CN112073206B - Method for realizing reliable multicast oriented data packet filtering based on programmable network hardware equipment - Google Patents

Abstract

The invention relates to the field of data packet processing of programmable network hardware devices in software defined networks, in particular to a data packet filtering method for reliable multicast based on the data packet processing capability of programmable network hardware devices. The receiving end can send filtering information to the SDN controller; the SDN controller controls the data processing and forwarding behavior of the programmable network hardware device, and realizes that the data packets with specific labels are sent to the specific receiving end. In order to support the upper-layer reliable multicast protocol, through the serial number function of the programmable network hardware device, two serial numbers are written to the data packet before and after the programmable network hardware device implements filtering, so that the receiver can determine that the data packet is missing due to Whether it is filtered or lost, the advantage lies in that: in the process of network transmission, the filtering operation of data packets is realized, so that each receiving end only receives the filtered data packets, and the bandwidth occupation of the receiving end is reduced.

Description

A Reliable Multicast-Oriented Packet Filtering Method Based on Programmable Network Hardware Devices

technical field

The invention relates to the field of data packet processing of programmable network hardware devices (such as programmable network hardware devices in software-defined networks), in particular to a data packet processing capability based on programmable network hardware devices, which realizes reliable multicast-oriented Packet filtering method.

Background technique

At present, a practical general reliable multicast protocol is the PGM (Reliable Multicast Programming) protocol. The basic idea is that when the receiving end receives the data packet, it judges whether there is any missing according to the sequence number on the data packet: when it is found that there is no missing, upload the data packet to the upper-layer application; when it is found to be missing, send NAK ( NegativeAcknowledgement) packet, indicating the sequence number of the missing packet. The sender retransmits the data packet according to the sequence number of the missing data packet in the received NAK data packet. The advantage of this approach is that in a multicast scenario with multiple receivers, there will not be a large number of Acknowledgement (ACK) packets (that is, for each received packet, the receiver sends feedback to the sender to confirm the packet). received) in the network to avoid ACK storms.

Packet filtering is to filter the data packets through the labels on the data packets, and its purpose is to make the upper-layer application only receive and process the data packets with specific labels. In a reliable multicast scenario, since the sender does not have a connection to each receiver, the sender cannot perform filtering unless the filter labels specified by all receivers are the same. The current common practice is that the receiving end filters the data packets after receiving all the data packets. The receiving end ensures that all data packets are received through the underlying reliable multicast PGM protocol, and then performs filtering operations on the data packets according to the filtering policy set by the upper layer application and according to the labels of the data packets (that is, the filtering operation is implemented on top of the reliable multicast protocol. ). However, the disadvantage of this method is that even if the application program on the receiving end sets a filtering policy, all data packets will be transmitted to the receiving end, which affects the system performance of the receiving end: a large number of data packets occupy bandwidth resources; packet filtering operations occupy computing resources ( thereby increasing the delay). Therefore, for the consideration of system performance optimization, the filtering operation needs to be implemented in the process of data packet transmission in the network, so as to reduce the impact on the system performance of the receiving end.

The concept of programmability of network hardware devices (such as switches) comes from Software-Defined Networking (SDN). The idea of SDN is to make the switch focus on the forwarding function by separating the control logic from the network switches (including the Layer 3 switches) and concentrating on the SDN controller. The SDN controller can deliver forwarding policies to the underlying switches through its southbound interface based on the state of the entire network and upper-layer application requirements; the switches forward and modify the received data packets based on the forwarding policies. Compared with the three-layer switch with fixed control logic in the traditional network (such as implementing OSPF protocol, etc.), by separating the control logic from it, its programmability can be greatly improved, thereby increasing the flexibility of data packet processing in the network . The ability of programmable network hardware equipment to process data packets is based on the realization of the underlying chip; due to the differences of chips, there is no unified standard for the ability of programmable network hardware equipment to process data packets. The data packet processing capability of the programmable network hardware device considered in the present invention includes: forwarding the data packet based on the label and assigning a serial number to the data packet.

SDN-based programmable network hardware devices can flexibly implement filtering operations on data packets in the network. When the receiving end has filtering requirements for the data packets, the receiving end can send corresponding filtering request information to the SDN controller. After the SDN controller receives the request, it sends the data packet processing policy to the underlying programmable network hardware device, and then realizes the filtering operation of the data packet in the network.

However, the packet filtering method based on programmable network hardware equipment needs to solve the following problems in the reliable multicast scenario: the filtering operation in the programmable network hardware equipment is implemented in the lower layer of the reliable multicast protocol, so when the receiver runs When the program of the reliable multicast protocol finds that the data packet is missing, it needs to judge whether the loss is caused by the filtering operation in the network or the data packet is lost during the transmission process. For a reliable multicast protocol, when a packet loss is found, a NAK packet needs to be sent to trigger retransmission at the sender; while for packet filtering, no action is required. If all missing data packets are judged as filtering, it will affect the reliability of data transmission; if all missing data packets are judged as lost, it will greatly increase the system delay. How to allow the receiver to receive enough information to determine whether the missing data packets are lost or filtered is the most important problem in realizing the data packet filtering method based on programmable network hardware devices and reliable multicast.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the deficiencies of the prior art, to provide a method for filtering data packets based on programmable network hardware equipment in network transmission, and to solve the problem of judging the lack of data packets by the receiving end in a reliable multicast scenario.

In order to achieve the above purpose, a method for implementing reliable multicast-oriented data packet filtering based on programmable network hardware devices is designed. The device is connected, the SDN controller is connected to the programmable network hardware device through the southbound interface, which is used to issue data packet processing policies to it, and the two receiving ends are connected to the SDN controller to send filtering information to it. The specific method is as follows :

1. The receiving end sends filtering information to the SDN controller;

2. After the SDN controller receives the filtering information from the receiving end, it sends the data packet processing policy to the underlying programmable network hardware device through its southbound interface;

3. When sending data packets, the sender adds different labels to different data packets according to the upper-layer application policy, and sends the data packets with labels to the configured programmable network hardware device;

4. The programmable network hardware device processes the received data packets according to the issued data packet processing strategy: the first step is to write the pre-filtering sequence number to each data packet before filtering; the second step, according to the data The packet label matches different matching items; the third step is to write different filtered sequence numbers to the data packets by matching different matching items; the fourth step is to forward the data packets to different ports by matching different matching items. Program network hardware devices.

The steps of the receiving end judging the data packet loss after receiving the data packet include:

1. When receiving two consecutive data packets i and j before and after, obtain the original sequence number OS, the sequence number PFS before filtering, and the sequence number AFS after filtering, namely (OS _i , PFS _i , AFS _i ) and (OS _j ) , PFS _j , AFS _j );

2. Calculate the value of AFS _j -AFS _i ;

3. If the value of AFS _j -AFS _i is not equal to 1, it means that there is packet loss after filtering, and the judgment is ended;

4. Calculate the value of ( _PFSj - _PFSi )-( _OSj -OSi ₎ ;

5. If the value of (PFS _j -PFS _i )-(OS _j -OS _i ) is not equal to 0, it means that there is packet loss before filtering, otherwise there is no packet loss, and the judgment is ended.

When the sender sends a data packet, it will contain the corresponding original sequence number OS, that is, a continuously increasing number. The programmable network hardware device maintains a counter for each matching item, representing how many data packets match the item. , since the value of the counter will automatically increase with the matching of the data packet, which has the property of continuous increment, so by writing this value into the data packet, the operation of serial numbering of the data packet can be realized.

Before the programmable network hardware device matches the data packet label, it writes the PFS to the data packet, and calculates whether the difference between the PFS and the OS is equal through the receiving end, so that it can judge whether there is a data packet loss before filtering.

After the programmable network hardware device matches the label of the data packet, it writes the AFS to the data packet, and calculates whether the AFS increases continuously through the receiving end, so that it can judge whether there is a data packet loss after filtering.

Compared with the prior art, the present invention has the following advantages:

1. It can reduce the bandwidth occupancy rate and reduce the delay of the system;

2. With the help of programmable network hardware equipment, the function of serial numbering data packets can effectively solve the problem of the receiver's judgment of missing data packets in the reliable multicast scenario, so it will not affect the normal operation of the upper-layer reliable multicast protocol.

Description of drawings

Figure 1 is a schematic diagram of the overall architecture of packet filtering in a network;

Fig. 2 is the basic schematic diagram of the internal forwarding strategy of the programmable network hardware device;

Fig. 3 is the flow chart of the programmable network hardware device filtering and writing serial number;

Fig. 4 is the flow chart that the receiving end carries out packet loss judgment;

FIG. 5 is a schematic diagram of the programmable network hardware device completing filtering and writing serial numbers.

Detailed ways

The present invention will be further described below in conjunction with examples. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to Figure 1, it is a schematic diagram of filtering data packets in the network by using SDN programmable network hardware equipment, which includes a sender and two receivers (A and B), and the sender and the two receivers pass through the programmable network. Hardware device connection. The SDN controller is connected to the programmable network hardware device through the southbound interface, and can issue data packet processing policies to it. The two receivers are connected to the SDN controller to which filtering information can be sent. The whole process is as follows:

1. The receiving end sends filtering information to the SDN controller, that is, sending the data packet of a specific label to a specific receiving end;

2. After the SDN controller receives the filtering information from the receiving end, it sends a forwarding policy to the underlying programmable network hardware device through its southbound interface, that is, it controls the forwarding of the data packet according to the label of the data packet;

3. When sending data packets, the sender adds different labels to different data packets according to the upper-layer application policy, and sends the data packets with labels to the configured programmable network hardware device;

4. According to the internal configuration and the label on the data packet, the programmable network hardware device forwards it, and sends the data packet with the specified label to the specified receiver to complete the filtering operation of the data packet.

Referring to FIG. 2, it is a basic schematic diagram of the internal forwarding strategy of the programmable network hardware device. The forwarding strategy inside the programmable network hardware device can be regarded as a matching table, wherein the matching field is the label in the data packet; the action after matching is port forwarding. When a packet arrives at a programmable network hardware device, it contains a tag added by the sender. The internal matching table of the programmable network hardware device performs matching according to the data packet label; when the matching is successful, the corresponding action is performed. As shown in Figure 2, when the label of the data packet is A, the first item in the table can be successfully matched, and then forwarded from port 1 to finally reach the receiving end A; while the data packet with the label B will only be forwarded to the receiving end End B will not be forwarded to receiving end A to achieve the filtering effect. At the same time, through the real-time delivery of the forwarding policy of the underlying programmable network hardware device through the SDN controller, the real-time configuration of the filtering policy can be realized and the system flexibility can be improved.

The above process is to use the forwarding function of the programmable network hardware device to realize the operation of filtering the data packets. In order to solve the problem of judging the lack of data packets, the programmable network hardware device will be used to serialize the data packets. Based on the reliable multicast protocol, in order to enable the receiver to determine whether the data packet is lost during the transmission process, the data packet will contain the corresponding serial number when it is sent (to distinguish it from the subsequent serial number, it is called the original serial number here, OriginalSequence, OS), that is, a continuously increasing number, such as 1, 2, 3, etc. Similar functions can also be implemented in programmable network hardware devices. The programmable network hardware device maintains a counter for each matching entry in the matching table, representing how many packets match the entry. Since the value of the counter will automatically increase with the matching of the data packets, it has the property of continuous increment, so by writing the value into the data packets, the operation of serial numbering of the data packets can be realized.

Referring to FIG. 3, FIG. 3 shows the operation flow performed in the programmable network hardware device in order to solve the problem of judging the lack of data packets. First, the first step of the process is to serialize all received packets. Since this operation occurs before the filtering operation, the generated sequence number is called Pre-Filter Sequence (PFS). Through the combination of PFS and OS, it can be judged whether packet loss occurs before filtering. After the writing of the PFS is completed, the filtering operation is performed based on the label on the data packet (ie, the second step of the process). Since each receiver can specify multiple filter conditions (ie, multiple filter labels), and form a filter group, the data packet can determine which filter group it belongs to during the process from matching to the end. Finally, through the mapping between the filter group and the receiver, it can be determined which port should be sent out. In order to support scalability, the filter group label needs to be written after determining which filter group a data packet belongs to, so that subsequent general switches can forward according to the filter group label (so when the programmable network device is directly connected to the receiving end, this parameter can be omitted). step). When multiple receivers have the same filter conditions, the number of filter groups will be less than the number of receivers. The third step in the process is to assign sequence numbers to packets belonging to different filter groups. Since this operation occurs after the filtering operation, the generated sequence number is called the After-Filter Sequence (AFS). Through AFS, it can be judged whether data packets are lost after filtering, that is, in the subsequent network transmission process.

When the data packet arrives at the receiving end, the missing of the data packet can be judged according to the OS, PFS and AFS on each data packet. The flow of this judgment is shown in FIG. 4 . First, according to two consecutively received data packets i and j, the corresponding values of OS, PFS and AFS can be obtained. For data packet i, there are (OS _i , PFS _i , AFS _i ); (OS _j , PFS _j , AFS _j ). Secondly, calculate the value of AFS _j -AFS _i , if it is not equal to 1, it means that there is a packet loss between packets i and j. The principle is relatively simple: for each filter group, when no packets are lost, the AFS value will increase continuously. Therefore, when any two consecutive data packets i and j before and after, if the value of AFS _j -AFS _i is not equal to 1, it means that there is a data packet loss between the data packets i and j. It needs to be handed over to the upper-layer reliable multicast protocol for subsequent operations. If the value of AFS _j -AFS _i is equal to 1, it is necessary to further judge whether a packet loss occurs before filtering. To determine whether data packet loss occurs before filtering, the value of (PFS _j -PFS _i )-(OS _j -OS _i ) needs to be calculated. If it is equal to 0, it means that no data packet loss occurs. The principle is: for the receiving end, if the PFS of the received data packets is continuous and the OS is not continuous, it means that there must be packet loss before the data packets reach the programmable network hardware device; if the PFS is not continuous, it means that data packet filtering occurs. The number is the difference between the PFS (when the AFS is continuous), and if the packet loss occurs before filtering, the difference between the PFS and the OS must not be equal (when the AFS is continuous, that is, there is no loss after filtering), so By calculating the value of (PFS _j -PFS _i )-(OS _j -OS _i ), it can be judged whether packet loss occurs before filtering.

For the reliable multicast scenario, the present invention mainly considers the scenario of original data packet transmission. However, for the processing procedures such as NAK and retransmission of data packets, the content specified by the existing reliable multicast protocol is still adopted. Since the reliable multicast protocol type fields of the NAK packet header, retransmission packet header, etc. (that is, the type of the data packet, such as original, retransmission, etc.) are different from the type field in the original packet header, the receiving end can distinguish and take Different logic for processing.

Embodiment 1, consider the scenario in FIG. 1 , that is, a sender, an SDN controller, a programmable network hardware device, and two receivers A and B. Receiver A(B) sends filtering information to the SDN controller specifying that it only accepts packets with label A(B). The SDN controller configures the programmable network hardware device to process related data packets according to the received filtering information. When the sender sends a data packet, it will add a label A or B to the data packet. The present invention does not make special provisions for the communication protocol between the receiving end and the SDN controller and between the SDN controller and the programmable network hardware device. For the former, the SDN controller needs to know which data packets with which labels need to be sent to which receiver; for the latter, the communication between the SDN controller and the underlying programmable network hardware devices can use the standard SDN southbound interface.

After the data packet with the label is sent from the sender, operations such as adding a sequence number and forwarding need to be performed on the programmable network hardware device. The corresponding programmable network hardware device configuration is shown in Figure 5. The configuration contains two match tables (Table 1 and Table 2). When a data packet arrives at the programmable network hardware device, it is first matched in Table 1. Table 1 contains a match item, and its match label field is a pass match, that is, all data packets can match this item. The action of this match is to write the pre-filter sequence number to the corresponding field of the packet. This function utilizes the counter function of the programmable network hardware device, that is, records the number of data packets matching this item (the counting function can be completed automatically), and writes the value of the counter into the corresponding field of the data packet.

After completing the writing of the serial number before filtering, it is necessary to filter the data packet and write the serial number after filtering. Table 2 of Figure 5 completes the latter two processes. When the data packet completes the matching of Table 1, it reaches Table 2. Table 2 contains two matches, corresponding to the two labels A and B, respectively. Each match in Table 2 implements two operations: packet filtering and writing of filtered sequence numbers. For a match item, by setting the corresponding match field, packet filtering can be achieved. The first matching label field in Table 2 is A, that is, only the data packets with the A label can match this item; similarly, the second item can only match the data packets with the B label. The data packet sending ports of the first item and the second item correspond to receivers A and B respectively, so that receiver A can only receive data packets with label A; receiver B can only receive data with label B Bag. Similar to the writing of the sequence number before filtering, the action of each matching item in Table 2 also includes the writing of the sequence number after filtering, that is, before the packet is forwarded, the sequence number after filtering needs to be written into the corresponding field of the data packet. Since the counter corresponds to each match, it is possible to maintain a different filtered sequence number for each filter group. Finally, all data packets will contain two new sequence numbers PFS and AFS in addition to the original sequence number (OS) written by the sender after the processing in Table 1 and Table 2.

Consider the data packets and labels sent by the sender as shown in the following table:

Table 1: Packet OS and label sent by sender

Original Serial Number (OS) packet label 1 A 2 A 3 B 4 A 5 B

The sender sends a total of five data packets (that is, the sequence numbers are from 1 to 5), and the corresponding labels are A, A, B, A, and B, respectively. Therefore, receiver A should only receive packets with OS 1, 2, and 4; receiver B should only receive packets with OS 3 and 5.

The following table shows the corresponding data packets written to PFS and AFS after passing through programmable network hardware devices:

Table 2: Writing to PFS and AFS

Since there is no data packet loss when the data packet reaches the programmable network hardware device, the value of PFS is consistent with the value of OS. The value of AFS increases continuously for different filter groups (ie, label A or B), that is, for the data packet of label A, the AFS value is 1, 2, and 3; for the data packet of label B, the AFS value is 1 and 2.

Finally, after the data packet is forwarded to different receivers according to the label, the receivers A and B receive the data packets as shown in the following table:

Table 3: Packets received by the receiver

The receiving end analyzes the received data packet according to the process of FIG. 4, and can determine that the data packet is not lost during the transmission process. For example, consider that the receiver receives two consecutive data packets with OS 2 and 4, first calculate the difference between their AFS as 1, indicating that no packet loss occurs after filtering; secondly, calculate (PFSj-PFSi)-(OSj-OSi) equal to zero ( The corresponding OS of the data packet j is 4; the corresponding OS of the data packet i is 2), indicating that there is no packet loss before filtering.

As can be seen from Table 3, when the packet loss occurs after filtering, it can be judged simply by the value of AFS: when the AFS value of the received packet does not continuously increase, it can be judged that the packet is lost after filtering.

When a packet is lost before filtering, it can be judged by the value of PFS. The following table shows the scenarios in which packets with OS 2 are lost before reaching the programmable network hardware device:

Table 4: Packets are lost before filtering

OS packet label 1 A 3 B 4 A 5 B

Since the programmable network hardware device does not analyze the OS value of the received data packet, but only performs a simple sequence number operation, the value of the PFS will not be consistent with the value of the OS. The following table shows the PFS and AFS written by the data packets through the programmable network hardware device:

Table 5: Writing to PFS and AFS

OS PFS packet label AFS 1 1 A 1 3 2 B 1 4 3 A 2 5 4 B 2

Compared with Table 2, the value of PFS is not consistent with the value of OS due to packet loss before arrival. Therefore, when the receiving end receives the data packet, it can determine whether the data packet is lost before filtering by calculating (PFSj-PFSi)-(OSj-OSi). The data packets received by receivers A and B are as follows:

Table 6: Packets received by the receiver

Receiver A calculates (3-1)-(4-1) is not 0, so it can be judged that there is a loss of other data packets between the two data packets; while receiver B calculates (4-2)-(5-3 ) is 0, so no additional NAK packets are sent to the sender.

Claims (5)

1. A method for realizing reliable multicast-oriented data packet filtering based on programmable network hardware equipment is characterized by comprising a sending end and two receiving ends, wherein the sending end and the two receiving ends are connected through the programmable network hardware equipment, an SDN controller is connected with the programmable network hardware equipment through a southbound interface and used for issuing a data packet processing strategy to the SDN controller, and the two receiving ends are connected with the SDN controller and used for sending filtering information to the SDN controller, and the specific method comprises the following steps:

(1) a receiving end sends filtering information to an SDN controller;

(2) after receiving the filtering information of the receiving end, the SDN controller issues a data packet processing strategy to bottom-layer programmable network hardware equipment through a southbound interface of the SDN controller;

(3) when a sending end sends a data packet, adding different labels to different data packets according to an upper application strategy, and sending the data packet with the label to the configured programmable network hardware equipment;

(4) the programmable network hardware equipment processes the received data packet according to the issued data packet processing strategy: firstly, writing a sequence number before filtering into each data packet before filtering; secondly, matching different matching items according to the data packet labels; thirdly, writing different filtered sequence numbers into the data packet by matching different matching items; and fourthly, forwarding the data packet to different ports to leave the programmable network hardware equipment by matching different matching items.

2. The method according to claim 1, wherein the step of determining packet loss at the receiving end after receiving the data packet comprises:

(1) when two consecutive data packets i and j are received, the original serial number OS, the serial number PFS before filtering and the serial number AFS after filtering are obtained, namely (OS) _i , PFS _i , AFS _i ) And (OS) _j , PFS _j , AFS _j ）；

(2) Calculating AFS _j - AFS _i A value of (d);

(3) if AFS _j - AFS _i If the value of (1) is not equal to 1, indicating that packet loss exists after filtering, and ending the judgment;

(4) calculation (PFS) _j - PFS _i )-(OS _j - OS _i ) A value of (d);

(5) if (PFS) _j - PFS _i )-(OS _j - OS _i ) If the value of (1) is not equal to 0, the packet loss exists before the filtering, otherwise, the packet loss does not exist, and the judgment is finished.

3. The method as claimed in claim 1, wherein the sender will include the corresponding original serial number OS when sending the data packet, that is, a continuously increasing number, and the programmable network hardware device maintains a counter for each matching entry, which represents how many data packets match the entry, and since the value of the counter will automatically increase with the matching of the data packets and has the continuously increasing property, the data packet can be serialized by writing the value into the data packet.

4. The method as claimed in claim 2, wherein the programmable network hardware device writes PFS into the packet before matching the packet tag, and calculates whether the difference between PFS and OS is equal through the receiving end, so that it can determine whether there is a packet loss before filtering.

5. The method as claimed in claim 2, wherein the programmable network hardware device writes AFS into the data packet after matching the tag of the data packet, and calculates whether the AFS is continuously incremented by the receiving end, so that it can determine whether the data packet is lost after filtering.

Images