JP2002101124A - Inter-network repeating apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To guarantee minimum bandwidth for highly important data in an inter-network repeating apparatus that repeats variable-length packets. SOLUTION: Guaranteed traffic schedulers (43-1 to 43-3) and best-effort traffic schedulers (44-1 to 44-3), both arranged in each input interface, respectively control the takeoff rates of the cues stored in guaranteed traffic input cues (42G-11 to 13, 42G-21 to 23, 42G-31 to 33) and in best-effort traffic input cues (42B-11 to 13, 42B-21 to 23, 42B-31 to 33). By the above arrangements, a guaranteed band is guaranteed for each interface, and a takeoff rate of the guaranteed data from each input interface can be set in the guaranteed band.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inter-network relay device for interconnecting a plurality of LANs (local area networks) and WANs (wide area networks) and relaying data between them.

[0002]

2. Description of the Related Art FIG. 5 is a general internal block diagram of a router which is a relay device between networks.

As shown in FIG. 1, an inter-network relay device has one or more WAN or LAN line interfaces (Line # 1 to Line # 3), and each port is connected to a PMD (Physical Layer).
r Medium Dependent) and PHY
(Physical Layer) 11-1 to 11-3
It consists of each function of.

Further, data link controllers 12-1 to 12-1 corresponding to a data link layer for each line interface.
12-3 is provided, and a packet transmission / reception process with the line is performed. In FIG. 5, three line interfaces are provided for convenience.

When the packet forwarder 13 receives a packet from the data link controllers 12-1 to 12-3, the address of the packet, for example, IP (Intern)
etProtocol), a line interface to be transferred is determined, the packet is relayed as a packet addressed to the corresponding transmission line interface, and the packet buffer memory 1 is sent via the queuing manager 15.
Write to 4.

[0006] Conversely, the data link controller 12
In the case of transmission to -1 to 12-3, the packet read out from the packet buffer memory 14 via the queuing manager 15 is similarly subjected to the relay processing and transmitted to the data link controllers 12-1 to 12-3. Issue the request.

[0007] In the above-mentioned relay processing, a field of a packet header is checked or added / added as normally performed by an IP router.
Processing such as deletion / rewriting is included. However, the processing content may differ between when receiving from the line interface and when transmitting to the line interface.

[0008] The packet buffer memory 14 is a memory for temporarily holding packets received from the packet forwarder 13 via the queuing manager 15.

The queuing manager 15 has a function of managing a queue of packets to be relayed, and adjusts the order and timing of transmitting packet data stored in the packet buffer memory 14 to the packet forwarder 13 for each packet. I do.

The queuing manager 15 has a DMA controller function for the packet forwarder 13 to read / write packet data transmitted / received from / to the packet buffer memory 14, and the scheduler in the queuing manager 15 and this DMA controller It has an arbiter function for arbitrating access rights when accessing the packet buffer memory.

FIG. 6 is a diagram showing a packet processing system of a conventional inter-network relay device. Note that the same parts as those in FIG.

In FIG. 1, the portion of the line interface outside the data link controller is omitted, and each port of the line interface is explicitly divided into a transmission line and a reception line for convenience.

As shown in FIG. 1, a FIFO is provided for each input interface (Line # 1Rx to Line # 3Rx).
Buffers (FIFO-1 to FIFO-3) are provided. This FIFO buffer (FIFO-1 to FIFO-1)
-3) sequentially stores packet data input from the corresponding input interface.

The packet forwarder 13 analyzes the packet data stored in the FIFO buffers (FIFO-1 to FIFO-3) and converts the packet data to the output interface (Line # 1Tx to Line #) of the destination.
3Tx), queues (3-11 to 13-13,
3-21 to 3-23 and 3 to 31 to 33).

The above queues (3-11 to 13-13,
3-21 to 3-23, 3-31 to 33-33) are provided in the packet buffer memory 14.

For example, the packet forwarder 13 is inputted from an input interface (Line # 1Rx),
When the destination of the packet data stored in the FIFO-1 is the output interface (Line # 1Tx), the packet data is queued in the queue 3-11.

The input interface (Line #)
1Rx) and the destination of the packet data stored in FIFO-1 is the output interface (Line #).
If 2Tx), the packet data is queued in the queue 3-21.

Queues (3-11 to 13-13, 3-21 to 23-23, 3-3) provided for each output interface
Packets queued in 1-3-33) are extracted by the corresponding schedulers 4-1 through 4-3, and output to the packet buffer memory 14 via the output interfaces (Line # 1Tx-Line # 3T).
The queues are queued in queues 5-1 to 5-3 provided for each x).

FIG. 7 is a state transition diagram for explaining the operation of the schedulers 4-1 to 4-3.

Here, the input interface Line
A description will be given taking the scheduler 4-1 provided corresponding to # 1Rx as an example.

First, the scheduler 4-1 retrieves the packet data stored in the queue 3-1 corresponding to the input interface Line # 1Rx (S
1).

Then, all the first packet data stored in the queue 3-1 is taken out, that is, the queue 3-
When the first packet data of 1 has been completely extracted, the process proceeds to S2.

In S2, the scheduler 4-1 operates as follows:
The packet data stored in the queue 3-2 corresponding to the input interface Line # 2Rx is extracted (S1).

Then, similarly to S1, when the first packet data of the queue 3-2 is completely extracted,
The process proceeds to S3.

Similarly, in S3, the scheduler 4-
1 retrieves the packet data stored in the queue 3-13, and returns to the processing of S1 when the first packet data of the queue 3-13 is completely retrieved.

Although the scheduler 4-1 has been described here, the schedulers 4-2 and 4-3 perform the same processing for the corresponding queue.
If there is no packet data in the corresponding queue in each state, the processing in that state is skipped and the state transits to the next state.

[0027]

However, the above-mentioned conventional inter-network relay device has the following problems. That is, data transmitted between networks includes data having high importance and data having low importance, but in the conventional inter-network relay device, the importance (required communication quality) of these data is not considered. .

Therefore, when the network becomes congested, it becomes difficult to relay not only data of low importance (data of low required communication quality) but also data of high importance (data of high required communication quality). There was a problem that would.

As a variable-length packet switch which is one of the inter-network relay devices, IEEE802.
There is a 1P-compliant Ethernet (registered trademark) switch having a priority control function for transmitting a high-priority packet to a network based on the priority assigned to the packet and improving communication quality of specific traffic.

However, such a variable-length packet switch also guarantees the communication quality of data with high importance (data with high required communication quality) only by changing the order of sending packets based on the priority. Not something.

In an ATM Cell switch that exchanges fixed-length cells, which is one of the inter-network relay devices, a GFR (Guaranteed) that guarantees a minimum bandwidth is used.
(Frame Rate) QOS class is realized.

However, this ATM Cell switch relates to fixed-length cells, and cannot be applied to an inter-network relay device that relays variable-length packets.

The present invention has been made in view of the above circumstances, and provides an inter-network relay device that can guarantee a minimum bandwidth of highly important data in an inter-network relay device that relays variable-length packets. The purpose is to do.

[0034]

Therefore, first, in order to achieve the above object, a first invention of the present invention is provided for each input interface and outputs a predetermined output to an output interface corresponding to the input interface. And an assurance traffic input queue for storing assurance data requiring quality of the same, and a best effort traffic for storing data other than the assurance data provided for each input interface and output to an output interface corresponding to the input interface. An input queue, a best effort traffic scheduler for controlling a rate of extracting data other than the guaranteed data stored in the best effort traffic input queue, and the best effort traffic scheduler. A guaranteed traffic scheduler for controlling the rate of extracting data other than the guaranteed data extracted from the fort traffic input queue and the guaranteed data stored in the guaranteed traffic input queue; and a guarantee traffic scheduler provided for each output interface. And an output queue for storing data extracted by the guaranteed traffic scheduler.

According to the present invention, the guaranteed traffic scheduler and the best effort traffic scheduler can control the extraction rates of the queues stored in the guaranteed traffic input queue and the best effort traffic input queue, respectively. The guaranteed bandwidth is guaranteed for each interface, and the rate at which guaranteed data is extracted from each input interface can be set within the guaranteed bandwidth.

According to a second aspect of the present invention, in the first aspect, a queue is provided for each output interface and detects congestion of a corresponding output interface based on data stored in the output queue. The inter-network relay device further comprising a monitor, wherein the best effort traffic scheduler and the guaranteed traffic scheduler change the extraction rate when the corresponding queue monitor detects congestion.

According to the invention, when the congestion of the output interface is detected by the queue monitor, the guaranteed traffic scheduler and the best effort traffic scheduler enter the guaranteed traffic input queue and the best effort traffic input queue, respectively. Controls the rate of retrieval of stored queues. Thereby, when network congestion is detected, communication quality of guarantee data can be ensured.

Further, according to the third invention of the present invention, in the first invention or the second invention, the packet data input from each input interface is classified based on the output interface and the communication quality class. An inter-network relay device, further comprising a packet forwarder that outputs the packet data to the guaranteed traffic input queue and the best effort traffic input queue provided for each of the input interfaces based on the packet data.

Further, according to a fourth aspect of the present invention, in the second aspect, when the congestion of the corresponding output interface is detected by the queue monitor, the congestion is stored in the guaranteed traffic input queue of the guaranteed traffic scheduler. 3. The inter-network relay device according to claim 2, wherein the obtained guarantee data extraction rate is determined based on the following equation.

(Guaranteed bandwidth occupied by the traffic of the corresponding input interface) / (Maximum bandwidth that can be taken by the corresponding output interface) Further, the fifth invention of the present invention provides the second invention, wherein
The extraction rate of data other than the guaranteed data stored in the best effort traffic input queue of the guaranteed traffic scheduler when the congestion of the corresponding output interface is detected by the queue monitor is determined based on the following equation. An inter-network relay apparatus characterized in that:

{(Maximum bandwidth available for corresponding output interface) − (Guaranteed bandwidth available for corresponding output interface)} / (Maximum bandwidth available for corresponding output interface) The extraction rate of data other than the guaranteed data stored in the best effort traffic input queue of the best effort traffic scheduler when the congestion of the corresponding output interface is detected by the queue monitor is determined based on the following equation. An inter-network relay apparatus characterized in that:

{(Maximum bandwidth available for corresponding input interface) − (Guaranteed bandwidth for corresponding input interface)} / {(Maximum bandwidth available for input interface) − (Guaranteed bandwidth for input interface)} According to a seventh aspect of the present invention, there is provided a first storage means for storing guarantee data output to an output interface corresponding to an input interface, and storing data other than the guarantee data output to an output interface corresponding to the input interface. Second storage means for storing, a best effort traffic scheduler for individually controlling the extraction rate of data other than the guarantee data stored by the second storage means, and the best effort traffic scheduler. Said warranty A scheduler for guaranteed traffic for individually controlling the extraction rate of the data other than the data and the guaranteed data stored by the first storage means, and the data extracted by the guaranteed traffic scheduler corresponding to the input interface. Means for outputting to an output interface that performs the operation.

Further, an eighth invention of the present invention is an input queue for guaranteed traffic which is provided for each input interface and stores assurance data required for a predetermined quality to be output to an output interface corresponding to the input interface. A best effort traffic input queue provided for each input interface and storing data other than the guarantee data output to an output interface corresponding to the input interface; and the best effort traffic input queue stored in the best effort traffic input queue. A best effort traffic scheduler for individually controlling the extraction rate of data other than guaranteed data; and a best effort traffic scheduler for extracting data from the best effort traffic input queue by the best effort traffic scheduler. Said the non-assurance data of the data,
A guaranteed traffic scheduler for individually controlling the extraction rate of the guaranteed data stored in the guaranteed traffic input queue, and a data provided for each of the output interfaces and stored by the guaranteed traffic scheduler. An output queue, provided for each of the output interfaces, and a queue monitor for detecting congestion of a corresponding output interface based on data stored in the output queue, wherein the congestion of the output interface corresponding to the queue monitor is provided. Is detected, the corresponding scheduler for guaranteed traffic and the scheduler for best effort traffic are notified of congestion, and when the corresponding queue monitor is notified of congestion, the best effort The traffic scheduler changes the extraction rate of data other than the guaranteed data stored in the best-effort traffic input queue, and the guaranteed traffic scheduler changes the best-effort traffic scheduler from the best-effort traffic input queue. An inter-network relay method characterized by changing a rate of extracting data other than the extracted guarantee data and guarantee data stored in the guarantee traffic input queue.

Further, a ninth invention of the present invention is a security traffic input queue which is provided for each input interface and stores guarantee data required to be output to an output interface corresponding to the input interface and requiring a predetermined quality. A best effort traffic input queue provided for each of the input interfaces and storing data other than the guarantee data output to an output interface corresponding to the input interface; and the best effort traffic input queue extracted from the best effort traffic input queue. A guaranteed traffic scheduler for controlling the extraction rate of data other than the guaranteed data and the guaranteed data stored in the guaranteed traffic input queue, and a guaranteed traffic scheduler provided for each output interface; Is a network between the relay apparatus characterized by comprising an output queue for storing data retrieved by use scheduler, respectively.

According to this invention, the scheduler for guaranteed traffic can control the extraction rate of the queue stored in the input queue for guaranteed traffic, so that the guaranteed bandwidth can be guaranteed for each interface, and the guaranteed bandwidth can be controlled within the guaranteed bandwidth. It is possible to set the rate of taking out the guarantee data from each input interface.

In a tenth aspect of the present invention based on the first aspect, the best effort traffic scheduler extracts data other than the guarantee data stored in the best effort traffic input queue in packet units. The guaranteed traffic scheduler extracts the guaranteed data stored in the guaranteed traffic input queue in packet units.

According to the invention, the scheduler for best effort traffic and the scheduler for guaranteed traffic read data in packet units.
It is possible to prevent transfer data from being destroyed at the time of reading.

Further, an eleventh invention of the present invention is the ninth invention, wherein the guaranteed traffic scheduler extracts the guaranteed data stored in the guaranteed traffic input queue in packet units. It is an inter-network relay device.

According to such an invention, since the guaranteed traffic scheduler reads data in packet units, it is possible to prevent transfer data from being destroyed at the time of reading.

[0050]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an inter-network relay device according to an embodiment of the present invention will be described with reference to the drawings.

The internal block diagram of the inter-network relay device is the same as that shown in FIG. 5, and will not be described here. However, in the present embodiment, functions such as a queuing manager and a packet forwarder are different.

FIG. 1 is a diagram showing a packet processing system of an inter-network relay device according to an embodiment of the present invention. In the figure, the portion of the line interface outside the data link controller is omitted, and for convenience, each port of the line interface is explicitly shown as a transmission line and a reception line.

As shown in the figure, a FIFO is provided for each input interface (Line # 1Rx to Line # 3Rx).
Buffers 41-1 to 41-3 are provided. This F
The packet data input from the corresponding input interface is sequentially stored in the IFO buffers 41-1 to 41-3.

The packet forwarder 31 analyzes the packet data stored in the FIFO buffers 41-1 to 41-3 and provides the packet data for each output interface (Line # 1Tx to Line # 3Tx) of the destination. Queue (42G-11 to 42G-13, 42
B-11 to 42B-13, 42G-21 to 23, 42B
-21 to 42B-23, 42G-31 to 33, 42B-
Queue to 31-33).

The queues (42G-11 to 42G)
-13, 42B-11 to 42B-13, 42G-21 to
23, 42B-21 to 42B-23, 42G-31 to 3
3, 42B-31-33) are provided in the packet buffer memory 32.

The queues are queues for the bandwidth guarantee class (42G-11 to 42G-13, 42G-21 to 42G-21).
23, 42G-31-33) and queues for the best effort class (42B-11-13, 42B-21-2)
3, 42B-31-33).

For example, the packet forwarder 31 is input from an input interface (Line # 1Rx),
If the destination of the packet data stored in the FIFO 41-1 is the output interface (Line # 1 Tx) and is a packet for a band guarantee class, the packet data is queued in the queue 42G-11. I do.

The input interface (Line #)
1Rx) and the destination of the packet data stored in the FIFO 41-1 is the output interface (Lin).
e # 2Tx), and if the packet is for a best effort class, the packet data is queued in the queue 42B-21.

The schedulers 44-1 to 44-3 for the best effort class are provided for each output interface (Line # 1Tx to Line # 3Tx) and correspond to the corresponding queues (42B-11 to 42B-13, 4B).
2B-21 to 23B, and 42B-31 to 33) respectively.

The corresponding queue monitors 48-1 to 48-4
When the congestion of the output interface is notified from 8-3, the corresponding queue (42B-11 to 42B-13,
42B-21 to 23B and 42B-31 to 33).
The queuing manager 33 controls the queuing of packet data to each queue.

Scheduler 43-1 for Bandwidth Guarantee Class
To 43-3 are output interfaces (Lin).
e # 1Tx to Line # 3Tx), and corresponding queues (42G-11 to 13, 13G-21 to 23G,
42G-31 to 33-3) and FIFO memories 45-1 to 45-3 for storing packet data transmitted from schedulers 44-1 to 44-3 for the best effort class.
From the packet data.

The scheduler 4 for these bandwidth guarantee classes
The packet data extracted by 3-1 to 43-3 are queued in queues 46-1 to 46-3 provided for each output interface (Line # 1Tx to Line # 3Tx) in the packet buffer memory 32.

The FIFOs 47-1 to 47-3 are output interfaces (Line # 1Tx to Line # 3Tx).
The buffer is provided for each packet, buffers packet data transmitted from the corresponding queue 46-1 to 46-3, and sequentially outputs the corresponding output interface (Line # 1 Tx to Li #).
ne # 3Tx).

Next, the operation of the inter-network relay device according to the embodiment of the present invention will be described.

The input interface (Line # 1Rx)
Through a line # 3Rx) and subjected to the relay processing by the packet forwarder 31, the FIFOs 41-1 to 41-in the packet forwarder 31.
3 via the output interface (Line # 1 Tx ~
Lines # 3Tx) and communication quality classes (bandwidth guarantee class, best effort class) and corresponding queues (42G-11 to 42G-13, 42B-).
11 to 42B-13, 42G-21 to 23, 42B-2
1-42B-23, 42G-31-33, 42B-31
To 33).

In the scheduler (44-1 to 44-3) for the best effort class, the input interface (Line # 1Rx to Li # Li) on the packet buffer memory 32 is used.
ne # 3Rx) queue group (4
2B-11 to 42B-13, 42B-21 to 23, 42
B-31 to 33), the packet data is read out in what order and at what timing, and the FIFO memory 4 in the next stage is read out.
Scheduling processing of whether to write in 5-1 to 45-3 is performed.

The scheduler 4 for the band guarantee class
In 3-1 to 43-3, the input interfaces (Line # 1Rx to Line #) on the packet buffer memory 32
3Rx) queues (42G-11 to 42G-13, 4
2G-21-23, 42G-31-33) and scheduler for best effort class (44-1-44-3)
In the queue group composed of the next-stage FIF0 memories (45-1 to 45-3), packet data is read out in any order and at any timing, and is written to the queues 46-1 to 46-3 on the next-stage packet buffer memory 32. This scheduling process is performed.

The queuing manager 33 writes the transmission packet data on the packet buffer memory 32 by the bandwidth guarantee class schedulers 43-1 to 43-3.
Buffer queue for the corresponding output interface (4
6-1 to 46-3), the packet forwarder 3
1 for the packet data transmission request.

In addition, the queuing manager 33
Output interface queue (transmission queue) 46-1
Queue monitor 48- for monitoring the queue length for each.
1 to 48-3.

The queue monitors 48-1 to 48-3 are connected to the bandwidth guarantee class schedulers 43-1 to 43-3 by congestion signal lines 49-1a to 49-3a, respectively.
The queue monitors 48-1 to 48-3 and the best effort class schedulers 44-1 to 44-3 are connected by congestion signal lines 49-1b to 49-3b.

When the length of the transmission queues 46-1 to 46-3 exceeds a preset threshold value, it is determined that congestion has occurred on the output interface (Line # 1Tx to Line # 3Tx) side, and the congestion signal line (49-1a-4
9-3a, 49-1b to 49-3b), the congestion notification is made to the schedulers 43-1 to 43-3 for the bandwidth guarantee class and the schedulers 44-1 to 44-3 for the best effort class, respectively.

When the queue length falls below a preset threshold value (a value different from the threshold value used for judging the occurrence of congestion may be set in order to provide hysteresis), the notification is canceled. You.

Scheduler 4 for Best Effort Class
4-1 to 44-3 and scheduler 43 for bandwidth guarantee class
When receiving congestion notifications for the output interfaces (Line # 1Tx to Line # 3Tx) corresponding to itself, -1 to 43-3 change the scheduling algorithm. When the congestion notification is released, the process returns to the normal scheduling algorithm. The post-change scheduling will be described later with reference to FIGS.

Upon receiving the packet data from the input interfaces (Line # 1Rx to Line # 3Rx), the packet forwarder 31 receives the input interface number, the output interface number to which the packet is to be transferred, the communication quality class (bandwidth guarantee class, best effort Class), writes packet data into the packet buffer memory 32, and writes the corresponding queue (42G-11 to 42G-11).
42G-13, 42B-11 to 42B-13, 42G-
21 to 23, 42B-21 to 42B-23, 42G-3
1 to 33 and 42B-31 to 33), and notifies the queuing manager 33 of the reception.

When a packet transmission processing request is received from the queuing manager 33, packet data is read from the transmission queues 46-1 to 46-3 in the packet buffer memory 32, and the corresponding output interface (L
The packet is output to ine # 1Tx to Line # 3Tx).

FIG. 2 is a state transition diagram of the bandwidth guarantee class schedulers 43-1 to 43-3 in the queuing manager 33.

S10, S11, S21 to S2 (k +
1), S31 to S3 (k + 1) are symbols for identifying each state [k = 3 in the example of FIG. 1].

Normally, in the idle state S10, when a packet reception notification is received from the switch interface,
Transit to the status check state S11.

If the packet data exists in any of the queues in the queue group to be processed and the queue monitor indicates a congestion-free state, the packet data is read from the queue of the reception line input interface Line # 1Rx. Transition to S21.

Therefore, the packet data is sequentially read from the head in a unit of a certain number of bytes, written into the next FIFO memory, and when the reading to the end of the packet is completed, the packet data is read from the queue of the next input interface Line # 2Rx. The state transits to the reading state S22, and thereafter, the packet data is similarly read from the queue of the reception line corresponding to the current state, and written to the next-stage FIFO.

When the number of transitions has reached the number of all input interfaces, the state transits to a state S2 (k + 1) for reading packet data from the FIFO output from the best-effort class scheduler. The data is sequentially read out and written into the FIFO memory of the next stage, and when reading up to the end of the packet is completed, the state returns to S11.

At this point, if the packet data still exists in any of the queues in the queue group, the process returns to S21.
State transition. If there is no packet data in the corresponding queue in each state, the processing in that state is skipped and the state transits to the next state.

If the packet data exists in any of the queue groups to be processed and the congestion signal line indicates a congestion state, the input interface Li
The state transits to the state S31 in which packet data is read from the queue of ne # 1Rx.

In S31, the read pointer (indicating the address on the packet buffer memory) for sequentially reading the packet data from the head is moved in a unit, for example, a byte unit. If the pointer reaches the end of the packet data, all of the corresponding packet data is read out and written into the next FIFO memory.

When the pointer movement is repeated up to a preset number of times VAL_L1 or when the pointer reaches the end of the packet data and all the packet data is written to the FIFO memory of the next stage, the queue of the next input interface Line # 2Rx is Then, the state shifts to a state S32 where packet data is read from the CPU, and thereafter, a pointer operation for reading packet data from the corresponding queue and writing the packet data in the next-stage FIFO is similarly performed.

When the number of transitions has reached the number of all reception interfaces, the scheduler output FI for the best effort class is output.
State S3 (k + 1) of reading packet data from FO
The packet data read pointer (indicating the address in the FIFO memory) is sequentially moved in a certain unit as in the case of the reception interface queue, and when the end of the packet data is reached, the entire packet data is read and the next Write to the FIFO memory of the stage.

When the pointer operation is repeated up to a preset number of times VAL_BE, or when the pointer reaches the end of the packet data and the entire packet data is written into the next-stage FIFO memory, the state returns to the state of S11.

At this point, if packet data still exists in any of the queues in the queue group, the process returns to S31.
State transition. If there is no packet data in the corresponding queue in each state, the processing in that state is skipped and the state transits to the next state.

VAL_L1 to VAL_LK are determined based on the ratio of the guaranteed bandwidth occupied on the transfer destination line by the traffic (packet sequence) input from the corresponding input interface. VAL_BE is determined based on the difference between the maximum bandwidth that the output interface of the transfer destination can take and the total guaranteed bandwidth. VAL_L1 to VAL
_LK and VAL_BE satisfy the following relationship.

[0090]

(Equation 1)

[0091]

(Equation 2)

In the above equation, the “guaranteed bandwidth of the output interface” is equal to the sum of the guaranteed bandwidths of the traffic of the output interface for the destinations transferred at each input interface.

In the state S11, if there is no packet data in any of the queue groups to be processed, the state transits to the idle state S10.

FIG. 3 shows a scheduler 44-1 to 4-4 for the best effort class in the queuing manager 33.
It is a state transition diagram of 4-3. State S2 (k + 1) from the state that can be taken by the bandwidth guarantee class scheduler shown in FIG.
And state S3 (k + 1) deleted. Also, the number of times the queue is read in the output interface congestion state is different.

VAL_BE_L1 to VAL_BE_LK
Is determined based on the ratio of the guaranteed bandwidth occupied by the traffic (packet sequence) input from the corresponding input interface on the input interface. VAL_B
E_L1 to VAL_BE_LK satisfy the following relationship.

[0096]

(Equation 3)

Note that the schedulers (44-1 to 44-3) for the best effort class and the schedulers 43-1 to 43-3 for the band guarantee class respectively correspond to the corresponding queues (42G-11 to 42G-13, 42B). -11 to 11
42B-13, 42G-21 to 23, 42B-21 to 4
2B-23, 42G-31-33, 42B-31-3
From 3), the packet data is read by the algorithm described above, but if the read packet data is in the middle of the packet, the next-stage FIFO memory (45-1 to 45-3) or the queue (46-46) is read. 1-46
-3) does not output the read packet data, but completely reads the entire packet and then reads the next FIFO memory (45-1 to 45-3) or queue (46-1 to 46).
-3) Output the packet data.

Therefore, according to the inter-network relay device of the present embodiment, when the congestion of the output interface is detected, each scheduler corresponding to the output interface where the congestion is detected changes the packet extraction rate. Therefore, it is possible to guarantee the minimum bandwidth of data requiring communication quality for each input interface.

FIG. 4 shows the structure of the packet buffer queue on the packet buffer memory.

The queue is composed of j buffers in which packet data is stored and the head pointer (address on the memory)
And a queue descriptor indicating

The queue descriptor has j entries corresponding to each packet buffer, and each entry is composed of a field indicating the buffer head pointer and a field indicating the packet data size in the buffer.
These entries are used cyclically.

In addition, the queue descriptor has a field indicating the total data size which is the total of queued packet data. When the scheduler in the queuing manager and the DMA controller operate the queue, the descriptor access is performed between the two. It has a semaphore field to perform exclusion of

In the queuing manager, there are registers for storing the head pointer and the end pointer of the queue descriptor, and registers for storing the head pointers of the head entry and the end entry of the actual queue.

An entry existing between the positions indicated by the head pointers of the head entry and the tail entry of the queue corresponds to the queue. If the head pointer value of the head entry is equal to the head pointer value of the tail entry, the queuing manager determines that there is no packet data in the queue.

When the various schedulers read the packet data or when the DMA controller reads the packet data to output the packet data to the packet forwarder, the first entry in the queue descriptor is referred to by referring to the pointer storage register of the first entry. Then, the position of the head packet data in the queue is derived from the buffer head pointer in the entry.

After reading the packet data, the value of the pointer storage register of the first entry is replaced with the value of the pointer of the next entry.

When the DMA controller writes the packet data to input the packet data from the packet forwarder, the position of the last entry on the queue descriptor is determined by referring to the pointer storage register of the last entry. Since the next entry following the end of the queue indicates the position of the free buffer, the packet data is written to that buffer.

After the writing of the packet data is completed, the value of the pointer storage register of the last entry is replaced with the value of the pointer of the next entry.

Further, the packet data size is added to the total data size field on the queue descriptor.

Conversely, when the various schedulers read the packet data, or when the DMA controller reads the packet data to output the packet data to the packet forwarder, the packet data size is subtracted from the total data size field. .

If the value of the total data size field is not 0, packet data exists in the queue.

That is, in the present embodiment, when reading data from the input queue, each scheduler moves a pointer for reading packet data by a certain unit and keeps moving the pointer until the pointer reaches the end of the packet data. The read rate is controlled by changing the number of read units to be read at once without outputting from the queue.

In the present embodiment, a case has been described in which scheduling is performed by the scheduler for the bandwidth guarantee class and the scheduler for the best effort class. However, the change of the extraction rate by the scheduler for the best effort class is indispensable. Not a component.

That is, the best effort class queues (42B-11 to 42B-13, 42B-21 to 2)
3, 42B-31-33), the packet is stored in each queue (4) without changing the extraction rate by the scheduler for the best effort class.
2B-11 to 42B-13, 42B-21 to 23, 42
B-31 to 33), packet data may be taken out in order and stored in the corresponding FIFOs (45-1 to 45-3).

In this case, the change of the extraction rate is performed only by the scheduler for the band guarantee class.

Therefore, according to the inter-network relay device of the present embodiment, even when the output interface is congested, the bandwidth set for each input interface is guaranteed.
In addition, the remaining bandwidth of the output interface can be allocated at a rate of best effort traffic for each input interface.

As a result, it is possible to guarantee the minimum bandwidth for each input interface, and to maintain fairness by performing packet relay at a rate corresponding to the limit bandwidth of the input interface.

Also, in a network for transferring variable-length packets, by guaranteeing the minimum bandwidth between the input and output interfaces of the relay device, it is possible to predict the minimum throughput of communication passing through that section. .

The present invention is not limited to the above embodiments, and can be variously modified at the stage of implementation without departing from the scope of the invention. In addition, when an invention is extracted by omitting some constituent elements from all the constituent elements described in the embodiment, when practicing the extracted invention, the omitted part may be omitted by a well-known common technique. It is supplemented.

Further, the method described in the embodiment can be implemented by a computer (computer) as a program (software means) such as a magnetic disk (floppy (registered trademark) disk, hard disk, etc.)
It can be stored in a recording medium such as an optical disk (CD-ROM, DVD, MO, etc.), semiconductor memory (ROM, RAM, flash memory, etc.), and can also be transmitted and distributed via a communication medium. The program stored on the medium side includes a setting program for causing the computer to execute software means (including not only the execution program but also tables and data structures) to be executed in the computer. The recording medium referred to in the present specification is not limited to a medium for distribution, but includes a storage medium such as a magnetic disk or a semiconductor memory provided in a computer or a device connected via a network.

[0121]

As described above in detail, according to the present invention,
In an inter-network relay device that relays variable-length packets, it is possible to provide an inter-network relay device that can guarantee a minimum bandwidth for data of high importance.

[Brief description of the drawings]

FIG. 1 is a diagram showing a packet processing system of an inter-network relay device according to an embodiment of the present invention.

FIG. 2 is a state transition diagram of bandwidth guarantee class schedulers 43-1 to 43-3 in the queuing manager 33.

FIG. 3 is a state transition diagram of a best effort class scheduler 44-1 to 44-3 in the queuing manager 33.

FIG. 4 is a diagram showing a structure of a packet buffer queue on a packet buffer memory.

FIG. 5 is a general internal block diagram of a router as an inter-network relay device.

FIG. 6 is a diagram showing a packet processing system of a conventional inter-network relay device.

FIG. 7 is a state transition diagram for explaining operations of the schedulers 4-1 to 4-3.

[Explanation of symbols]

3-11 to 13-13, 3-21 to 23-23, 3-31 to 31
3-33: queue, 4-1 to 4-3: scheduler, 5-1 to 5-3: queue, 11-1 to 11-3: PMD / PHY, 12-1 to 12-3: data link controller, 13: Packet forwarder, 14: Packet buffer memory, 15: Queuing manager, 31: Packet forwarder, 32: Packet buffer memory, 33: Queuing manager, 41-1 to 41-3: FIFO, 42G-11 to 13, 42G-21 to 23, 42G-3
1 to 33: queues for bandwidth guarantee classes, 42B-11 to 13, 42B-21 to 23, 42B-3
1 to 33: queue for best effort class; 43-1 to 4: scheduler for bandwidth guarantee class; 44-1 to 4: scheduler for best effort class; 45-1 to 3: FIFO memory; 3 ... Queue, 47-1 to 3 ... FIFO, 48-1 to 3 ... Queue Monitor, 49-1a, 1b, 49-2a, 2b, 49-3a, 3
b: Congestion signal line.

Claims (11)

[Claims] An input queue for assurance traffic, which is provided for each input interface and stores assurance data required for a predetermined quality to be output to an output interface corresponding to the input interface, and is provided for each input interface. A best effort traffic input queue for storing data other than the guaranteed data output to the output interface corresponding to the input interface, and a retrieval rate of data other than the guaranteed data stored in the best effort traffic input queue. A scheduler for best effort traffic to be controlled, and a data other than the guarantee data extracted from the input queue for best effort traffic by the scheduler for best effort traffic. Data, a guaranteed traffic scheduler for controlling the extraction rate of the guaranteed data stored in the guaranteed traffic input queue, and a data extracted by the guaranteed traffic scheduler provided for each output interface. An inter-network relay device, comprising: an output queue for storing. 2. A scheduler for the best effort traffic and the guarantee, the queue monitor being provided for each output interface and detecting congestion of a corresponding output interface based on data stored in the output queue. 2. The inter-network relay device according to claim 1, wherein the traffic scheduler changes the extraction rate when the corresponding queue monitor detects congestion. 3. Classifying packet data input from each input interface based on an output interface and a communication quality class, and categorizing the packet data based on the classification based on the input queue provided for each of the input interfaces. 3. The inter-network relay device according to claim 1, further comprising a packet forwarder that outputs the packet to the best effort traffic input queue. 4. The extraction rate of guaranteed data stored in the guaranteed traffic input queue of the guaranteed traffic scheduler when the congestion of a corresponding output interface is detected by the queue monitor is determined based on the following equation. 3. The inter-network relay device according to claim 2, wherein: (Guaranteed bandwidth occupied by the corresponding input interface traffic) / (maximum available bandwidth of the corresponding output interface) 5. The extraction rate of data other than guaranteed data stored in the best effort traffic input queue of the guaranteed traffic scheduler when the congestion of a corresponding output interface is detected by the queue monitor is represented by the following formula: 3. The inter-network relay device according to claim 2, wherein the inter-network relay device is determined based on the following. {(Maximum possible bandwidth of corresponding output interface) − (Guaranteed bandwidth of corresponding output interface)}
/ (Maximum bandwidth that the corresponding output interface can take) 6. The extraction rate of data other than the guaranteed data stored in the best effort traffic input queue of the best effort traffic scheduler when the congestion of the corresponding output interface is detected by the queue monitor is as follows: 3. The inter-network relay device according to claim 2, wherein the value is determined based on an equation. {(Maximum available bandwidth of corresponding input interface) − (Guaranteed bandwidth of corresponding input interface)}
/ Σ ｛(Maximum bandwidth available for input interface)-
(Guaranteed bandwidth of input interface) 7. A first storage unit for storing guarantee data output to an output interface corresponding to an input interface, and a second storage unit for storing data other than the guarantee data output to an output interface corresponding to the input interface. 2, a best-effort traffic scheduler for individually controlling a retrieval rate of data other than the guarantee data stored by the second storage means, and the guarantee data retrieved by the best-effort traffic scheduler. And a guaranteed traffic scheduler for individually controlling the extraction rates of the guaranteed data stored by the first storage means and the guaranteed data stored by the first storage means. Computer-readable information recording medium characterized by comprising a means for outputting to an output interface corresponding to the interface. 8. An input queue for assurance traffic provided for each input interface and storing assurance data required for a predetermined quality to be output to an output interface corresponding to the input interface, and provided for each input interface. A best effort traffic input queue for storing data other than the guaranteed data output to the output interface corresponding to the input interface, and a retrieval rate of data other than the guaranteed data stored in the best effort traffic input queue. A scheduler for best effort traffic that is individually controlled; and data other than the guaranteed data extracted from the input queue for best effort traffic by the scheduler for best effort traffic. A guaranteed traffic scheduler for individually controlling the extraction rate of the guaranteed data stored in the guaranteed traffic input queue, and a guaranteed traffic scheduler provided for each of the output interfaces. An output queue to be stored, and a queue monitor provided for each of the output interfaces, for detecting congestion of a corresponding output interface based on data stored in the output queue, and an output corresponding to the queue monitor. When the congestion of the interface is detected, the congestion is notified to the corresponding guaranteed traffic scheduler and the best effort traffic scheduler. When the congestion is notified from the corresponding queue monitor, the best The fort traffic scheduler changes the extraction rate of data other than the guaranteed data stored in the best effort traffic input queue, and the guaranteed traffic scheduler changes the best effort traffic input queue by the best effort traffic scheduler. A method of changing the extraction rate of data other than the guarantee data extracted from the guarantee data and the guarantee data stored in the guarantee traffic input queue. 9. An assurance traffic input queue, provided for each input interface, for storing assurance data required for a predetermined quality to be output to an output interface corresponding to the input interface, and provided for each input interface. An input queue for best effort traffic storing data other than the guarantee data output to an output interface corresponding to the input interface, data other than the guarantee data extracted from the input queue for best effort traffic, and the guarantee A guaranteed traffic scheduler for controlling the extraction rate of the guaranteed data stored in the traffic input queue, and a guaranteed traffic scheduler provided for each of the output interfaces. An inter-network relay device, comprising: an output queue for storing output data. 10. The best-effort traffic scheduler extracts data other than the guaranteed data stored in the best-effort traffic input queue in packet units, and the guaranteed traffic scheduler sends the data to the guaranteed traffic input queue. 2. The inter-network relay device according to claim 1, wherein the stored guarantee data is extracted in packet units. 11. The inter-network relay device according to claim 9, wherein the guaranteed traffic scheduler extracts the guaranteed data stored in the guaranteed traffic input queue in packet units.