KR100637933B1 - Packet forwarding method and apparatus in router of subscriber access network - Google Patents

Abstract

A packet forwarding method in a subscriber access network connecting device and an apparatus thereof are provided to extract an IP(Internet Protocol) head of a packet one time only, and to process QoS(Quality of Service) profile information acquisition, forwarding information acquisition, and virtual queue number acquisition in parallel by using multiple engines, thereby improving data processing performance. A receiving buffer(31) receives inputted packets. A head extractor(32) extracts a head from the buffer(31). The first to third forwarding engines(33-35) take charge of packet classification, forwarding information acquisition, and packet queueing information acquisition, then simultaneously acquire classification, forwarding information, and queueing information for predetermined packets if head information of the packets is delivered, and configure meta data of the packets. A meta data storage(36) stores metal data structures for each packet configured by the acquired information. A transmission buffer(37) stores the packets processed into the data, before the packets are outputted to a switch fabric. A scheduler(38) determines a transmission sequence of the stored packets by using the meta data.

Description

Packet forwarding method and apparatus in apparatus of router access subscriber

1 is a block diagram illustrating a general configuration of an access device that performs packet forwarding in a subscriber access network.

2 is a data flow diagram illustrating a conventional packet processing procedure for packet forwarding.

3 is a functional block diagram showing the configuration of a packet forwarding apparatus according to the present invention.

4 is a flowchart showing a processing procedure of a packet received from a subscriber in the packet forwarding method according to the present invention.

5 is a flowchart illustrating a packet processing procedure transmitted to a subscriber side in the packet forwarding method according to the present invention.

6 is a schematic diagram illustrating a queuing process in the packet forwarding method according to the present invention.

* Description of the symbols for the main parts of the drawings *

31: Receive Buffer

32: head extraction unit

33 to 35: 1st to 3rd forwarding engine

36: metadata storage

37: Send buffer

The present invention relates to a method and apparatus for packet forwarding in a subscriber access network capable of processing a large amount of data at high speed in packet forwarding in a subscriber access network connected to a subscriber and providing a fair service among subscribers.

The subscriber access network for connecting a plurality of subscriber stations to a core network (for example, the Internet, etc.) is provided with access equipment for transmitting and receiving packets of the multiple subscriber stations through an appropriate path. This is called a router.

FIG. 1 is a block diagram illustrating a general configuration for forwarding processing in the above-described router system. The hardware configuration includes a plurality of line cards 10 connected to a core network or a subscriber and performing packet transmission and forwarding. And a switch fabric 13 for performing data exchange between the plurality of line cards 10 and a microprocessor (MPU) 14 for controlling forwarding of the plurality of line cards 10.

The line card 10 is a part that actually performs packet forwarding. The line card 10 includes a network processor 11 that is responsible for forwarding packets and a physical layer processor 12 that is connected to a network and a subscriber. The processor 11 is configured with a packet forwarding engine (PFE) 11a for packet forwarding processing, and the microprocessor 14 has a routing function that performs a routing function and an operation and management (OAM) function of the equipment. The engine 15 is configured.

The data exchange between the line card 10 is made through the switch fabric 13, the control packet between the forwarding engine 11 and the routing engine 15 is via an internal local area network (LAN).

2 illustrates a data processing procedure performed between the input terminal and the output terminal of the line card 10 according to the conventional forwarding method in the above-described router system.

The procedure of the line card 10 is largely based on ingress data processing for receiving a packet from a media device connected to a network and outputting the packet to the switch fabric 13, and receiving a packet from the switch fabric 13 to the network. It can be classified into output egress data processing.

The input data processing procedure is performed as follows. When receiving a packet from a media device (not shown) (S211), after assembling into a data frame (S212), 6-tuples are extracted from the head of the packet to obtain QoS information, and thus the packet To classify (S213). At this time, each packet flow is classified according to the packet transmission rate and data size, and the color of each classified packet is displayed as green, yellow, and red.

Next, after the metering and statistical information on the received packet is collected and the statistical information is processed (S214), the packet forwarding information is looked up according to the destination address of the received packet (S215).

Subsequently, after performing RED (Random Early Discard) / WRED (Weighted RED) to discard the packet in advance in order to prevent the congestion of the packet (S216), the received packet is switched in a predetermined order through queue scheduling and queue management. Transmission to the output port connected to the fabric 13 (S217, S218). At this time, since the switch interface is CSIX (Common Switch Interface), the packet is delivered in a "C-frame" format, which is a cell unit of fixed size.

Next, the output data processing process is as follows.

When a packet is input from the switch fabric 13 (S221), the metering process and statistical information are collected in the same manner as in the input data processing described above with respect to the packet, and then the congestion control process by RED / WRED is performed (S222 to S223). ).

The packet not discarded in the congestion control process is encapsulated by updating and adding L2 (Layer 2) head information (S224), and then transmitting the packet to the media device in a predetermined order through scheduling and queue management (S225, S226).

In the above-described input / output data processing, a WRR (Weighted Round Robin) algorithm is generally used between ports and a DRR (Deficit Round Robin) algorithm is used between queues. The statistical information processing calculates the number of packets transmitted and received in units of flows and the number of bytes, and collects the number of packets transmitted in units of ports and interfaces, the number of bytes, and the packet arrival interval time.

As described above, the conventional forwarding procedure extracts the IP head of the received packet, classifies the packet to find QoS profile information, and then extracts a destination address from the IP head of the received packet and then looks up using the same. In this way, the forwarding information was found out, and the data was processed in a sequential flow.

Therefore, the process of extracting the head of the packet is repeatedly performed, and the delay time of packet processing is increased because the look-up for forwarding is performed after packet classification.

In addition, in the conventional queue management for packet transmission, since a method of queuing the packets in the order in which they are received is used, both subscribers that use a lot of traffic and subscribers that use less traffic during network congestion are the same. Positive packet loss is seen. Therefore, it was not possible to provide fair data traffic between subscribers.

SUMMARY OF THE INVENTION The present invention has been proposed to solve the above-mentioned conventional problems, and an object thereof is to provide a service that is capable of processing a large amount of data at high speed in packet forwarding in a subscriber access network and providing a fair service among subscribers. It is to provide a packet forwarding method and apparatus in a network.

As a means for achieving the above object, the present invention provides a subscriber access including a plurality of line cards connected to an external network to perform transmission and reception of packets, and a switch fabric for transferring packets between the plurality of line cards. A packet forwarding method in a network access device, the method comprising: extracting an IP head of the received packet when receiving a packet from an external network; Using the extracted IP head information through multiple engines, acquiring profile information for QoS and security, obtaining forwarding information, and obtaining a virtual queue index in parallel; Setting metadata describing the packet with the obtained QoS profile information, forwarding information, and virtual queue index information; And performing queuing and scheduling of the received packet with reference to the set metadata, and delivering the received packet to the switch fabric.

In addition, as another configuration means for achieving the above object, the present invention, a receiving buffer for storing the received packet; A head extracting unit extracting an IP head of a packet stored in the receiving buffer; When the head information of the received packet is transmitted from the head extractor to obtain packet classification, forwarding information acquisition, and packet queuing information for QoS and security, respectively, classification, forwarding information, and First to third forwarding engines which acquire queuing information and set the metadata of the corresponding packet; Meta data storage unit for storing the packet-specific metadata; A transmission buffer for storing the packets processed by the first to third forwarding engines before transmission; And a scheduler for determining and transmitting a transmission order between packets stored in a transmission buffer by referring to scheduling information of metadata stored in the metadata storage unit.

3 is a functional block diagram showing a packet forwarding apparatus according to the present invention, which is applied in the network processor 11 in the router system shown in FIG.

Referring to FIG. 3, the packet forwarding apparatus according to the present invention includes a receiving buffer 31 for receiving a packet input from a media device, a head extracting unit 32 for extracting a head from the receiving buffer 31, Responsible for packet classification, forwarding information acquisition, and packet queuing information acquisition for QoS and security, respectively, and when head information of a predetermined packet is transferred from the head extractor 32, the classification, forwarding information, and queuing information for the corresponding packet are simultaneously Packet-specific metadata data structure comprising first to third forwarding engines 33 to 35 to acquire and configure metadata of the corresponding packet, and information obtained by the first to third forwarding engines 33 to 35. A meta data storage unit 36 for storing the data, a transmission buffer 37 for storing the data processed by the first to third forwarding engines 33 to 34 until output to the switch fabric 13, and In the metadata storage unit 36 Using the metadata includes a scheduler 38 for determining a transmission order of the packets stored in the transmission buffer (37).

The packet forwarding apparatus having the above-described configuration extracts the head of a packet only once using multiple engines, simultaneously performs classification of the packet, look-up for acquiring forwarding information, and acquiring queuing information. As a delay of the forwarding process is prevented, the detailed action is as follows.

When a packet received from an external network (or subscriber network) through a media device is input to the network processor 10 of FIG. 1, the packet is sequentially stored in the reception buffer 31 until data processing for forwarding is completed. The head extractor 32 extracts the necessary information from the head of the received packet stored in the receive buffer 31. At this time, a 6-tuple of the IP heads of the received packet is extracted to extract a plurality of forwarding engines; That is, it transfers to the 1st-3rd forwarding engines 33-35.

The six-tuple extracted by the head extractor 32 corresponds to a source IP address, a destination IP address, a source port, a destination port, a protocol, and a differentiated service code point (DSCP).

The first forwarding engine 33 calculates the 6-to-flow hash key extracted as described above, retrieves the QoS profile table using the calculated hash, and obtains QoS profile information of a perfectly matched entry, If there is no perfect match, the default QoS profile information set in the routing system is obtained, and the second forwarding engine 34 looks up the forwarding table using the destination IP address among the 6-tuples extracted as described above. The third forwarding engine 33 obtains a virtual queue number for each subscriber using the source IP address among the extracted 6-tuples.

In the above, the QoS profile table and the forwarding table are configured in the routing processor 15 and provided to a plurality of network processors 11 and stored in the network processor 11.

When the first to third forwarding engines 33 to 35 respectively acquire QoS profile information, forwarding information, and virtual queue number, the first to third forwarding engines 33 store these in the metadata storage 36 to configure metadata describing the corresponding packet. do.

 In this case, since the first to third forwarding engines 33 to 35 need to set information in the same memory as the metadata storage unit 36, synchronization between the first to third forwarding engines 33 to 35 is necessary. . The most basic method for this purpose, according to the complexity of the first to third forwarding engine (33 to 35), the third forwarding engine (35) responsible for obtaining the virtual queue number in order of access to the metadata storage (36), The first forwarding engine 33 in charge of obtaining QoS profile information and the second forwarding engine 34 in charge of obtaining forwarding information are determined in this order.

In addition, a register for transferring information between the first to third forwarding engines 33 to 35 is provided, and whether or not metadata is configured is displayed as 0 or 1, and the first to third forwarding engines ( 33 to 35 access metadata only when the flag for the forwarding engines having priority accesses becomes 1, in which case, all three flags for the first to third forwarding engines 33 to 35 are all. If it is 1, it can be determined that the construction of the metadata is completed.

When the configuration of the meta data is completed by the first to third forwarding engines 33 to 35 as described above, the received packet is stored in the transmission buffer 37 and then the switch fabric port set through the scheduler 38. Is output. The scheduler 38 refers to scheduling information of a corresponding packet through metadata stored in the metadata storage 36, and sequentially outputs it to the switch fabric 13 according to the priority of each packet.

Table 1 below shows the data structure of the above-described metadata configured to describe the information of the packet until the packet is input and output from the network processor 11 of the router system. Meta data is generally composed of eight words (LW: Long Word) having a 32-bit size, and has a buffer, size information, forwarding information, QoS profile information, and the like.

LW beat size Field name Explanation 0 31: 0 32 buffer_next Next buffer to drive in the packet chain One 31:16 16 buffer_size Buffer size [bytes] 15: 0 16 offset The starting offset of the data in the buffer [bytes] 2 31:16 16 packet_size Total packet size including all chained buffers 15:12 4 free_list_id Free list ID for buffer 11: 8 4 rx_start Receive Status Flag 7: 0 8 header_type Header Type in Offset Bytes in Packet 3 31:16 16 input_port Input port of input side network processor 15: 0 16 output_port The output port number of the line card to which the packet is to be sent. 4 31:16 16 next_hop_id Identifier of Next Hop IP Node 15: 8 8 fabric_port Output port on switch fabric to differentiate destination directions 7: 4 4 reserved Not currently used 3: 0 4 nexthop_id_type Specify an ID in the table that looks up next_hop_id 5 31: 0 32 flow_id Flow ID (QoS Flow ID or MPLS Label / Flow ID) 6 31:16 16 class_id Relation identifier of the queue connected to the output port 15: 2 14 queue_id Virtual queue separator 1: 0 2 color_id Packet drop processing level (green, yellow, red) 7 31: 0 32 packet_next Indicator of next packet (not used in cell mode)

In the metadata of the above-described structure, forwarding information corresponds to output_port, next_hop_id, and fabric_port, which are obtained by the second forwarding engine 34 in charge of the forwarding lookup. In this case, the next_hop_id is information necessary for the packet to know the next transmission position when the packet is transmitted from the output terminal of the network processor. The next_hop_id is an index of an L2 information table such as an Ethernet head, and fabric_port is a switch fabric to which the packet is to be transmitted. 13) port number.

The QoS profile information corresponds to flow_id, class_id, and color_id, which are obtained by the first forwarding engine 33 responsible for obtaining QoS profile information. At this time, flow_id is a flow delimiter of the corresponding packet, and represents a bandwidth of traffic, class_id represents a priority during packet scheduling, and color_id is a factor representing priority of discarding a packet during congestion control.

The queue_id and the class_id correspond to the queuing and scheduling information. In this case, the queue_id represents a queue identifier for setting a virtual queue on a subscriber basis, and the class_id represents a class of packet scheduling. Here, class_id is obtained by the first forwarding engine 33, and the remaining queue_id is obtained by the third forwarding engine 35 in charge of setting the virtual queue number.

When the present invention is applied, the network processor 11 of FIG. 1 sets the metadata of the above-described structure through simultaneous operations of the first to third forwarding engines 33 to 35, and uses the information set as described above. Will be forwarded. As described above, duplicate extraction of the IP head of the packet is eliminated, and the lookup of the QoS profile table and the forwarding table is performed simultaneously, thereby significantly reducing the time delay of the forwarding procedure.

4 is a flowchart illustrating a packet forwarding method according to the present invention, which corresponds to an input data processing process in the network processor 11.

Referring to FIG. 4, when the present invention is applied, when the network processor 11 of the router system shown in FIG. 1 receives a packet from the outside through a media device (S300), the 6-tuple of the received IP packet is received. Extract (S310).

Then, through the multiple engines, the lookup and information acquisition of the QoS profile table using the extracted 6-tuple, the lookup and information acquisition of the forwarding table, and the acquisition of the virtual queue number (queue id) are performed simultaneously (S320, S330, S340).

In the process of obtaining the QoS profile information (S320), after hashing the extracted 6 tuples to obtain a hash key (S321), profile information that is completely matched by searching the QoS profile table with the calculated hash key Find (S322). As a result of the search, if there is a perfectly matched QoS profile, the corresponding QoS profile information (flow_id, class_id, color_id) is obtained and set in the metadata, and if the completely matched QoS profile does not exist, the value set as default QoS profile information Is set in the metadata (S323 to S325).

In the lookup process of the forwarding table (S330), the destination IP address is obtained from the extracted 6-tuple, and the forwarding table is searched by a longest prefix match (LPM) method (S332). As a result of the search, if there is a matching entry in the forwarding table, the next hop information (next_hop_id, output_port, fabric_port) is obtained from the entry and set in the metadata (S333, S334). On the contrary, if there is no matching entry in the forwarding table, an interrupt for the corresponding exceptional situation is sent to the routing processor 15 or the received packet is discarded (S335).

In the above, the virtual queue number acquisition process (S340), the source IP address is obtained from the extracted 6-tuple (S341), the virtual queue number set to the subscriber corresponding to the source address (S342), and the search The virtual queue number queue_id set for each subscriber is obtained and set in the metadata (S343). The current metadata structure uses a 14-bit virtual queue index for the virtual queue number (queue_id) as shown in Table 1 above. Thus, approximately 16000 subscribers can be distinguished. However, if the number of subscribers to be accommodated in the routing system increases, the structure of the metadata should be extended.

In this case, it is preferable to set the order of access to the metadata data structure in order to synchronize the multiple engines, and in the present embodiment, the metadata data structure according to the complexity of each engine performing the process (S320, S330, S340). The order of access is determined in order of virtual queue number setting, QoS profile information setting, and next hop information setting.

By the above-described processes (S320, S330, S340), when the construction of the metadata indicating the information of the received packet is completed (S350), it is possible to check the traffic bandwidth (measurement from the flow_id of the meta data) through the metering. Check whether the band is available (S360), and if the check result is included in the available band, the packet is delivered to the output port of the switch fabric 13 through scheduling (S380). In the above scheduling, the priority is determined according to scheduling information (class_id) in the meta data, and is output in the order of high priority.

In addition, the present invention, in order to ensure fairness for each subscriber in performing packet forwarding in the router system, unlike the conventional method of queuing packets in order of receiving the packets, assigning a virtual queue for each subscriber so that the data is fair among the subscribers. To provide traffic. The virtual queue assignment can be applied to the queuing step of both the ingress data processing and the egress data processing of the line card 10.

FIG. 5 is a flowchart showing a forwarding procedure for allocating virtual queues for each subscriber according to the present invention, and exemplifies output data processing for transferring data input from the switching fabric 13 to a subscriber through a media device. FIG. 6 is a diagram illustrating the processing of the virtual queue allocation process shown in the flowchart of FIG. 5.

Referring to FIG. 5, the network processor 11 receives a packet having a C-frame format from the switch fabric 13 (S510), and acquires the virtual queue number of FIG. 4 from metadata of the corresponding packet (S340). In operation S520, a virtual queue number (queue_id), which is set separately for each subscriber, is obtained, and a virtual queue is allocated for each subscriber. By allocating a virtual queue for each subscriber in this way, the number of buffers for each subscriber is set to be the same.

Then, the weight is set by dividing the quality service and the general service in the virtual queue allocated to each subscriber (S530).

6, for example, subscribers 1 and 2 are subscribers who have registered quality service (QoS service) and two general services, and subscriber n is a subscriber who has only registered quality service without registering quality service. Here, quality service is a service that must be guaranteed quality sensitive to the band, delay variation, loss rate, etc. of IP broadcasting, VoD, video call, VoIP, etc., and general service is sensitive to band, delay, loss, etc. like P2P. It does not receive general Internet service. In the embodiment of FIG. 6, a weight of five times the general service is assigned to the quality service, so that the service is given priority over the general service.

That is, according to the above, by allocating virtual queues for each subscriber and setting the same number of buffers, quality services can be preferentially processed over general services while maintaining fairness among subscribers. At this time, the general service serves as a best-effort (BE) transmission method within a range allowed by network resources.

As described above, after performing queuing, scheduling is performed to transmit a packet having a high priority (S540 and S550).

At this time, the scheduling scheme can be applied to the WRR (Weight Round Robin) and DRR (Deficit Round Robin) algorithm that can be easily implemented, while ensuring maximum fairness among subscribers.

Such subscriber-specific virtual queue allocation and service-specific weighting are applied to packets input from the media device and output to the switch fabric according to the flow identifier (flow_id) and the class identifier (class_id) of the metadata configured by the method of FIG. The same can be applied.

As described above, the present invention, in performing the forwarding to guarantee the QoS of the subscriber access network, extracts the IP head of the packet only once, obtains QoS profile information, obtains the forwarding information, and obtains the virtual queue number using multiple engines. By processing the data in parallel, there is an excellent effect of improving data processing performance and enabling high-speed data processing.

In addition, by setting a virtual queue for each subscriber using a subscriber address, a virtual queue can be allocated to each subscriber to provide fairness for queue preemption between subscribers in the network. Analyze and understand the services used by

Claims (15)

A packet forwarding method in a subscriber access network access equipment comprising a plurality of line cards connected to an external network and performing packet transmission and forwarding processing, and a switch fabric for transferring packets between the plurality of line cards, If the line card receives a packet, extracting an IP head of the received packet; Using the extracted IP head information through multiple engines, acquiring profile information for QoS and security, obtaining forwarding information, and obtaining a virtual queue index in parallel; And Setting metadata describing the packet with the obtained QoS profile information, forwarding information, and virtual queue index information; And And queuing and scheduling received packets with reference to the set metadata. The method of claim 1, wherein the setting of the metadata is performed. The packet forwarding method in the subscriber access network access equipment, characterized in that the virtual queue index information, QoS profile information, forwarding information. The method of claim 1, wherein extracting the IP head Extracting a source IP address, a destination IP address, a source port, a destination port, a protocol, and 6-tuple of a differential service code point (DSCP) from a received packet. . The method of claim 3, The profile information acquisition for QoS and security is performed by searching the QoS profile table by perfect matching with the extracted 6-tuple calculated hash key to obtain completely matched QoS profile information. Packet forwarding method on the device. The method of claim 3, The acquiring of the forwarding information includes searching for a forwarding table with a destination IP address among the extracted 6-tuples to obtain matching forwarding information. The method of claim 3, The acquiring the virtual queue index information may include retrieving a subscriber queue number set in units of subscribers from the extracted 6-tuple by using a source IP address to obtain a matching subscriber queue number. Packet forwarding method on the device. The method of claim 4, wherein The QoS profile information includes a flow identifier (flow_id) of a packet indicating a traffic band, a class identifier (class_id) indicating a class of packet scheduling, and a factor (color_id) indicating a priority when discarding a packet in a congestion control process. Packet forwarding method in a subscriber access network access equipment, characterized in that. The method of claim 5, The forwarding information includes a next hop identification value (next_hop_id) when a packet is transmitted from a line card to a network, an output port number (output_port) of a line card to which a packet is to be transmitted, and a port number of a switch fabric to which the packet is to be transmitted ( packet forwarding method in a subscriber access network access device comprising: fabric_port). The method of claim 6, The virtual queue index information is a packet forwarding method in a subscriber access network access equipment, characterized in that the queue identifier (queue_id) for setting a virtual queue on a subscriber basis. The method of claim 1, wherein performing queuing and scheduling of the packet comprises When the line card receives a packet, allocating a virtual queue on a subscriber basis by referring to virtual queue index information of metadata set in the received packet; And Scheduling the virtual queue allocated by the subscriber, and transmitting the packet allocated to each virtual queue according to a scheduling order. The method of claim 10, And assigning weights to the packets of the virtual queue set in the subscriber unit by dividing the quality service and the general internet service. And scheduling the quality service packet in preference to the general service packet according to the weight. A reception buffer for storing the received packet; A head extracting unit extracting an IP head of a packet stored in the receiving buffer; When the head information of the received packet is transmitted from the head extractor to obtain packet classification, forwarding information acquisition, and packet queuing information for QoS and security, respectively, classification, forwarding information, and First to third forwarding engines which acquire queuing information and set the metadata of the corresponding packet; A metadata storage unit for storing packet-specific metadata; A transmission buffer for storing the packets processed by the first to third forwarding engines before transmission; And And a scheduler for determining and transmitting a transmission order between packets stored in a transmission buffer with reference to scheduling information of metadata stored in the metadata storage unit. The method of claim 12, The first to third forwarding engine, Subscriber access network access equipment, characterized in that to access the metadata storage in order of the third forwarding engine responsible for packet queuing information acquisition, the first forwarding engine responsible for packet classification, the second forwarding engine responsible for forwarding information acquisition Packet forwarding device in. The method of claim 13, The metadata storage unit includes first to third flags indicating whether access to each of the first to third forwarding engines is provided. And the first to third forwarding engines set metadata information in order with reference to the first to third flag values, respectively. The method of claim 13, The head extracting unit extracts 6-tuples of a source IP address, a destination IP address, a source port, a destination port, a protocol, and a differentiated service code point (DSCP) from the IP head of the received packet, and includes first to third forwarding engines. Packet forwarding apparatus in subscriber access network access equipment, characterized in that delivered to.

Images