JPH02257737A - Congestion control system in packet exchange network - Google Patents

Abstract

PURPOSE:To shorten the congestion time by allowing an exchange having congestion to inform the fact of occurrence of congestion to each adjacent exchange and allowing each exchange adjacent receiving the said notice to limit the transmission of a packet to the exchange in the state of congestion. CONSTITUTION:Each exchange in a packet exchange network is discriminated as to whether it is in congestion state and when the congestion state is discriminated, a congestion exchange 1 is informed the fact of the congestion state to adjacent exchanges 2-5 of, e.g. the exchange 1. The exchange receiving the notice applies output limit of a packet stored in its own exchange and to be outputted to the exchange 1 in congestion state. Thus, the congestion consecutive time is shortened.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an axle wheel control method for avoiding and resolving congestion that occurs in a packet-switched network that performs packet-switched communications.

(Prior art) Conventionally, when transmitting data digitally, this digital data is divided into pieces of predetermined length, and a header having an address etc. is added to each divided message to form a packet. A packet communication method has been proposed in which data is transmitted in units of .

A packet switching network is a communication network for performing such packet communications, and is composed of a plurality of exchanges and communication paths that alternately connect them.

In such a packet-switched network, the communication network becomes congested when calls far exceed the traffic capacity or when equipment malfunctions.

Conventionally, control methods for the axles that occur in packet switching networks include a priority control method in which packets stored in a switch or terminal buffer are sent out in order of priority, or an end-to-end flow using a response method. A so-called window control method that embodies the control method is known.

(Problems to be Solved by the Invention) However, such a conventional axis-lateral control system has the following problems.

In other words, in a packet-switched network that performs high-speed data communication, the packet streams generated from each terminal are multiplexed and exchanged at high speed. Instantaneous width portions that overlap with the period of manual input occur stochastically, but with conventional control methods, in such cases, a large number of packets are discarded regardless of their importance.

The present invention has been made to solve these problems in the prior art, and provides an axis-lateral control method that further shortens the width part duration and discards packets with a predetermined priority. The purpose is to provide the following.

[Structure of the invention]

(Means for Solving the Problems) In the axis-lateral control method in the packet switching network of the present invention, which has been made to achieve the above object, ■ each switch in the packet switching network is controlled in a congested state. If it is determined that the switch is in a congested state, it notifies neighboring switches of the congested switch that the congested switch is in the congested state, and the notified switch automatically Output restrictions are applied to packets stored in the switch that should be output to the switch in the congestion state.

Or, ■ Determine whether or not each switch in the packet switching network is in a congested state, and if it is determined to be congested, secure a route that allows packets to be transferred to that switch. A control packet is transferred to the terminal or terminal interface in order to limit the output of the packet, and the terminal or terminal interface that receives the control packet limits the output of the packet.

Furthermore, or, ■ It is determined whether each switch in the packet switching network is in a congested state, and if it is determined that the switch is in a congested state, the neighboring switch of the switch in that congested state is notified that it is in that congested state. The exchange notifies that it is in a congested state, and the exchange that receives the notification limits the output of packets to terminals or terminal interfaces that have secured a route that allows packet transfer to the congested exchange. The terminal or terminal interface that receives the control packet restricts the output of the packet.

(Function) In the width control method that takes the above-mentioned method (2), the exchange that caused the width notifies each adjacent exchange of the fact.

Then, each adjacent exchange that receives this notification restricts the sending of packets to the axis-side exchange for a predetermined period of time.

That is, for example, by stopping sending packets to the axis-side switch for a while, or by using a predetermined discard priority,
Packets are discarded from among the packets of each adjacent exchange to be sent to the axis-side exchange, and only those that are not discarded are sent to the axis-side exchange.

In the axis-lateral control method that takes the above-mentioned method (2), control packets are sent from the exchange that generated the width portion to each terminal or terminal interface directly or via an adjacent exchange. Each terminal or terminal interface that receives the control packet restricts packet transmission to the axle exchanger.

In the axis-lateral control method that takes the above-mentioned method (2), the exchange that has caused the congestion notifies each adjacent exchange of the fact. Then, instead of the shaft exchange, the adjacent exchange sends control packets to each terminal belonging to the adjacent exchange. Then, each terminal that receives the control packet restricts the sending of the packet to the axis-side switch.

(Example) Examples of the present invention will be described below with reference to the drawings.

FIG. 1 and FIG. 2 are a conceptual diagram showing an axis-lateral control method in a packet switching network according to a first embodiment of the present invention, and a flowchart showing a processing procedure, respectively.

In the case shown in FIG. 1, the exchange 1 is an exchange with a width section, and is shown with diagonal lines to distinguish it from other exchanges. Elements 2 to 5 are the exchanges adjacent to exchange 1 which produced the width.

A chain line extending from the axle exchanger 1 to each adjacent exchanger indicates the occurrence of a width portion in the exchanger 1.

Next, based on FIG. 2, the processing procedure for canceling the axis transverse direction performed in this embodiment will be described. Let us consider the situation shown in FIG. 1.

In step (2), it is determined whether a width portion has occurred for each exchange.

In this embodiment, each exchange itself monitors the usage rate of its buffer, and when it is determined that the axis-horizontal state has been reached, the corresponding signal is set as an interrupt signal to the CPU of that exchange. Although this is done manually, it is also possible to use a monitoring device provided independently of each exchange to detect the usage rate of the buffer within each exchange at a predetermined period. If it is determined that a width portion has occurred in a certain exchange, for example, exchange 1 as shown in FIG. In such a case, an axis-lateral notification packet is sent to each of the exchanges 2 or 2 indicating that a width portion has occurred in that exchange.

After sending the axis-side notification packet to a certain adjacent exchange, for example, the exchange 2, the notification table in the exchange 1 is updated in step (2), that is, data indicating that the axis-side notification packet has been sent to the exchange 2 is set.

Thereafter, it is determined in step (3) whether the congestion of the transverse exchange has been resolved. In this judgment, if it is determined that the width part has been resolved, step ■
, and the axis lateral control is released. That is, the exchange 1 sends a control packet to each of the adjacent exchanges 2 to 5 to notify the cancellation of the axle wheel control. Each exchange that receives the axle wheel control release packet shifts to normal operation mode. Thereafter, in step (2), the contents of the notification table in the exchange 1 are reset, and the process returns to step (2).

Furthermore, if it is determined in step (2) that the congestion is still not resolved, it is determined in step (2) whether or not the axle notification packets have been sent to all adjacent exchanges, and even if the determination result is YES, Then, the process moves to step (2), where it is determined whether the width portion has been resolved. If it is determined that the problem has been resolved, the process moves to step (3), and the end processing of the axis-lateral control is performed as described above.

On the other hand, if it is determined in step (■) that there is an adjacent exchange that has not sent the axis-side notification packet,
The process moves to step (3), and the same process as described above is performed for the adjacent exchange.

The adjacent exchange that has received the shaft notification packet thereafter limits the output of packets to the width exchange until the congestion is resolved.

As a method for limiting the output, for example, packet transmission to the width switch may be completely stopped, or if this causes the buffer in the adjacent switch to overflow, the packets may be discarded. At this time, the present invention discards packets using a predetermined discard priority, instead of uniformly discarding all packets that are manually input after the buffer is full, as in the conventional method. .

That is, for example, each packet has discard priority data in its header depending on the type of its unique data.

Each adjacent exchange detects the one with the highest discard priority among newly input packets and packets stored in the buffer, and discards it. As a result, 1. It is possible to avoid discarding things that should not be discarded. Furthermore, when discarding packets in adjacent exchanges, the route information possessed by each packet is also taken into account. In other words, route information is used to determine whether each packet should be forwarded to a wide switch, and when it is, it is discarded with priority over packets that are not. .

With this method, a certain percentage of the packets in the switch's buffer and newly input packets will be discarded, but the number of packets discarded with this control method will be The operating speed is extremely fast compared to the operating speed of the terminal, so if you evaluate each terminal, the number will be small.

According to the method of this embodiment, when congestion occurs in a certain exchange, the above-mentioned restrictions are placed on the subsequent transmission of packets from adjacent exchanges to that exchange, so that the congestion that has occurred can be quickly resolved. In this case, not only is
The exchange that has caused the congestion can simply notify the adjacent exchanges of the fact that a width section has occurred within itself as a measure to resolve the axle, and the exchange that has created the width section will only have to make the operation more difficult. It becomes possible to reduce the amount of further measures that must be taken to eliminate the axle for the congested exchanger in the current situation.

It should be noted that the gist of this embodiment can be carried out by other methods, and for example, it is also possible to configure the processing procedure shown in FIG. 2' so that the determination in step (2) is not performed. At this time, a determination as to whether the width portion has been resolved is made in step (3) after sending the axle notification packet to all adjacent exchanges.

Next, a second embodiment of the present invention will be described.

FIG. 3 and FIG. 4 are a conceptual diagram showing an axle wheel control method in a packet switching network according to the second embodiment, and a flowchart showing a processing procedure, respectively.

What is shown in FIG. 3 corresponds to the axle situation used as an example in the description of the first embodiment above, and elements having the same reference numerals as in FIG. 1 are as previously described. have the same meaning. For example, element 1 shown with diagonal lines
is the exchanger that caused the width part. Also, elements 6 to 11
is a terminal whose output packet passes through the shaft exchanger 1, and 12 is an exchanger that is not adjacent to the width exchanger 1. Also, from the shaft wheel exchanger 1 to each terminal 2 to 7
The wavy line toward represents the flow of control packets described below.

Elimination of the width portion in the exchange 1 is performed as follows. In other words, in step 1 shown in Fig. 4, it is determined whether a width portion is generated for each exchange! P11 is determined.

If it is determined that a width portion has occurred in a certain exchange, for example, exchange 1 as shown in FIG.
In step (2), control packets are sequentially sent to the terminals that are sending the packets via the exchange, that is, terminals 6 to 11 in the example shown in FIG.

After sending a control packet to a certain terminal, for example, terminal 6, in step 2, the control packet sending table in the exchange 1 is updated, that is, data indicating that the control packet has been sent to terminal 6 is set.

Thereafter, it is determined in step (3) whether or not the width of the axle wheel exchanger has been resolved. In this determination, if it is determined that the shaft lateral problem has been resolved, the process moves to step (3), and the shaft wheel control is canceled. That is, the exchange 1 sends a control packet to each of the terminals 6 to 11 to notify the cancellation of the axis-lateral control. Each exchange that receives the width section control release packet shifts to normal operation mode. Thereafter, in step (2), the contents of the control bucket I and the sending table in the exchange 1 are reset, and the process returns to step (2).

In step (2), it is determined whether the control packet has been sent to all target terminals, that is, terminals 6 to 11 in the case shown in FIG.

In the determination in step ■, if it is determined that the control packet has not been sent to all target terminals, step Shift, and repeat steps ① to ② below. At step ■, Y
If it is determined to be E or S, the process moves to step ■, and if it is determined that the horizontal axis has been resolved, then step ■
, move on to step ■ and step ■.

Each terminal that receives the control packet recognizes the control packet and limits the packet output.

As in the case of the first embodiment, various methods can be used to limit the output. For example, the output of packets may be completely stopped, or the output may be limited based on a predetermined discard priority. Some packets may be output by discarding the packets.

Various modifications can be made to this embodiment.

For example, the control packet may include information indicating the extent of the horizontal axis. As a result, if the mode of output restriction at each terminal is changed depending on the degree of horizontal axis, more rational width control can be performed. The mode of output restriction may be changed autonomously on the terminal side.

Further, various methods can be used for sending the control packet. For example, the control unit of a switch can store all the routing headers used to route calls (packet streams) accommodated by the switch to the originating and terminating stations, and this information can be used to create control packets. Transfer to a terminal or terminal interface. That is, the control unit of the exchange attaches a routing header to a control packet using the stored routing header information of each packet stream, and transfers the control packet to the terminal or terminal interface to be transferred.

Alternatively, the routing header of the packet input into the input buffer of the exchange may be added to the header of the control packet as is and transferred to the terminal side using the return packet method.

In other words, a packet switching network of transmission paths is created in which the same channel number is set for the same packet flow (call) for the line from one exchange A to the other exchange B, and the line from B to A in the opposite direction. The uplink channel number and the downlink channel number are set to the same value for one call. In such a packet switching network, the header part of a packet received by the input buffer section of the exchange in the axis-horizontal state is copied as is and sent back to the terminal side as a routing header of a control packet. This makes it possible to send control packets extremely easily, which in turn shortens the axis-transverse time.

At this time, you may choose not to transfer duplicate control packets to the same terminal or terminal interface, or you may transfer control packets unconditionally when the axis is in the horizontal state without being aware of duplication. However, in the latter case, the hardware that executes the control is simpler. In addition, the terminal interface device has a function to store the distance from itself to each exchange, such as the number of hops between the exchanges, and furthermore, the terminal interface device has a function to remember the distance from the own device to each exchange, such as the number of hops between the exchanges, and furthermore, the terminal interface device can recognize the switch in the horizontal axis state in the control packet. Based on this information, the terminal interface device may check the distance between itself and the exchange in the transverse axis state, and if the distance is greater than a certain value, the output may not be restricted.

Furthermore, even if the terminal or terminal interface device receives the control packet, it may not perform the control during the axle, for example, in a case where the packet sending rate is extremely high.

By the method of this embodiment, congestion can be resolved quickly, and even if packets are discarded at the terminal or terminal interface for congestion control, it is done using an appropriate discard priority, so it is not like the conventional technology. This solves the problem of uniformly discarding packets regardless of the type of data.

Next, a third embodiment of the present invention will be described. FIG. 5 is a conceptual diagram showing the axle wheel control method in the packet switching network of the third embodiment.

If it is determined that exchanger 1 is in the axis-lateral state, exchanger 1
sends a control packet to each adjacent exchange 2 to 5.

At the same time, control packets are sent to each terminal 9, 10 that directly sends packets to the exchange 1 without going through other exchanges. The flows of packets sent to adjacent exchanges and control packets sent directly to terminals are shown in FIG. 5 by dashed lines and dashed lines, respectively.

an adjacent exchange that inputs the control packet from the axis-horizontal exchange 1;
For example, 2 further sends an a1q control packet to each terminal subordinate to itself, for example, terminal 6 in the case of exchange 2.

The timing for determining whether to release the axle is determined every time a control packet is sent out, as in the second embodiment.

The packet output restriction performed by each terminal that receives the control packet is as already described in the second embodiment.

In this way, in this embodiment, the exchange that caused the axis horizontal transmission indirectly sends control packets to each terminal that is not directly connected via the adjacent exchange. By doing so, compared to the case of the second embodiment, it is possible to reduce the burden on the axis-lateral control of the exchange 1, which is busy with the original packet communication processing due to the generation of an axis. That is, in the example shown in FIG. 5, for example, a control packet is sent to the terminal 6 from the adjacent exchange 2 of the horizontal exchange 1, but the burden of horizontal control on the horizontal exchange 1 is reduced accordingly. .

Next, a fourth embodiment of the present invention will be described. FIG. 6 shows a conceptual diagram of the axis-lateral control method of this embodiment.

In FIG. 6, the chain lines extending from the exchange 1 that caused the horizontal axis to each exchange represent the flow of packets for notifying the horizontal axis in the exchange 1, as in the case of the first embodiment. This embodiment differs from the first embodiment in that notification of the shaft horizontal state is sent not only to adjacent exchanges but also to other exchanges.
Other than that, there is basically no difference from the first embodiment.

That is, for example, when the exchange 12 receives the axis-side notification from the exchange 1, it further sends a control packet to the terminal 11 that it manages. The terminal 11 that has input the control packet restricts the sending of packets via the exchange 1.

However, in this embodiment, since the shaft-lateral status is notified to a wider range of exchanges, in the case of the first embodiment, there is a risk that problems such as the shaft-horizontal waves spreading to adjacent exchanges may occur. This problem is solved.

At this time, if the distance between the switch that received the notification and the switch next to the axis where the problem occurred is large, such as the number of hops between the switches, control packets should not be transferred to the target terminal or terminal interface. It can also be done.

In other words, in order to eliminate the horizontal problem in exchange 1, it is not necessary to limit the output of all terminals whose output packets pass through the switch 1, and to eliminate the horizontal problem at an early stage, it is necessary to This is because it is more effective to notify a nearby exchanger than to perform axle wheel control by notifying an exchanger that takes time to notify.

Various embodiments other than those described above are possible.

For example, the axis-lateral control may be canceled autonomously by a terminal that has received the control packet. In this case for example the control bucket!・The control may be canceled after a certain period of time after receiving the message, or the terminal may send a data packet,
On the other hand, if a control packet for shaft lateral control is not returned, it may be assumed that the shaft lateral control has been resolved, and the shaft wheel control at that terminal may be canceled.

In an embodiment in which terminals or terminal interfaces are controlled, when a packet switching network is configured such that multiple routes connecting a common exchange are managed as one route group, Alternatively, the exchange may send a control notification to a terminal or a terminal interface that uses one of the routes in the group of routes it manages.

〔Effect of the invention〕

As is clear from the above, according to the present invention, the horizontal axis time is significantly shortened, and even when packets are unavoidably discarded, this is done based on the discard priority that takes into account the type of data, etc. The effect is that packets that would otherwise be discarded are not discarded.

[Brief explanation of drawings]

FIG. 1 and FIG. 2 are conceptual diagrams of the axle wheel control system in the packet switching network of the first embodiment of the present invention, respectively;
3 and 4 are conceptual diagrams of the axle wheel control method in the packet switching network according to the second embodiment of the present invention, and flowcharts, and FIGS. FIG. 12 is a conceptual diagram of an axle wheel control method in a packet switching network according to a fourth embodiment. 1.2, 3, 4, 5.12...exchange 6.7, 8, 9
, 10.11...Terminal representative Hidekazu Toshi Miyoshi Figure 1 Figure 2 Figure 5 Figure 6

Claims (3)

[Claims] (1) A congestion control method that performs control to eliminate a congestion state when a congestion state occurs in a packet switching network, which determines whether each switch in the packet switching network is in a congestion state, If it is determined that the switch is in a congested state, the neighboring switches of the switch in the congested state are notified that the switch in the congested state is in the congested state, and the notified switch is stored in its own switch. A congestion control method in a packet switching network characterized by restricting the output of packets that should be output to an exchange in a congested state. (2) A congestion control method that performs control to eliminate the congestion state when a congestion state occurs in the packet switching network, which determines whether each switch in the packet switching network is in a congestion state, If it is determined that the switch is in a congested state, a control packet is transferred to the terminal or terminal interface that has secured a route that allows the transfer of packets to that exchange to limit the output of the packet, and the control packet is A congestion control method in a packet-switched network, in which the terminal or terminal interface that receives the packet limits the output of the packet. (3) A congestion control method that performs control to eliminate a congestion state when a state of congestion occurs in a packet switching network, which determines whether each switch in the packet switching network is in a state of congestion; If it is determined that the switch is in a congested state, the neighboring switches of the switch in the congested state are notified that the switch in the congested state is in the congested state, and the notified switch sends a message to the switch in the congested state. Transfer a control packet to limit the output of packets to a terminal or terminal interface that has secured a route that allows packet transfer, and the terminal or terminal interface that receives the control packet limits the output of the packet. A congestion control method for packet-switched networks characterized by