data packet" to the end edge server along the reverse path of the cache task interest packet forwarding; The LCE strategy means that each node will cache a copy of the content in the data return path; The LRU strategy refers to a cache replacement strategy that uses the most recently cached content as a reference to delete the least recently used object in the cache queue from the cache to make room for new content. 10.The method of claim 9, wherein: The "cache status interest packet" contains the available cache capacity of the edge server.

Images