WO2021051431A1

WO2021051431A1 - Forwarding method for network node in named data network, device, system, and storage medium

Info

Publication number: WO2021051431A1
Application number: PCT/CN2019/107695
Authority: WO
Inventors: 李挥; 胡嘉伟; 邬江兴; 黄婷; 伊鹏; 马化军; 尹峰
Original assignee: 北京大学深圳研究生院; 国家数字交换系统工程技术研究中心; 佛山赛思禅科技有限公司
Priority date: 2019-09-18
Filing date: 2019-09-25
Publication date: 2021-03-25
Also published as: CN110417661A

Abstract

Provided are a forwarding method for a network node in a named data network, a device, a system, and a storage medium. The forwarding method comprises a FIB comprising a hash table and a prefix tree, and each name stored in the table has a proper prefix having a corresponding table entry in the table. A process of checking whether a prefix exists and adding a corresponding auxiliary table entry is called FIB reconfiguration. Table entries in a reconfigured FIB are classified into real table entries and non-real table entries, and the non-real table entries are classified into virtual table entries and semi-virtual table entries. The present invention provides an advantageous effect of implementing a FIB forwarding framework and a related algorithm that support random search, and resolves issues relating to backtracking and outdated entries. As shown by experimental evaluation, the present invention has almost no influence on time overheads of algorithms, thereby ensuring the efficiency and superior performance of random search algorithms. When used for designing a highly efficient NDN forwarding framework, the invention lays the foundations for fundamentally solving the scalability problem of NDNs.

Description

A forwarding method, device, system and storage medium for naming network nodes in a data network

Technical field

The present invention relates to the field of Internet technology, and in particular to a forwarding method, device, system and storage medium for naming network nodes in a data network.

Background technique

Information Central Networking (ICN), including Content Central Networking (CCN) and Named Data Networking (NDN), is a new type of Internet architecture that aims to solve the bottleneck of the current IP protocol. In the present invention, the NDN branch is mainly discussed. In the following, NDN will be referred to as the content center network. In NDN, the network cares more about the data itself rather than the location of the data. Through the decoupling of content and specific locations, as well as the design of appropriate forwarding and caching strategies, NDN realizes the multi-path transmission of data through an intelligent forwarding plane, and reduces the data request delay through the cache in the network, thereby achieving Efficient and reliable information distribution.

As its name suggests, NDN uses hierarchical names to refer to every resource in the network, and designs a name-based content request and forwarding mechanism. The NDN data transmission process relies on two types of data packets, namely, Interest Packet and Data Packet. NDN uses a consumer-driven approach to complete its content request: the consumer adds the name of the requested content to the interest package, and the interest package is posted to the network; after the routing node receives the interest package, it will be at this node In the forwarding information table (Forwarding Information Base, FIB), the content name in the interest packet is queried based on the Longest Prefix Matching (LPM) method to obtain the port for forwarding the packet; when the interest packet is When the holder of the content accepts it, the holder will generate a data packet of the content and trace it back to the consumer along the path of the interest packet.

NDN changes the message forwarding mechanism from address-oriented to name-oriented. In this way, NDN realizes the decoupling of content and location, and improves the flexibility and efficiency of the network. However, the name-oriented forwarding model also has its shortcomings: With the expansion of the network scale, the performance of the NDN forwarding plane tends to suffer a serious decline, including the rapid increase of forwarding delay, message congestion, and a large amount of memory. and so on. These problems are roughly caused by the following reasons:

1. The name-oriented forwarding plane uses the name as the search key, and the NDN name has unbounded length and variable styles. Therefore, various optimization algorithms based on regular IP addresses in the past are often difficult to apply to it; at the same time, the traditional linear The search algorithm has a time cost proportional to the expected length of the name, and it is difficult to achieve better results for NDN names that are much longer than the IP address. The above reasons lead to a large delay in the query process.

2. Because the objects referred to are more subtle, the number of NDN names is much larger than that of IP addresses, resulting in NDN FIBs often having an extremely large scale. Considering that the key value of FIB is an unbounded NDN name, when an appropriate compression strategy is not adopted, its memory footprint will far exceed the available capacity of existing equipment.

Compared with stable IP addresses, NDN names are extremely prone to change, resulting in the FIB in NDN facing much more frequent update requirements. For this reason, FIB design must support the efficiency of various operations to maintain its own real-time and effective.

For the above reasons, NDN is often difficult to apply in large-scale network practice, which is called the scalability problem of NDN. Among them, the name-oriented forwarding mechanism is the key point that caused this problem and started to solve the problem. Excellent forwarding plane design can ensure the efficiency and reliability of NDN network in large-scale applications, which has become the main focus of existing NDN researchers. One of the directions.

Prefix Tree (Trie), also known as dictionary tree, is a data structure commonly used in string matching. In the dictionary tree, an edge refers to a segmented component (bits, characters...) of a name, and a node refers to a specific name, and the content of the name is the combination of components from the root to all edges on the path of the node. In the storage structure based on the prefix tree, the prefix parts that are the same between the names are merged into the upstream path, thereby achieving the compression of the data capacity and the preservation of the logical relationship between the names.

Since the prefix tree naturally supports the LPM algorithm and has good space utilization efficiency, most NDN implementations use the prefix tree for the design of its forwarding plane. The disadvantage of the prefix tree is its slow query speed: its computational time cost is roughly proportional to the expected length of the name. At the same time, in each layer, the query algorithm needs to match all the outgoing edges of the node one by one to find the The descending child node. Therefore, the NDN forwarding architecture based on the prefix tree will cause a large query delay and affect the overall performance of the network.

Compared with the prefix tree, the query speed of the hash table will not be affected by the change of the name length and the size of the table entry, so it has better adaptability to the large-scale network environment. However, in order to deal with hash collisions, the hash table needs to also store the complete key (here, the content name) in the table entry, which causes a large storage overhead. At the same time, the original hash table structure does not support the longest prefix matching algorithm, and the simplest linear implementation has a larger query time. In order to cope with these problems, the existing NDN usually uses data compression schemes such as Footprint-Based Hash Table, and algorithm optimization schemes such as Random Search to improve the scalability of the system.

BACKGROUND OF THE INVENTION 1. In the 2013 article "Named data network on a router: Fast and DoS-resistant forwarding with hash tables", Cisco proposed a forwarding algorithm and architecture that can effectively respond to DoS attacks. The team designed a two-stage (2-stage) FIB search method, which effectively improves the efficiency of the algorithm when the query is missed, thereby enhancing the system's anti-attack ability. The brief flow of the algorithm is as follows:

1) When each name is inserted into the FIB, the forwarding plane will query the table for a prefix of length M. If the prefix does not exist in the table, it will insert the virtual entry corresponding to the prefix. The virtual table entry can only be used to support the algorithm flow of random search, and cannot be used to guide the forwarding of interest packets.

2) As shown in Figure 1, a 2-stage LPM query starts with a prefix of length M, and only when the query hits, will it enter the linear search process for the complete name. Under normal circumstances, M takes a smaller value (2 or 3), which can speed up the query for invalid prefixes, thereby improving the resistance to DoS attacks.

Background Art One disadvantage: In the case of a query hit, this method relies on a more time-consuming linear search. Since the time cost of a linear search is proportional to the expected length of the query string, in the face of NDN names with no upper bound in length, The algorithm still has a large space to be optimized.

Background Art 2: In the article "Reliably Scalable Name Prefix Lookup" by the University of Washington, Haowei Yuan equals to 2015, a FIB acceleration scheme based on binary search is proposed, which is summarized as follows:

1) FIB contains a series of hash tables numbered starting from 1, where the table numbered i only stores names with length i. For each name in the table, all its true prefixes must have corresponding entries in the FIB. If the corresponding entries do not exist, the corresponding virtual entries are inserted as padding.

2) As shown in Figure 2, in each iteration, the search range [L, H] is faced, and the algorithm will select the middle point (L+H)/2 of the range as the next search object, and query according to it The result determines the search range for the next iteration. When the search range is reduced to 0, the algorithm terminates.

Through binary search, the relationship between the time cost of the algorithm and the length of the name is reduced from linear to logarithmic, thereby greatly improving the computational efficiency of the forwarding plane.

The disadvantage of the second background technology: the above-mentioned binary query scheme and many similar random search algorithms. There are so-called backtracking problems (Backtracking Problem) and outdated entry problems (Outdated Entry Problem).

Background Art Three, Yi Wang of Tsinghua University proposed a series of random search algorithms based on statistical distribution in articles such as "Statistical Optimal Hash-based Longest Prefix Match". Different from the binary query, in its design scheme, the query length of each iteration is calculated by a priori statistical distribution, rather than simply selecting the center point of the range to be queried. As a result, the solution shortens the expected length of the query path, thereby improving the computational efficiency of the forwarding platform.

Disadvantages of background technology three: similar to background technology two, there are so-called backtracking problems (backtracking problems) and outdated entry problems (outdated entry problems).

Backtracking problem: In the above binary query process, when the algorithm body is executed, if the prefix of the last HIT is a virtual table entry, FIB cannot determine whether there is forwarding information corresponding to the name to be queried in the table. For this reason, the algorithm needs Go to the prefix of the last HIT and re-execute the linear search. For example, if there are names "/c1" and "/c1/c2/c3/c4" in the table, then "/c1/c2/" and "/c1/c2/c3" will be inserted with

As a virtual table entry. At this time, the 2-point query process for the name "/c1/c2/c3/c5" is:

Since all HIT prefixes on the query path are virtual, if the algorithm ends directly after the binary query, it will return the query failure. However, there is its LPM "/c1" in the table, resulting in a false negative (False Negative) error. In the traditional IP protocol, the forwarding algorithm solves this problem through a backtracking search, which is to go to the prefix of the last HIT (here "/c1/c2/c3") and perform a linear search. However, as described above , The name in the content network is unbounded and much longer than the length of the IP address, so this method is difficult to apply to NDN implementation.

Outdated entries problem: In the existing random search scheme, outdated virtual entries are difficult to delete in time. In the above example, if "/c1/c2/c3/c4" is deleted, then "/c1/c2/ "And "/c1/c2/c3" have lost their necessity and should be deleted in time to save storage costs. However, due to the diversity of NDN identification, unless the entire table is traversed, it is difficult for FIB to determine whether an auxiliary table entry has other entries inherited from it in the table. Therefore, the accumulation of outdated auxiliary entries will also cause a large amount of memory waste and affect the overall performance of the network.

Summary of the invention

technical problem

The solution to the problem

Technical solutions

In order to solve the problem of backtracking in random search, the present invention provides a forwarding method for naming network nodes in a data network, which includes a hash table and a prefix tree. The FIB is composed of the hash table and the prefix tree. For any name, all its true prefixes have corresponding entries in FIB. The process of checking whether the prefix exists and adding corresponding auxiliary entries is called FIB reconstruction. In the reconstructed FIB, the entries are divided into Real entries and non-real entries, non-real entries are divided into virtual entries and semi-virtual entries;

In the hash table, the name is used as the key, and the node in the prefix tree is used as the value, thereby realizing fast retrieval of forwarding information from the name to the prefix tree;

Each edge in the prefix tree represents a name component, and each node of the prefix tree represents a name. The node stores the forwarding information corresponding to the name and the corresponding category of the table item, as well as the pointer used to maintain the prefix tree structure;

Real entry: The name in the real entry refers to the actual data and is used to guide the forwarding of the interest packet; before the FIB reconstruction, all the entries are real;

Non-real entry: The auxiliary entry used to support the random search algorithm is called the non-real entry. The name in the non-real entry does not refer to any actual data, and cannot be used to guide the forwarding of interest packets;

Virtual table entry: if a non-real table entry does not have any real prefix, the non-real table entry is called a virtual table entry; when the random search process ends with a virtual table entry, it ends directly without any false negative error;

Semi-virtual entry: if the non-real entry has a real prefix, the non-real entry is called a semi-virtual entry and needs to be searched back;

FIB: forwarding information table;

The forwarding method includes a searching step, and the searching step includes:

Step 1: Query the name in the interest packet through a random search algorithm to obtain the port for forwarding the message;

Step 2: Determine whether the prefix of the last HIT is real, if yes, return the prefix as the query result, otherwise go to step 3;

Step 3: Determine whether the prefix of the last HIT is virtual, if yes, return the query failure, otherwise go to step 4;

Step 4: Trace back up to the first real node in the prefix tree, and return its corresponding forwarding information.

As a further improvement of the present invention, in order to solve the problem of obsolete entries, the forwarding method further includes a deletion step. The deletion step is used to discover and reclaim obsolete non-real entries, and the deletion step includes:

First, determine whether there is a child node in the corresponding entry of the name to be deleted in the FIB, if so, execute the first deletion step, otherwise execute the second deletion step;

First delete step: Determine whether the parent node of the corresponding entry is virtual, if it is virtual, then execute the first delete sub-step, if the parent node of the corresponding entry is real or semi-virtual, then the entry category of the name is modified Semi-virtual

The first deletion sub-step: modify the category of the entry corresponding to the name to virtual, and then traverse the subtree rooted at the name. If a node meets the first and second conditions, then the category of the node is modified to virtual , The first condition: the category is semi-virtual, and the second condition: there is no real node on the path from the node to the name;

The second deletion step: delete the entry, and then progressively check all the true prefixes of the name, if the corresponding node of the prefix meets the first and second points, then delete the node, the first point: the category is not In fact, the second point: it is a leaf node.

The present invention also provides a forwarding device for naming network nodes in a data network, which includes a hash table and a prefix tree. The FIB is composed of the hash table and the prefix tree. For any name stored in the FIB, all the names are The true prefix has corresponding entries in FIB. The process of checking whether the prefix exists and adding corresponding auxiliary entries is called FIB reconstruction. In the reconstructed FIB, the entries are divided into real entries and non-real entries. Entries, non-real entries are divided into virtual entries and semi-virtual entries;

FIB: forwarding information table;

The forwarding device includes a search unit, and the search unit includes:

Query module: used to query the name in the interest packet through a random search algorithm, so as to obtain the port for forwarding the message;

The first judgment module: used to judge whether the prefix of the last HIT is true, if yes, return the prefix as the query result, otherwise execute the second judgment module;

The second judgment module: used to judge whether the prefix of the last HIT is virtual, if yes, return the query failure, otherwise execute the return module;

Return module: used to backtrack up to the first real node in the prefix tree and return its corresponding forwarding information.

The present invention also provides a forwarding system for naming network nodes in a data network, including: a memory, a processor, and a computer program stored on the memory, and the computer program is configured to be implemented when called by the processor The steps of the forwarding method of the present invention.

The present invention also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of the forwarding method of the present invention when called by a processor.

The beneficial effects of the invention

Beneficial effect

The beneficial effects of the present invention are: the present invention realizes a FIB forwarding architecture and related algorithms that support random search, and completely solves the backtracking problem and outdated entry problems in it. At the same time, the experimental evaluation shows that the present invention hardly affects the algorithm. This ensures the efficiency and superiority of the random search algorithm, and lays an important foundation for designing an efficient NDN forwarding architecture and completely solving the scalability problem of NDN.

Brief description of the drawings

Description of the drawings

Figure 1 is a two-stage FIB query process diagram of the background technology;

Figure 2 is a FIB query process diagram based on dichotomy in the background art;

Figure 3 is a forwarding information table based on a hash table and a prefix tree of the present invention;

Figure 4 is a flowchart of the search steps of the present invention;

Figure 5 is a schematic diagram of the search process of the present invention;

Figure 6 is a flowchart of the insertion steps of the present invention;

Figure 7 is a flowchart of the deletion steps of the present invention;

Figure 8 is a performance comparison diagram of binary query and linear query.

Invention embodiment

Embodiments of the present invention

The present invention discloses a forwarding method for naming network nodes in a data network. The present invention focuses on solving the backtracking problem in random algorithms, which can cause so-called false negative errors (False Negative Error), which in turn leads to interest packets being covered. Discard for no reason. In addition, the present invention also realizes timely detection and deletion of obsolete entries, thereby completing efficient recovery of memory.

The present invention designs a forwarding architecture based on the combination of a hash table and a prefix tree, where the hash table is used to support quick search, and the tree structure is used to store the logical relationship between names. We categorized the entries in the hash table to completely solve the problem of backtracking and outdated entries. The present invention also designs matching operation algorithms to ensure the stability and correctness of the architecture.

For ease of introduction, the present invention uses the most basic binary search in random search as a case. In order to support the binary search process, for any name stored in the table, all its true prefixes must have corresponding entries in the table at the same time. The process of checking whether the prefix exists and adding the corresponding auxiliary entries is called FIB reconstruction ( FIB reconstruction). In the reconstructed FIB, the table entries can be divided into the following two categories:

1. Real Entry: The names in the real entry all refer to actual data, and can be used to guide the forwarding of interest packets. Before FIB reconstruction, all entries are real.

2. Non-Real Entry: The auxiliary entry used to support the random search algorithm is called a non-real entry. The name in the entry does not refer to any actual data and cannot be used Guide the forwarding of interest packages.

Next, for the convenience of addressing, we record the names in the real entries as real names, and real prefixes and semi-virtual names have similar meanings.

In the introduction of the backtracking problem in the background technology, it can be seen that the root cause of the problem is "it is impossible to judge whether a non-real table item has a real prefix". For this reason, a backtracking process must be performed on all non-real table item pairs, resulting in relatively In order to cope with this problem, we divide the non-real entries into the following two categories:

1. Virtual Entry: If a non-real entry does not have any real prefix, we call the entry a virtual entry. When the binary search process ends with a dummy entry, it can end directly without any false negative errors.

2. Semi-virtual entry (Semi Virtual Entry): In contrast, if the non-real entry has a real prefix, the entry is called semi-virtual and requires a backtracking lookup.

The further subdivision of categories helps us to optimize the backtracking query process and forms one of the foundations of the architecture design of the present invention.

The present invention designs a forwarding information table (FIB, Forwarding Information Base) based on a hash table and a prefix tree, where the hash table is used to support fast search, and the prefix tree is used to store the logical relationship between names. The main structure of FIB is shown in Figure 3:

In the hash table, the name (such as /c1/c4/c5) is used as the key, and the node (e1) in the prefix tree is used as the value. This achieves rapid retrieval of forwarded information from the name to the tree.

The edges in the prefix tree all represent a name component (Name Component, such as /c1), and each node represents a name, that is, the splicing of all components on the path from the node to the root. The node stores the forwarding information corresponding to the name and the corresponding category of the entry, as well as pointers used to maintain the tree structure. Since the name does not need to be recorded repeatedly in the tree, the introduction of the tree structure will only bring limited additional storage overhead.

The introduction of the prefix tree is for the following three purposes:

1. The prefix tree helps us modify the categories of non-real entries in a dynamic environment. For example, if "/c1" is inserted into the FIB, all virtual entries with the prefix "/c1" need to be modified to semi-virtual, due to the diversity of NDN identification, it is not necessary to use the tree Under the premise of the structure, the modification can only be realized by traversing the entire table.

2. The prefix tree helps us speed up the backtracking process in the search algorithm: because each node in the tree has only one parent node, the process of backtracking does not require name or component query matching operations, so it has a faster search speed.

3. The prefix tree helps us clear outdated non-real entries in time, that is, those leaf nodes whose types are non-real. In the algorithm process of the table, as long as it is ensured that the types of all leaf nodes are not real, the cleaning of redundant table entries can be completed.

As shown in Figure 4, the forwarding method of the present invention includes a search step, and the search step includes:

Step 1: Query the name in the interest packet through a random search algorithm to obtain the port for forwarding the message. For example, the random search algorithm is a binary search algorithm;

Step 2: After the binary search algorithm ends, first, judge whether the prefix of the last HIT is real, if it is, the LPM search has been successful, and return the prefix as the query result, otherwise go to step 3;

Step 3: Determine whether the prefix of the last HIT is virtual. If it is, you can be sure that there is no matching real prefix in the table, and return the query as a failure, otherwise go to step 4;

Step 4: It can be determined that the entry is a semi-virtual entry, and then it can be known that the entry has a real prefix, go back up in the prefix tree structure to find the real entry and return, that is, back up in the prefix tree to The first real node returns its corresponding forwarding information. Since the backtracking process in the prefix tree does not involve name matching and search, this process has minimal time overhead.

As shown in Figure 5, in the process of querying the name "/c1/c2/c3/c6/c7", the final HIT prefix of the binary search is "/c1/c2/c3", which is a semi-virtual entry. This enters the backtracking process: the algorithm starts from "/c1/c2/c3", goes back up until it encounters the first actual entry "/c1", and returns its corresponding forwarding information.

The present invention is also applicable to other advanced schemes based on binary search, thereby completely solving the backtracking problem in random search.

As shown in FIG. 6, the forwarding method further includes an inserting step. The inserting step is used to insert the name into the FIB, and the inserting step includes:

First, determine whether the name to be inserted in the FIB query exists, if it is, then execute the first insertion step, otherwise execute the second insertion step;

The first inserting step includes:

Step 1: Determine whether the entry corresponding to the name is a real entry, if so, update its forwarding information, otherwise proceed to step two;

Step 2: Determine whether the entry corresponding to the name is a virtual entry, if it is, then perform step three, otherwise perform the modification step;

Step 3: Modify all virtual entries in its subtree to semi-virtual entries, and then perform the modification steps;

Modification steps: modify its category as a real entry and add forwarding information.

In summary, in the first insertion step, if the name to be inserted already has a corresponding entry in the FIB, the issue of newly added entries is not involved at this time. The situation of a real entry is trivial, so we only consider the case of a non-real entry: first, the category of the corresponding entry will be modified to real; if the entry is originally virtual, the virtual entry in the subtree needs Modify its category to semi-virtual. If the entry is originally semi-virtual, the subtree does not need to be modified.

The second inserting step includes:

First, query the LPM of the name to be inserted in the FIB, if HIT is real, then execute the first processing step, if MISS or HIT is virtual, then execute the second processing step;

The first processing step: insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding semi-virtual entry if it does not exist;

The second processing step: insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding virtual entry if it does not exist.

In summary, in the second insertion step, if the name to be inserted does not have a corresponding entry, the corresponding real entry is inserted, and the prefix is checked progressively from back to front to ensure that they all exist in the FIB. If a certain prefix is found to be absent, the corresponding non-real entry is inserted. The process continues until the algorithm reaches the LPM or the root node, and thus determines the category of the non-real node inserted in this process.

As shown in FIG. 7, the forwarding method further includes a deletion step. The deletion step is used to discover and reclaim outdated non-real entries, and the deletion step includes:

Through the deletion step of the present invention, the category of non-real entries can be kept correct in a dynamic environment, and outdated non-real entries can be discovered and recycled in time, thereby ensuring the efficiency and stability of the forwarding plane. Therefore, the present invention completely solves the problem of outdated entries.

The invention also discloses a forwarding device for naming network nodes in a data network. FIB includes a hash table and a prefix tree. For any name stored in the table, all its true prefixes have corresponding entries in the table. , The process of checking whether the prefix exists and adding corresponding auxiliary entries is called FIB reconstruction. In the reconstructed FIB, the entries are divided into real entries and non-real entries, and non-real entries are divided into virtual tables Entries and semi-virtual entries;

FIB: forwarding information table;

The forwarding device includes a search unit, and the search unit includes:

The forwarding device also includes an inserting unit, the inserting unit is used to insert a name into the FIB, and the inserting unit includes:

First, determine whether the name to be inserted in the FIB query exists, if so, execute the first insertion module, otherwise execute the second insertion module;

The first plug-in module includes:

The first insertion judgment module: used to judge whether the entry corresponding to the name is a real entry, if so, then update its forwarding information, otherwise execute the second insertion judgment module;

Second insertion judgment module: used to judge whether the entry corresponding to the name is a virtual entry, if so, then enter the execution module, otherwise enter the modification module;

Execution module: used to modify all virtual entries in its subtree to semi-virtual entries, and then execute the modification module;

Modification module: used to modify its category as a real entry and add forwarding information;

The second plug-in module includes:

First, query the LPM of the name to be inserted in the FIB, if HIT is real, then execute the first processing module, if MISS or HIT is virtual, then execute the second processing module;

The first processing module: used to insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding semi-virtual entry if it does not exist;

The second processing module: used to insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding virtual entry if it does not exist;

LPM: Longest prefix match.

The forwarding device further includes a deletion unit, the deletion unit is used to discover and reclaim outdated non-real entries, and the deletion unit includes:

First, determine whether there are child nodes in the corresponding entry of the name to be deleted in the FIB, if so, execute the first deletion module, otherwise execute the second deletion module;

First delete module: used to determine whether the parent node of the corresponding entry is virtual, if it is virtual, then execute the first delete submodule, if the parent node of the corresponding entry is real or semi-virtual, then the name of the entry Modified the category to semi-virtual;

The first delete sub-module: used to modify the category of the entry corresponding to the name to virtual, and then traverse the subtree rooted at the name, if a node meets the first and second conditions, then the category of the node is modified Is virtual, the first condition: the category is semi-virtual, and the second condition: there is no real node on the path from the node to the name;

The second delete module: used to delete the entry, and then progressively check all the true prefixes of the name, if the corresponding node of the prefix meets the first and second points, then delete the node, the first point: category It is not real, and the second point: it is a leaf node.

The random search algorithm is a binary search algorithm. In the query module, the binary search algorithm is used to query the name in the interest packet to obtain the port forwarding the message; virtual table entry: if a non-real entry does not have any real prefix, then The non-real entry is called a virtual entry; when the binary search process ends with a virtual entry, it ends directly without any false negative error.

The present invention also discloses a forwarding system for naming network nodes in a data network, comprising: a memory, a processor, and a computer program stored on the memory, and the computer program is configured to be implemented when called by the processor The steps of the forwarding method of the present invention.

The present invention also discloses a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of the forwarding method of the present invention when called by a processor.

The present invention realizes a FIB forwarding architecture and related algorithms that support random search, and completely solves the backtracking problem and outdated entry problems in it. At the same time, experimental evaluation shows that the present invention hardly affects the time cost of the algorithm, thereby ensuring The efficiency and superiority of the random search algorithm have laid an important foundation for us to design an efficient NDN forwarding architecture and completely solve the scalability problem of NDN.

With respect to the FIB forwarding data structure proposed in the present invention, the performance comparison between binary query and linear query is performed under different FIB table name prefix scales. Count the total time of one million name lookups and calculate the average lookup time (unit: us). The average number of components of the name prefix in the FIB table is 4.2, and the experimental result when the average number of components of the name prefix found is 11.2 is shown in Figure 8. The average throughput of the binary query (the number of queries that can be processed/s) is 1.80 times that of the linear query, and when the scale of the FIB table is larger, that is, 1 billion, the present invention can still achieve a throughput of 0.2M.

Table 1 FIB table scale and generation and construction time

At the same time, the FIB forwarding data structure proposed by the present invention supports larger-scale name prefix storage and query, as shown in Table 1. The data source of this experiment is the self-generated NDN identification. Since NDN is a new type of network architecture and has not yet been used in large-scale practical applications, it is difficult for us to obtain large-scale NDN data samples from actual network traffic. Therefore, we analyzed some statistical properties in the current network flow, simulated a large number of NDN identifiers, and generated a large-scale FIB table on this basis, so as to realize the evaluation and research of the present invention.

The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For those of ordinary skill in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

A forwarding method for naming network nodes in a data network, including a hash table and a prefix tree. It is characterized in that the FIB is composed of a hash table and a prefix tree. For any name stored in the FIB, all its true prefixes The process of having corresponding entries in FIB, checking whether the prefix exists and adding corresponding auxiliary entries is called FIB reconstruction. In the reconstructed FIB, the entries are divided into real entries and non-real entries. Non-real entries are divided into virtual entries and semi-virtual entries;

In the hash table, the name is used as the key, and the node in the prefix tree is used as the value, thereby realizing fast retrieval of forwarding information from the name to the prefix tree;

Each edge in the prefix tree represents a name component, and each node of the prefix tree represents a name. The node stores the forwarding information corresponding to the name and the corresponding category of the table item, as well as the pointer used to maintain the prefix tree structure;

Real entry: The name in the real entry refers to the actual data and is used to guide the forwarding of the interest packet; before the FIB reconstruction, all the entries are real;

Non-real entry: The auxiliary entry used to support the random search algorithm is called the non-real entry. The name in the non-real entry does not refer to any actual data, and cannot be used to guide the forwarding of interest packets;

Virtual table entry: if a non-real table entry does not have any real prefix, the non-real table entry is called a virtual table entry; when the random search process ends with a virtual table entry, it ends directly without any false negative error;

Semi-virtual entry: if the non-real entry has a real prefix, the non-real entry is called a semi-virtual entry and needs to be searched back;

FIB: forwarding information table;

The forwarding method includes a searching step, and the searching step includes:

Step 1: Query the name in the interest packet through a random search algorithm to obtain the port for forwarding the message;

Step 2: Determine whether the prefix of the last HIT is real, if yes, return the prefix as the query result, otherwise go to step 3;

Step 3: Determine whether the prefix of the last HIT is virtual, if yes, return the query failure, otherwise go to step 4;

Step 4: Trace back up to the first real node in the prefix tree, and return its corresponding forwarding information.
The forwarding method according to claim 1, wherein the forwarding method further comprises an inserting step, and the inserting step is used to insert a name into the FIB, and the inserting step includes:

First, determine whether the name to be inserted in the FIB query exists, if it is, then execute the first insertion step, otherwise execute the second insertion step;

The first inserting step includes:

Step 1: Determine whether the entry corresponding to the name is a real entry, if so, update its forwarding information, otherwise proceed to step two;

Step 2: Determine whether the entry corresponding to the name is a virtual entry, if it is, then perform step three, otherwise perform the modification step;

Step 3: Modify all virtual entries in its subtree to semi-virtual entries, and then perform the modification steps;

Modification steps: modify its category as a real entry and add forwarding information;

The second inserting step includes:

First, query the LPM of the name to be inserted in the FIB, if HIT is real, then execute the first processing step, if MISS or HIT is virtual, then execute the second processing step;

The first processing step: insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding semi-virtual entry if it does not exist;

The second processing step: insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding virtual entry if it does not exist;

LPM: Longest prefix match.
The forwarding method according to claim 1, wherein the forwarding method further comprises a deleting step, the deleting step is used for discovering and recycling outdated non-real entries, and the deleting step comprises:

First, determine whether there is a child node in the corresponding entry of the name to be deleted in the FIB, if so, execute the first deletion step, otherwise execute the second deletion step;

First delete step: Determine whether the parent node of the corresponding entry is virtual, if it is virtual, then execute the first delete sub-step, if the parent node of the corresponding entry is real or semi-virtual, then the entry category of the name is modified Semi-virtual

The first deletion sub-step: modify the category of the entry corresponding to the name to virtual, and then traverse the subtree rooted at the name. If a node meets the first and second conditions, then the category of the node is modified to virtual , The first condition: the category is semi-virtual, and the second condition: there is no real node on the path from the node to the name;

The second deletion step: delete the entry, and then progressively check all the true prefixes of the name, if the corresponding node of the prefix meets the first and second points, then delete the node, the first point: the category is not In fact, the second point: it is a leaf node.
The forwarding method according to any one of claims 1 to 3, wherein the random search algorithm is a binary search algorithm, and in step 1, the name in the interest packet is queried by the binary search algorithm, so as to obtain the information of the forwarded message port;

When the binary search process ends with a dummy entry, it ends directly without any false negative errors.
A forwarding device for naming network nodes in a data network, including a hash table and a prefix tree. It is characterized in that the FIB is formed by the hash table and the prefix tree. For any name stored in the FIB, all its true prefixes The process of having corresponding entries in FIB, checking whether the prefix exists and adding corresponding auxiliary entries is called FIB reconstruction. In the reconstructed FIB, the entries are divided into real entries and non-real entries. Non-real entries are divided into virtual entries and semi-virtual entries;

In the hash table, the name is used as the key, and the node in the prefix tree is used as the value, thereby realizing fast retrieval of forwarding information from the name to the prefix tree;

Each edge in the prefix tree represents a name component, and each node of the prefix tree represents a name. The node stores the forwarding information corresponding to the name and the corresponding category of the table item, as well as the pointer used to maintain the prefix tree structure;

Real entry: The name in the real entry refers to the actual data and is used to guide the forwarding of the interest packet; before the FIB reconstruction, all the entries are real;

Non-real entry: The auxiliary entry used to support the random search algorithm is called the non-real entry. The name in the non-real entry does not refer to any actual data, and cannot be used to guide the forwarding of interest packets;

Virtual table entry: if a non-real table entry does not have any real prefix, the non-real table entry is called a virtual table entry; when the random search process ends with a virtual table entry, it ends directly without any false negative error;

Semi-virtual entry: if the non-real entry has a real prefix, the non-real entry is called a semi-virtual entry and needs to be searched back;

FIB: forwarding information table;

The forwarding device includes a search unit, and the search unit includes:

Query module: used to query the name in the interest packet through a random search algorithm, so as to obtain the port for forwarding the message;

The first judgment module: used to judge whether the prefix of the last HIT is true, if yes, return the prefix as the query result, otherwise execute the second judgment module;

The second judgment module: used to judge whether the prefix of the last HIT is virtual, if yes, return the query failure, otherwise execute the return module;

Return module: used to backtrack up to the first real node in the prefix tree and return its corresponding forwarding information.
The forwarding device according to claim 5, wherein the forwarding device further comprises an inserting unit, the inserting unit is configured to insert a name into the FIB, and the inserting unit comprises:

First, determine whether the name to be inserted in the FIB query exists, if so, execute the first insertion module, otherwise execute the second insertion module;

The first plug-in module includes:

The first insertion judgment module: used to judge whether the entry corresponding to the name is a real entry, if so, then update its forwarding information, otherwise execute the second insertion judgment module;

Second insertion judgment module: used to judge whether the entry corresponding to the name is a virtual entry, if so, then enter the execution module, otherwise enter the modification module;

Execution module: used to modify all virtual entries in its subtree to semi-virtual entries, and then execute the modification module;

Modification module: used to modify its category as a real entry and add forwarding information;

The second plug-in module includes:

First, query the LPM of the name to be inserted in the FIB, if HIT is real, then execute the first processing module, if MISS or HIT is virtual, then execute the second processing module;

The first processing module: used to insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding semi-virtual entry if it does not exist;

The second processing module: used to insert the real entry corresponding to the name, find all the true prefixes of the name in the FIB, and insert the corresponding virtual entry if it does not exist;

LPM: Longest prefix match.
The forwarding device according to claim 5, wherein the forwarding device further comprises a deleting unit, the deleting unit is configured to discover and reclaim outdated non-real entries, and the deleting unit comprises:

First, determine whether there are child nodes in the corresponding entry of the name to be deleted in the FIB, if so, execute the first deletion module, otherwise execute the second deletion module;

First delete module: used to determine whether the parent node of the corresponding entry is virtual, if it is virtual, then execute the first delete submodule, if the parent node of the corresponding entry is real or semi-virtual, then the name of the entry Modified the category to semi-virtual;

The first delete submodule: used to modify the category of the entry corresponding to the name to virtual, and then traverse the subtree rooted by the name, if a node meets the first and second conditions, then the category of the node is modified Is virtual, the first condition: the category is semi-virtual, the second condition: there is no real node on the path from the node to the name;

The second delete module: used to delete the entry, and then progressively check all the true prefixes of the name, if the corresponding node of the prefix meets the first and second points, then delete the node, the first point: category It is not real, and the second point: it is a leaf node.
The forwarding device according to any one of claims 5 to 7, wherein the random search algorithm is a binary search algorithm, and in the query module, the name in the interest packet is queried by the binary search algorithm to obtain the information of the forwarded message port;

When the binary search process ends with a dummy entry, it ends directly without any false negative errors.
A forwarding system for naming network nodes in a data network, comprising: a memory, a processor, and a computer program stored on the memory, and the computer program is configured to implement rights when called by the processor The steps of the forwarding method described in any one of 1-4 are required.
A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to implement the forwarding method according to any one of claims 1 to 4 when called by a processor A step of.