Multicast replication method and device for cross-chip process processing unit

Abstract

The present invention discloses a multicast replication method and device for cross-chip process processing units, wherein the method comprises: a source-end chip process processing unit receives a multicast message; a pre-replication engine in the source-end chip process processing unit generates an index item according to the multicast information and message feature information of the multicast message, and matches a preset replication table to determine the distribution of valid members of the multicast message; the replication table is pre-set in the chip by the upper-layer system according to the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; the pre-replication engine in the source-end chip process processing unit selects the replication multicast message according to the distribution information of the valid members, and sends it to the multicast table entry module in the corresponding destination-end chip process processing unit. The method and device can solve the problem of replication bandwidth waste in multi-PP concurrent design in LAG scenarios.

Description

Technical Field

[0001] The present invention relates to computer network communication technology, and in particular to a multicast replication method and device for a cross-chip process processing unit. Background Art

[0002] Multicast communication is widely used in many large networks, such as operator networks, enterprise networks, industrial networks, etc. However, for high-bandwidth chips with multi-slice designs, multicast applications are mainly concentrated in operator networks.

[0003] With the development of the Internet, the bandwidth of a single switching chip can now reach 51.2Tbps. The reason why the bandwidth of the switching chip can be so large is due to the concurrent architecture of the multi-chip process processing unit (Processer Pipeline, PP). Generally, each PP has an upper limit for its bandwidth processing, which depends on the main frequency of the chip process. The main frequency of the most advanced network switching chip is about 1.4GHz, and the bandwidth of each PP it can bring is 3.2Tbps. As a result, if you want to increase the bandwidth of the switching chip by increasing the chip main frequency, the research and development progress will be very slow. Therefore, the main way to increase the bandwidth of the switching chip is to rely on the concurrent design of multiple PPs.

[0004] This solution is excellent in terms of both performance and cost, but there are still some problems in specific applications. For example, in scenarios where Link Aggregate (LAG) is used, for users, the duplicate messages forwarded by the chip as a whole sometimes seem to be unable to reach the expected switching chip bandwidth processing capacity. Summary of the invention

[0005] The technical problem to be solved by the embodiments of the present invention is how to solve the problem of waste of replication bandwidth in multi-PP concurrent design in a LAG scenario.

[0006] In order to solve the above problems, the technical solution provided by the embodiment of the present invention is as follows: a multicast replication method across chip process processing units, which is applied to the aggregation port scenario, including: the source chip process processing unit receives the multicast message; the pre-replication engine in the source chip process processing unit generates an index item based on the multicast information and message feature information of the multicast message, and matches a preset replication table to determine the distribution of valid members of the multicast message; the replication table is pre-set in the chip by the upper-level system according to the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; the pre-replication engine in the source chip process processing unit selects the replication multicast message according to the distribution information of the valid members, and sends it to the multicast table entry module in the corresponding destination chip process processing unit.

[0007] Optionally, the above-mentioned multicast replication method of the cross-chip process processing unit, wherein the multicast information of the multicast message includes a multicast Group value; the message feature information includes a multicast message Hash value; the generating of index items according to the multicast information and the message feature information of the multicast message includes: taking the multicast Group value with a high 10-bit width and the Hash value with a low 8 bits in the multicast message to generate an Index index item.

[0008] Optionally, the above-mentioned multicast replication method across chip process processing units, wherein, after the multicast table entry module in the corresponding destination chip process processing unit is sent, it also includes: the multicast table entry module in the destination chip process processing unit copies the multicast message to the local multicast group members, and then forwards the multicast message from the corresponding port.

[0009] Optionally, the above-mentioned multicast replication method of the cross-chip process processing unit is characterized in that the aggregation port scenario is: only one multicast group exists in the aggregation port of the cross-chip process processing unit.

[0010] Optionally, in the above-mentioned multicast replication method of the cross-chip process processing unit, the upper-layer system is a software development kit software.

[0011] In order to solve the above-mentioned technical problems, an embodiment of the present invention also discloses a multicast replication device across chip process processing units, which is applied in an aggregation port scenario and includes: a receiving unit, arranged in the source-end chip process processing unit, for receiving multicast messages; a processing unit, a pre-replication engine arranged in the source-end chip process processing unit, for generating index items according to the multicast information and message feature information of the multicast message, and matching a preset replication table to determine the distribution of valid members of the multicast message; the replication table is pre-set in the chip by the upper-layer system according to the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; a first forwarding unit, a pre-replication engine arranged in the source-end chip process processing unit, for selecting and replicating the multicast message according to the distribution information of the valid members, and sending it to the multicast table entry module in the corresponding destination-end chip process processing unit.

[0012] Optionally, the above-mentioned multicast replication device of the cross-chip process processing unit, wherein the multicast information of the multicast message includes the multicast Group value; the message feature information includes the multicast message Hash value; the processing unit takes the multicast Group value with a high 10 bits of width and the Hash value with a low 8 bits in the multicast message to generate an Index index item.

[0013] Optionally, the above-mentioned multicast replication device across chip process processing units further includes: a second forwarding unit, a multicast table entry module arranged in the destination chip process processing unit, for copying multicast messages to local multicast group members, and then forwarding multicast messages from corresponding ports.

[0014] Optionally, in the above-mentioned multicast replication device of the cross-chip process processing unit, the aggregation port scenario is: only one multicast group exists in the aggregation port of the cross-chip process processing unit.

[0015] Optionally, in the above-mentioned multicast replication device of the cross-chip process processing unit, the upper-layer system is a software development kit software.

[0016] Compared with the prior art, the technical solution of the present invention has the following advantages:

[0017] In the multi-PP architecture and LAG scenario, the embodiment of the present invention uses a pre-replication engine (Pre-Multicast Expansion Table, Pre Met) at the source PP to use the multicast information and message information in the multicast message as an index to determine whether the destination PP actually has a demand for replicating and forwarding the multicast message, processes the routing strategy in advance, and then decides whether to replicate the message to the corresponding PP. In this way, the LAG member egress can be determined based on the pre-replication, thereby avoiding the problem of wasted replication bandwidth and achieving efficient bandwidth utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the network topology environment used for multicast message processing in the existing multi-PP scenario;

[0019] Figure 2 This is a diagram of the network topology environment used to handle the difficulties encountered in multicast message processing in existing multi-PP scenarios;

[0020] Figure 3 is a flow chart of a cross-PP multicast replication method in an embodiment of the present invention;

[0021] Figure 4 is a schematic diagram of a replication table in a multicast replication method across PPs in an embodiment of the present invention;

[0022] Figure 5 is a logical schematic diagram of a cross-PP multicast replication method in an embodiment of the present invention;

[0023] Figure 6 It is a structural diagram of a cross-PP multicast replication device in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] Currently, the main way to increase the bandwidth of the switching chip is to rely on the concurrent design of multiple PPs. This solution is excellent in terms of both performance and cost, but there are still some problems in specific applications. For example, in the scenario where LAG is applied, for users, the copy messages forwarded by the chip as a whole sometimes seem to be unable to reach the expected switching chip bandwidth processing capacity.

[0025] like Figure 1 As shown in the figure, it is a 4PP architecture of a 12.8Tbps chip. It can be seen that in addition to PP, the replication capacity of Met (Multicast Expansion Table) which is mainly responsible for message replication is also limited to only one-fourth, 3.2Tbps. If you want to achieve a replication capacity of 12.8Tbps, each PP needs to have an independent replication engine and a matching pre-replication engine (Pre-Multicast Expansion Table, Pre Met).

[0026] When the ports where multicast messages need to be replicated appear on different PPs, the Pre Met at the source end will first be replicated to the PP corresponding to the egress, and then the Met in each PP will replicate the local multicast group members. Figure 1 As shown in the figure, PP1 has messages that need to be copied to ports 8 and 9 on PP2, and ports 13 and 15 on PP3. Therefore, Pre Met1 on PP1 needs to copy two messages to PP2 and PP3 respectively, and then Met2 on PP2 and Met3 on PP3 each copy two messages to the local corresponding ports and forward them out.

[0027] This solution is excellent in terms of performance and cost, but it may cause problems in some scenarios, such as when dealing with LAG ports. Figure 2 As shown: Assuming that port 7 and port 13 form a LAG port, the message copied to this LAG port can only select a port to go out through the hash value. However, since Pre Met cannot distinguish which exit is selected by the hash value, a problem occurs.

[0028] In order to solve the problem of multicast replication to cross-PP LAG, the hash value can be calculated after the Met of each destination PP, and it can be determined whether the port selected by the hash is local. If it is local, it will be forwarded; if not, it needs to be discarded. However, since each Met can only determine whether the member's egress is local after the replication message is completed, the part that is not local and is discarded has actually occupied the replication bandwidth of the Met. Therefore, from the user's point of view, the replication message sent out by the overall chip cannot reach the expected 12.8Tbps capacity.

[0029] In order to solve the above technical problems, this embodiment discloses a multicast replication method across PPs, which is applied in the aggregation port LAG scenario. In order to make the above objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Figure 3 As shown, the cross-PP multicast replication method may include:

[0030] Step S101, the source PP receives a multicast message;

[0031] Still Figure 2 Taking the topology diagram shown as an example, in this embodiment, the high-bandwidth switch chip with multi-Slice design has 4 PPs. At this time, PP1 as the source PP receives a multicast message and needs to send it to the destination PP, and sends it out through ports 8 and 9 of PP2 and ports 13 and 15 of PP3. By aggregating ports LAG, ports 8, 9, 13, and 15 are bound into one port to avoid problems such as Layer 2 loops. It can be understood by those skilled in the art that the number of PPs in the high-bandwidth switch chip with multi-Slice design can be other than 4, but any other number, which is within the scope of this embodiment.

[0032] Step S102, generating index items according to the multicast information and message feature information of the multicast message, and matching a preset replication table to determine the distribution of valid members of the multicast message;

[0033] In a specific implementation, Pre Met at the source PP end uses the multicast Group of the multicast message as the above-mentioned multicast information, and uses the Hash value as the message feature information to form an index item, reads a preset replication table, and then determines the valid members in the multicast message, that is, the distribution of LAG ports in each PP.

[0034] Because the upper layer system can obtain the correspondence between the multicast and LAG groups when creating the multicast and LAG groups, the upper layer system pre-sets a replication table in the chip according to which PPs the members of the multicast group are distributed on, thereby further defining the correspondence between each set index item and the member on the PP in the replication table, so that each message can be accurately controlled to be replicated to the valid member across the PP LAG to avoid waste. In a specific implementation, the upper layer system is a software development kit (SDK).

[0035] The above hash value can be calculated based on the content of the message, for example, through the five-tuple of the message (source IP address, source port, destination IP address, destination port and transport layer protocol). The five-tuple can distinguish different sessions, and the corresponding session is unique, so a fixed hash value can be determined. The multicast Group is obtained by searching the routing table or other tables, such as the forwarding information table (FIB), which is information that can be obtained in the normal forwarding process. The above relevant information can be obtained before PreMet at the source PP end.

[0036] In the above implementation, the conventional multicast Group can be set to 10 bits, the Hash value can be set to 8 bits, the Group is placed in the high position and the high 10 bits are taken, and the hash value is placed in the low position and the low 8 bits are taken for search. Figure 4 As shown, the 1K Group and the 256-bit Hash can form an 18-bit index item index[17:0]=[17:8][7:0], and use this index to look up the table and read the corresponding Valid table item content in the replication table. Taking the 4 PP scenario of this embodiment as an example, each item in the above replication table is 4 bits, corresponding to whether the message needs to be replicated to the four PPs in this embodiment. For example, if it does not need to be replicated to PP1, it needs to be replicated to PP2, it needs to be replicated to PP3, and it does not need to be replicated to PP4, then the Valid value corresponding to the Index is 0110, so it can be determined that the valid members of the multicast message are distributed in PP2 and PP3.

[0037] In a specific implementation, the aggregation port scenario may be that only one multicast group exists in the aggregation port of the cross-chip process processing unit. This embodiment solves the problem of wasting replication resources when the message is replicated to the cross-PP LAG. However, in the above implementation, the index composed of the hash value and the multicast group value may sometimes be relatively large. As mentioned above, the conventional hash value is 8 bits, that is, 2 8 There are 256 types, and the maximum number of multicast groups can reach 2 10 , that is, 1M. Then, in the architecture applied to 4 PPs, the added replication table needs to include 256×1M×4=1G bits of table entries, which will be a very large overhead for the network switching chip. Therefore, when this embodiment is used, it has a very high application value in the scenario where the multicast group demand is small or only part of the multicast group exists. For example, when making a Fabric chip for an OTN device, the Fabric chip can be planned so that only one multicast group exists across PP LAGs, so that the replication table entry only consumes 256 bits.

[0038] Step S103: Pre Met of the source PP selects to copy the multicast message and sends it to Met of the corresponding destination PP.

[0039] As described in step S102, the source PP Pre Met can match the replication table according to the index item to obtain the distribution information of the valid members of the multicast message at each PP end, and then selectively replicate the message to send the multicast message to the required destination PP Met.

[0040] Step S104: The Met of the destination PP copies the multicast message to the local multicast group member, and then forwards the multicast message from the corresponding port.

[0041] The copy message of the Pre Met of the source PP in the above step S103 is to copy the multicast message to the corresponding PP, and the copy object is the PP. The copy in this step S104 is to copy the multicast message to the corresponding local multicast member, thereby forming a two-level copy.

[0042] After the destination Met copies the corresponding number of messages, it performs LAG port routing calculation based on the hash value of the message, forwards the message on the port of the multicast member corresponding to the PP, and discards the message on the port that is not a multicast member.

[0043] Through the multicast replication method of a cross-chip process processing unit of this embodiment, the problem of replication bandwidth waste can be solved in a multi-PP architecture, and the efficient use of bandwidth can be achieved. At the same time, this embodiment also takes into account the table entry overhead brought about, and further selects the appropriateness of the application scenario for application when using it, which has extremely high application value.

[0044] In order to solve the above technical problems, the embodiment of the present invention also discloses a device for multicast replication of cross-chip process processing units, which can be applied to network scenarios of aggregation ports, such as Figure 6 As shown, the device for multicast replication across chip process processing units may include: a receiving unit, arranged in the source-end chip process processing unit, for receiving multicast messages; a processing unit, a pre-replication engine arranged in the source-end chip process processing unit, for generating index items based on the multicast information and message feature information of the multicast message, and matching a preset replication table to determine the distribution of valid members of the multicast message; the replication table is pre-set in the chip by the upper-layer system based on the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; a first forwarding unit, a pre-replication engine arranged in the source-end chip process processing unit, for selecting a replicated multicast message based on the distribution information of the valid members, and sending it to the multicast table entry module in the corresponding destination-end chip process processing unit.

[0045] In a specific implementation, the multicast information of the multicast message includes a multicast Group value; the message feature information includes a multicast message Hash value; the processing unit takes the multicast Group value of the upper 10 bits and the Hash value of the lower 8 bits in the multicast message to generate an Index index item.

[0046] In a specific implementation, the multicast replication device of the cross-chip process processing unit may also include: a second forwarding unit, a multicast table entry module arranged in the destination chip process processing unit, used to copy the multicast message to the local multicast group members, and then forward the multicast message from the corresponding port.

[0047] In a specific implementation, the aggregation port scenario may be that only one multicast group exists in the aggregation port of the cross-chip process processing unit.

[0048] In a specific implementation, the upper-layer system may be software development kit software.

[0049] Since the embodiment of the present invention and the first embodiment are based on the same inventive concept, the specific solutions of the present embodiment can refer to the corresponding contents of the first embodiment and will not be described in detail here.

[0050] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which can include: ROM, RAM, disk or CD, etc.

[0051] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A multicast replication method for a cross-chip process processing unit, characterized in that: Applicable to the aggregation port scenario, including: The source chip process processing unit receives the multicast message; The pre-copy engine in the source chip process processing unit generates index items according to the multicast information and message feature information of the multicast message, and matches a preset copy table to determine the distribution of valid members of the multicast message; the copy table is pre-set in the chip by the upper system according to the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; the multicast information of the multicast message includes the multicast Group value; the message feature information includes the multicast message Hash value; the generation of index items according to the multicast information and message feature information of the multicast message includes: taking the multicast Group value with a high 10-bit width and the Hash value with a low 8-bit width in the multicast message to generate an Index index item; The pre-copy engine in the source-end chip process processing unit selects the copy multicast message according to the distribution information of the valid members, and sends it to the multicast entry module in the corresponding destination-end chip process processing unit; After the multicast table entry module in the corresponding destination chip process processing unit is sent, it also includes: performing LAG port routing calculation according to the hash value of the message, the multicast table entry module in the destination chip process processing unit copies the multicast message to the local multicast group member, and then forwards the multicast message from the corresponding port.

2. The multicast replication method of the cross-chip process processing unit according to claim 1, characterized in that: The aggregation port scenario is that only one multicast group exists in the aggregation port of the cross-chip process processing unit.

3. The multicast replication method of the cross-chip process processing unit according to claim 1, characterized in that: The upper layer system is software development kit software.

4. A multicast replication device for a cross-chip process processing unit, characterized in that: Applicable to the aggregation port scenario, including: A receiving unit, provided in the source chip process processing unit, for receiving multicast messages; A processing unit, a pre-copy engine arranged in a source-end chip process processing unit, is used to generate an index item according to the multicast information and message feature information of the multicast message, and match a preset copy table to determine the distribution of valid members of the multicast message; the copy table is pre-set in the chip by the upper-layer system according to the distribution of multicast group members in each chip process processing unit and the member information of the aggregation port; the multicast information of the multicast message includes a multicast Group value; the message feature information includes a multicast message Hash value; the processing unit takes the multicast Group value of the upper 10 bits and the Hash value of the lower 8 bits in the multicast message to generate an Index item; A first forwarding unit, a pre-copy engine provided in the source-end chip process processing unit, is used to select a copy of the multicast message according to the distribution information of the valid members, and send it to the multicast entry module in the corresponding destination-end chip process processing unit; The second forwarding unit is used to perform LAG port routing calculation according to the Hash value of the message. The multicast entry module in the destination chip process processing unit is used to copy the multicast message to the local multicast group member and then forward the multicast message from the corresponding port.

5. The multicast replication device for cross-chip process processing units according to claim 4, characterized in that: The aggregation port scenario is that only one multicast group exists in the aggregation port of the cross-chip process processing unit.

6. The multicast replication device for cross-chip process processing units according to claim 4, characterized in that: The upper layer system is software development kit software.

Images