CN113364685B

CN113364685B - Distributed MAC table item processing device and method

Info

Publication number: CN113364685B
Application number: CN202110534463.6A
Authority: CN
Inventors: 吕高锋; 卓超; 李韬; 全巍; 刘汝霖; 李存禄; 乔冠杰; 莫露莎
Original assignee: National University of Defense Technology
Current assignee: National University of Defense Technology
Priority date: 2021-05-17
Filing date: 2021-05-17
Publication date: 2023-03-14
Anticipated expiration: 2041-05-17
Also published as: CN113364685A

Abstract

The invention belongs to the technical field of computer communication, and provides a distributed MAC (media access control) table entry processing device and method. The device comprises MAC table management units deployed on each assembly line, wherein each MAC table management unit is used for learning according to a message source MAC address, and the MAC table management units on each assembly line are interacted based on learning notification information, aging refreshing notification information and miss notification information so as to maintain the synchronization of the MAC table of each assembly line.

Description

Distributed MAC table item processing device and method

Technical Field

The invention belongs to the technical field of computer communication, and particularly relates to a distributed MAC (media access control) table entry processing device and method.

Background

The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.

With the rapid development of big data and data centers, the demand of the rapid increase of network traffic on the processing performance of network devices is continuously increasing. The main stream interface of the switching chip as the core part of the two-layer switching is changed from the original one hundred mega/giga rate ethernet interface to the current 100G/400G ethernet interface, the switching capacity is further increased to the Tbps level, but limited by the maximum system clock frequency, and the performance of a single pipeline processing mode is difficult to be increased, so that the current switching chip mostly adopts a mode of parallel processing of a plurality of pipelines to meet the requirement of the switching capacity of the Tbps level.

The inventor finds that, for a switch chip implementing the ethernet two-layer switching function, MAC table entries are not completely configured in a centralized manner, but need to be self-learned and maintained by each pipeline, and when a plurality of pipelines are used to process MAC table lookup in parallel to improve the two-layer switching capacity, a method and an apparatus for maintaining the MAC table entries of each pipeline in synchronization are needed.

Disclosure of Invention

In order to solve the technical problems in the background art, the present invention provides a distributed MAC entry processing apparatus and method, which can maintain the MAC entries of each pipeline in synchronization when multiple pipelines process in ethernet two-layer switching.

In order to achieve the purpose, the invention adopts the following technical scheme:

a first aspect of the present invention provides a distributed MAC entry processing apparatus, including MAC entry management units deployed on each pipeline, where each MAC entry management unit is configured to learn according to a message source MAC address, and interact MAC entry management units on each pipeline based on learning notification information, aging refresh notification information, and miss notification information, so as to maintain synchronization of MAC entries of each pipeline.

Further, the MAC table item management unit includes a learning sub-module, a searching sub-module, an aging sub-module, and a relay sub-module.

Further, the learning submodule is configured to perform learning of the MAC table entry according to the source MAC address and the input port number of the packet, and generate learning notification information to notify the table entry management unit of another pipeline to update the table entry at the same time, and the learning submodule also needs to process the learning notification information sent by another pipeline to update the table entry.

Furthermore, the learning sub-module is further configured to process the miss notification information, and check whether the table management unit stores the table entries that need to be searched by another pipeline, that is, the table entries that cannot be searched by the table searching process of another pipeline.

Further, the learning sub-module is provided with entry identifiers, and the entry identifiers are used for distinguishing entries directly inserted by processing learning notification information and entries obtained by self-learning, so that the aging sub-module can distinguish and process the entries.

Furthermore, the searching submodule is configured to perform table entry searching according to the message destination MAC address, and generate miss notification information when the searching submodule of all table entry management units in the pipeline misses.

Further, the aging submodule is used for deleting the MAC table entry which is no longer in an active state, releasing the storage space, and periodically generating an aging refresh notification message to notify the active state of the current table entries of other pipelines;

further, the relay sub-module is configured to receive the learning notification information, the miss notification information, and the aging refresh notification information from the notification message bus and transmit the learning notification information, the miss notification information, and the aging refresh notification information to the corresponding sub-module, and meanwhile, make a decision as to whether the notification information continues to be transmitted on the bus.

Furthermore, the learning notification information, the aging refreshing notification information and the missing notification information are transmitted in sequence in each pipeline until the mission of the corresponding notification information is finished.

A second aspect of the present invention provides a distributed MAC entry processing method based on the above-mentioned distributed MAC entry processing apparatus, where the method is implemented in an entry management unit, and includes:

learning MAC table items according to the source MAC address and the input port number of the message, and generating and processing learning notification information to interact with table item management units of other pipelines;

deleting MAC table entries which are no longer in an active state, releasing a storage space, and periodically generating an aging refresh notification message to notify the active state of the current table entries of other pipelines;

each flow line searches in respective table item management unit according to the assigned message destination MAC address, outputs the hit result, and generates miss notification information if the hit result is not hit;

and processing the miss notification information, and checking whether the table item management unit stores table items required to be searched by other pipelines, namely the table items which cannot be searched by the table searching process of other pipelines.

Compared with the prior art, the invention has the beneficial effects that:

the invention provides a distributed MAC (media access control) table item processing device which comprises MAC table item management units arranged on each production line, wherein the MAC table item management units are used for learning according to message source MAC addresses, and the MAC table item management units on each production line are interacted based on learning notification information, aging refreshing notification information and miss notification information so as to maintain the synchronization of MAC table items of each production line, thereby finally solving the problem of MAC table item synchronization when a plurality of production lines process two-layer exchange with ultra-large capacity.

Advantages of additional aspects of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an exemplary embodiment of the invention and, together with the description, serve to explain the invention and are not intended to limit the invention.

Fig. 1 is a schematic diagram of an overall structure of a distributed MAC processing method according to an embodiment of the present invention;

fig. 2 is a flowchart of performing MAC table entry lookup according to an embodiment of the present invention;

fig. 3 is a flowchart of performing MAC table entry maintenance according to an embodiment of the present invention;

FIG. 4 is a diagram illustrating the content of three advertisement messages according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a relay flow of LAM/ARAM in each pipeline according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of the MAM relay process on each pipeline according to the embodiment of the present invention;

fig. 7 is a basic structure diagram of a MAC table entry management unit according to an embodiment of the present invention.

Detailed Description

The invention is further described with reference to the following figures and examples.

It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.

Interpretation of terms:

and (4) table item management: each flow line is provided with n MAC table management units, and each table management unit can respectively carry out learning, aging, searching, static configuration and the like of the table items so as to improve the table item management performance of each flow line.

A distributed system: the message is distributed to m pipelines according to rules (such as input port number, quintuple hash and the like) to be subjected to table lookup forwarding processing. The search of the MAC table items is completed by a production line, namely the production line searches in n table item management units at the same time according to the MAC address of the message destination, and n results are output after being compared. The maintenance functions of learning, aging and the like of the MAC table items are completed based on the table item management unit, each assembly line carries out Hash calculation according to the message source MAC address to obtain an index, and then assigns the message source MAC address to the table item management unit indicated by the index for learning, so that the table items can be distributed in the n table item management units for storage and maintenance. Each table item management unit completes the task of searching and maintaining the distributed table items, and the m x n table item management units work cooperatively to realize the function of two-layer MAC exchange, so that a distributed table item management system is formed.

And (4) table item synchronization: the n table item management units of each pipeline learn according to the message source MAC address, and the range of MAC address search is only the n table item management units of the pipeline, so each pipeline also needs to learn the MAC table items of other pipelines. The invention can make each pipeline interact the operation of the table items through the following three kinds of notification messages, thereby maintaining the synchronization of the MAC table items of each pipeline.

Learning notification information: when the table management unit of a certain assembly line learns according to the source MAC address assigned to the Message of the assembly line, and the content of the MAC table needs to be added or updated, learning notification information LAM (Learn Announce Message) is generated and is used for notifying the table management units of other assembly lines; other pipelines synchronize this entry operation with the LAM by processing it.

Aging refresh announcement message: in order to dynamically maintain the list items shared by the LAM and prevent the list items from being simply aged by the list item management unit, the list item management unit generating the LAM periodically generates an aging Refresh notification Message ARAM (Age Refresh Announce Message) to notify other pipelines that the current list items are still in an active state and cannot be aged; other pipelines keep the entries synchronized with them by processing the ARAMs.

Miss notification information: when all the n table item management units in a certain pipeline search the MAC address, miss notification information MAM (Miss Announce Message) is generated to notify other pipelines to assist in searching, and if other pipelines have matching table items, the Miss notification information MAM is notified by LAM so as to facilitate the subsequent same searching operation to be hit.

Example one

Referring to fig. 1, the distributed MAC entry processing apparatus of this embodiment relates to m pipelines (101), each pipeline (101) may independently perform processing such as MAC address learning and lookup on a packet, the processing capability of each pipeline (101) is limited by factors such as system clock frequency, and if B is obtained, the parallel processing capability of m pipelines (101) may theoretically reach m × B. Messages are typically dispatched to each pipeline (101) for processing according to rules (e.g., input port numbers) to load balance traffic.

In the embodiment, n table entry management units (102) are deployed in each pipeline (101) to improve the MAC address learning capability of the pipeline (101). The table item management unit (102) is a basic logic list for realizing the functions of learning, searching, aging and the like of table itemsA meta. Hash of message source MAC addresses dispatched to a pipeline (101) _L1 The index is calculated, and the message source MAC address is assigned to the table item management unit (102) pointed by the index to learn, so that the learning pressure of the single table item management unit (102) is relieved. The invention comprises m pipelines (101), each pipeline (101) is provided with n table management units (102), and m × n table management units (102) are provided, for convenience of description, a TMU is used _x,y A y-th table entry management unit (102) representing the x-th pipeline (101), where 0<x<＝m,0<y<＝n。

In the figure, 103 is an advertisement message bus for transmitting the above three kinds of advertisement messages. The announcement message is relayed in sequence in each pipeline (101) until the mission of the announcement message is finished.

As shown in fig. 7, the MAC entry management unit of this embodiment includes a learning sub-module (701), a searching sub-module (702), an aging sub-module (703), and a relay sub-module (704).

In a specific implementation, the learning submodule (701) is configured to perform MAC table entry learning according to a source MAC address and an input port number of a packet, and generate and process learning notification information to interact with table entry management units of other pipelines.

Specifically, the learning sub-module (701) is further configured to process the miss notification information, and check whether the table entry management unit stores a table entry that needs to be searched by another pipeline, that is, a table entry that cannot be searched by the table search process of another pipeline.

The learning submodule (701) is provided with a table entry identifier, and the table entry identifier is used for distinguishing and processing a table entry directly inserted by learning notification information and a table entry obtained by self-learning so as to facilitate the distinguishing and processing of the aging submodule.

In a specific implementation, the lookup sub-module (702) is configured to perform table entry lookup according to the MAC address of the message destination, and generate miss notification information when the lookup sub-modules of all table entry management units in the pipeline miss.

Fig. 2 is a flow chart of MAC table entry lookup according to the present invention. The search of the MAC table entry is completed based on m pipelines, each pipeline searches n table entry management units according to the assigned message destination MAC address, and n results are processed and then output. The specific implementation process is as follows:

step 201: after the message arrives, the destination MAC address is dispatched to the m pipelines according to rules (such as input port numbers or other rules);

step 202: the flow line simultaneously searches the MAC addresses in the n table item management units according to the assigned message destination MAC addresses;

step 203: if there is hit in the n table look-up results, outputting the hit table as the result of the table look-up operation;

step 204: if the n table lookup results are not hit, the table lookup operation result is a flooding operation, and miss notification information MAM is generated;

step 205: and forwarding the message according to the result of the table look-up operation.

The MAM issued in step 204 is used to search the target MAC address missed in this time in other pipelines, but the lookup is not performed in all n table entry management units of each pipeline, but is performed according to the Hash function Hash of the target MAC address _L1 The calculation result of (2) is selected.

In a specific implementation, the aging submodule (703) is configured to delete an MAC entry that is no longer in an active state, release a storage space, and periodically generate an aging refresh notification message to notify an active state of a current entry of another pipeline;

the aging submodule (703) may send out an ARAM only for the self-learned entry, and for the received ARAM, finds out the corresponding entry according to the storage address of the cache entry therein, and resets the aging timing thereof.

In a specific implementation, the relay sub-module (704) is configured to receive the learning notification message, the miss notification message, and the aging refresh notification message from the notification message bus and transmit the learning notification message, the miss notification message, and the aging refresh notification message to the corresponding sub-module, and meanwhile, make a decision on whether the notification message continues to be transmitted on the bus. If the transmission is continued, the step length is increased by one, otherwise, the notification message is discarded; in addition, the LAM and the ARAM generated by the learning submodule (701) and the aging submodule (703) need to be sent to the notification message bus (705).

Fig. 3 is a flowchart of MAC entry maintenance performed in this embodiment. The MAC table item maintenance is that the table items are added, changed and deleted through the functions of self-learning, aging and the like of the table items. In this embodiment, through the three announcement messages of LAM, ARAM, and MAM, the pipelines interact with each other, and maintain the synchronicity of their respective entries dynamically, so that the pipelines realize a complete two-layer switching function. The following describes the maintenance flow of the table entry with respect to the generation and processing process of three types of notification messages by using a certain pipeline. Steps 301-306 in fig. 3 are the generation and processing flow of LAM.

Step 301: after the message arrives, the source MAC address and the input port number of the message are dispatched to the m pipelines according to rules (such as input port numbers or other rules);

step 302: hash function Hash of source MAC address by pipeline _L1 Calculating, and further reassigning the source MAC address and the input port number to the table item management unit pointed by the result;

step 303: the entry management unit learns the assigned source MAC address and its input port number. If a new table entry is learned, namely the table entry content needs to be added or updated, updating the table entry;

step 304: generating learning announcement information LAM according to the content of the corresponding table item and the table item cache address, and sending the learning announcement information LAM to the bus so as to share the table item to other pipelines and realize the synchronization of the pipelines;

step 305: when a certain assembly line receives the LAM, finding a corresponding storage space in a corresponding table entry management unit according to the LAM, and inserting the table entry, namely directly covering the original table entry;

step 306: if the LAM has not been passed to all of the pipelines, it is sent on the bus to allow the pipelines that have not received the LAM to process.

In step 303, when the source MAC address is learned, the aging time of the table entry is reset regardless of whether a new table entry is learned. In step 305, the aging timer of the entry is also reset when the entry is inserted.

307-310 in the figure are ARAM generation and processing flows.

Step 307: when the aging time is up, checking whether the MAC table item is hit again by the learning function in the aging time of the round, namely whether the MAC table item needs to be aged;

step 308: if the table entry does not need to be aged, the table entry management unit generates an aging refresh notification message ARAM and sends the aging refresh notification message ARAM to the bus;

step 309: other pipelines receive the ARAM, find out the corresponding table entry according to the ARAM, mark it can't be aged in the aging time of the next round;

step 310: if the ARAM has not been passed to all of the pipelines, it is sent on the bus for processing by the pipelines that have not received the ARAM.

In the step 308, the entry directly inserted by the LAM is distinguished from the entry learned by self, and the aging refresh notification message ARAM may be sent only for the entry learned by self.

Steps 311-313 of fig. 3 are the MAM generation and processing flow.

Step 311: if the table item searching process is not hit, the HashL1 of the hash function is carried out according to the target MAC address, the MAM is generated, and the MAM is sent to the notification bus by a relay sub-module of the table item management unit where the index is located;

step 312: other pipelines receive the MAM, match in respective table item management units according to the MAM, if the table item matching is successful, go to step 304, otherwise go to step 313;

step 313: if the MAM has not been passed to all pipelines, it is sent to the bus for processing by the pipelines that have not received the MAM.

As shown in fig. 4, a schematic diagram of the information content included in the three advertisement messages defined in the present invention is shown.

The LAM (401) should contain the information: MAC address, port number port, memory address addr of cache table entry, LAM step length step. The MAC address and the port number port are the basic contents of the corresponding table entry; the memory address addr of the cache table entry is the address of the memory space where the corresponding table entry is located; the LAM step is used for recording how many pipelines the current LAM passes through, the initial value is 1, when the LAM passes through one pipeline, one is added, when the step is m-1, the LAM is indicated to pass through all pipeline processing, and the mission of the LAM is ended.

The ARMM (402) contains information: the storage address addr and the ARAM step size step of the cache table entry have similar meanings to those of the LAM.

The MAM (403) contains information: the MAC address addr and the MAM step size step. The MAC address is similar to the step length of the DMAC and the MAM of the message when the table look-up process is not hit.

Fig. 5 is a schematic diagram illustrating a relay flow of the LAM/ara advertisement message on each pipeline. LAM is similar to ARAM notice information transfer flow, generated by a table item management unit of a certain assembly line, and transferred on other assembly lines in a bus relay transfer mode in turn, until all the assembly lines receive notice information, the mission of LAM/ARAM is finished. The specific implementation process is as follows:

step 501: in the above-mentioned table item maintenance flow, a certain table item management unit TMU _a,b Generating an LAM/ARAM announcement message with an initial step size of 1;

step 502: TMU _a,b Passing this notification message to the table entry management unit TMU of the next pipeline _a+1,b (if a = m, then TMU _1,b (ii) a Subsequent transfer processes are similar);

step 503: TMU _a+1,b Receiving the notification message, processing according to the content of the notification message, namely inserting an entry or marking the entry not to be aged in the next round of aging time;

step 504: TMU _a+1,b And determining whether to continue the transmission according to the step size of the notification message. If the step size is m-1, which indicates that all pipelines have received the LAM/ARAM, go to step 506; if the step size is less than m-1, go to step 505;

step 505: the step length of the notification message is added by one and then is transmitted to the table entry of the next flow lineManagement unit TMU _a+2,b Go to step 507;

step 506: the message mission is finished and the transmission is stopped;

step 507: and analogizing the operations of steps 503-506 until the end.

As shown in fig. 6, a schematic diagram of the MAM relaying flow on each pipeline is shown. The MAM is not generated by the table management unit, but is generated by a pipeline in executing the table lookup process when all table entries of the table management unit are not hit, and the MAM is also transmitted in sequence by adopting a bus, but is not necessarily transmitted to all pipelines, and the MAM mission is finished as long as one of the pipelines is successfully matched. The specific implementation process is as follows:

step 601: in the table lookup process, when a message of a certain pipeline e is not hit and flooded in all table lookup performed by n table entry management units, the pipeline e generates miss notification information MAM, and performs hash function HashL1 calculation according to a destination MAC address to obtain an index f, and the table entry management unit TMU where the index is located _e,f The relay submodule of the system is sent to an announcement bus, and the initial step length of the MAM is 1;

step 602: the MAM is passed to the table entry management unit TMU of the next pipeline _e+1,f ；

Step 603: TMU _e+1,f After receiving the MAM, processing according to the MAM content, namely, the MAC address and the TMU in the MAM _e+1,f And matching the MAC addresses of the middle table entries. If the matching is successful, go to step 604, otherwise go to step 605;

step 604: TMU _e+1,f Generating an LAM for the successfully matched list item, performing the relay processing flow of the LAM, and turning to step 606;

step 605: TMU _e+1,f And deciding whether to continue the transmission according to the step size of the notification message. If the step size is m-1, which indicates that all pipelines have received the MAM, go to step 606; if the step size is less than m-1, go to step 607;

step 606: the MAM mission is over, stopping delivery;

step 607: adding one to the step length of the MAM and transmitting the step length to the MAMTable item management unit TMU of next pipeline _e+2,f ；

Step 608: and so on from steps 603-607 to the end.

Example two

The embodiment provides a distributed MAC entry processing method based on the above-mentioned distributed MAC entry processing apparatus, where the method is implemented in an entry management unit, and includes:

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A distributed MAC table item processing device is characterized by comprising MAC table item management units deployed on each assembly line, wherein each MAC table item management unit is used for learning according to a message source MAC address, and enabling the MAC table item management units on each assembly line to interact based on learning notification information, aging refreshing notification information and miss notification information so as to maintain the synchronization of MAC table items of each assembly line;

and the learning notification information, the aging refreshing notification information and the missing notification information are transmitted in sequence in each pipeline until the mission of the corresponding notification information is finished.

2. The distributed MAC entry processing apparatus according to claim 1, wherein the MAC entry management unit includes a learning sub-module, a lookup sub-module, an aging sub-module, and a relay sub-module, and the relay sub-module communicates with the learning sub-module, the lookup sub-module, and the aging sub-module through an announcement message bus, respectively.

3. The distributed MAC entry processing apparatus as claimed in claim 2, wherein the learning sub-module is configured to perform learning of the MAC entry according to a source MAC address and an input port number of the packet, and generate a learning notification message to notify an entry management unit of another pipeline to update the entry, and the learning sub-module also needs to process the learning notification message sent by the other pipeline to update the table.

4. The distributed MAC table entry processing apparatus of claim 2, wherein the learning sub-module is further configured to process the miss notification information to check whether the table entry management unit stores the table entries that need to be searched by other pipelines, that is, the table entries that cannot be hit by the table lookup process of other pipelines.

5. The distributed MAC entry processing apparatus of claim 2, wherein the learning sub-module is provided with an entry identifier, and the entry identifier is used to distinguish an entry directly inserted for processing the learning advertisement information from an entry obtained by self-learning, so that the aging sub-module discriminates processing.

6. The distributed MAC entry processing apparatus as claimed in claim 2, wherein the lookup sub-module is configured to perform entry lookup according to a destination MAC address of the message, and generate the miss notification information when the lookup sub-modules of all the entry management units in the pipeline miss.

7. The distributed MAC table entry processing apparatus of claim 2, wherein the aging submodule is configured to delete a MAC table entry that is no longer in an active state, release a storage space, and periodically generate an aging refresh notification message to notify other pipelines of the active state of the current table entry.

8. The distributed MAC entry processing apparatus of claim 2, wherein the relay sub-module is configured to receive the learning notification message, the miss notification message, and the aging refresh notification message from the notification message bus and transmit the learning notification message, the miss notification message, and the aging refresh notification message to the corresponding sub-module, and determine whether to continue transmitting the notification message on the bus.

9. A distributed MAC entry processing method based on the distributed MAC entry processing apparatus according to any one of claims 1 to 8, wherein the method is performed in an entry management unit, and comprises:

according to the source MAC address and the input port number of the message, learning of MAC table items is carried out, learning notification information is generated and processed to interact with table item management units of other pipelines;

each flow line searches in each table item management unit according to the assigned message destination MAC address, a hit result is output, and if the hit result is not hit, miss notification information is generated;

and processing the miss notification information, and checking whether the table item management unit stores table items required to be searched by other pipelines, namely table items which cannot be searched by the table searching process of other pipelines.