JPH0750673A - Congestion detection system and congestion control system - Google Patents

Abstract

PURPOSE:To effectively use a resource in a network and to collect congestion owing to the in-flow of burst data in an early stage. CONSTITUTION:An exchange 81 periodically monitors a buffer use ratio in a state management device 83, and uses a buffer use ratio difference (increase quantity) with a previous period. Thus, it is discriminated whether the input (in-flow) of regular data is the input (in-flow) of burst data or not. When it is judged to be the in-flow of burst data, the state transition of congestion is detected earlier than the case of the in-flow of regular data, and congestion is noticed to a calling station using a congestion route. The exchange 81 to which congestion is noticed checks up on the input traffic of the call using the congestion route by a traffic monitor device 82 and user I/F9X restricts congestion in accordance with a network state and input traffic. Thus, congestion owing to the in-flow of burst data can be detected in the early stage, and congestion can be controlled in accordance with the state in the network.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to congestion control in a network that handles burst data.

[0002]

2. Description of the Related Art As shown in "Packet Switching Technology and Its Applications" by Masaya Yamauchi (Corona Publishing Co., Ltd.), the conventional congestion detection method observes the number of buffers used per unit time and exceeds the reference value for a certain time or longer. In this case, congestion detection is used because the buffer usage rate is exceeded.

FIG. 8 is a block diagram showing a conventional policing control method for an ATM network disclosed in Japanese Patent Laid-Open No. 4-168834, which is 21X (X = 1, 2, 3, ...).
・) Is an ATM terminal, and the input monitoring device 23X (X = 1,
2, 3, ...) And connected to the call processing 250, and also 24
0 is an ATM switch. Next, the operation will be described. ATM based on the value declared by the user from the ATM terminal
The input monitoring device 23X (X = 1, 2 ...) Of the exchange monitors whether or not the user information to be transmitted is in violation of the declaration. Discard processing for leaky packet method, output processing to switch 240 after marking for marking method (The leaky packet method and marking method are written by Shigehiko Suzuki "Network Construction Technology for B-ISDN", 1990. The electronic information society spring national convention co-sponsored seminar, PP7-15) is conducted to notify the call processing device 250 of the report violation. When the call processing device 250 receives the notification of the report violation, the call processing device 250 confirms that the number of violations is within a predetermined number, and
A warning is issued to the terminal 21X (X = 1, 2 ...). ATM
When the terminal receives a warning, the declared value is corrected and the call processor 2
Send to 50. The call processing device 250 monitors the user information according to the modified declared value. If such a modification also causes a report violation, and even if the modification is repeated, if the violation occurs and the number of violations exceeds a predetermined number for the same call, the call processing device 250 forces the call in the connected state. It is a technology of shutting off.

[0004]

In the conventional congestion detection method, the congestion is detected by comparing the buffer usage rate with the congestion control buffer usage rate threshold value without distinguishing between normal data and burst data. However, in a network that handles burst data, there is a possibility that a large amount of congestion may occur in a short time due to the burst data, and the conventional congestion detection method is not suitable for control. In the congestion control, the conventional method regulates the call that outputs the user information more than the declared value regardless of the state in the network, so even if there is a margin in the network resource, There was a problem that frame discarding occurred and resources in the network could not be effectively used.

The present invention has been made to solve the above problems, and provides a congestion detection method for detecting congestion by distinguishing between normal data and burst data.
It is an object of the present invention to provide a congestion control method that makes effective use of resources in the network.

[0006]

According to a first aspect of the present invention, there is provided a congestion detection method, wherein in a network that handles burst data, a means for calculating a difference value (increase amount) from a buffer usage rate at the time of previous detection at the time of detection, A means for comparing the difference value and the reference value, and, when the difference value exceeds the reference value, it is determined that burst data inflow, the buffer utilization rate and burst data congestion detection buffer utilization rate threshold value The feature is that congestion is detected by comparison.

The congestion detection method according to a second aspect of the present invention is the congestion detection method according to the first aspect, characterized in that the burst data inflow is determined by a sum of difference values at least twice in the past.

According to another aspect of the congestion detection method of the present invention, at the time of detection, the sum of the difference value (increase amount) from the buffer usage rate at the previous detection and the buffer usage rate, and the normal data congestion detection buffer usage rate threshold value are used. It is characterized in that the congestion is judged by comparing

A congestion detection method according to a fourth aspect is the congestion detection method according to the third aspect, wherein a sum of a value obtained by multiplying a difference value by an arbitrary coefficient and a buffer utilization rate is a normal data congestion detection buffer utilization rate threshold value. It is characterized by comparing with and determining congestion.

According to a fifth aspect of the congestion control method, in congestion control when congestion is detected, means for notifying the originating station of the state in the network, means for monitoring the input traffic at the originating station, and the input Means for comparing the traffic and a traffic value (declared value) reported in advance by the originating station, and when the congestion in the network occurs, the input traffic of the call is examined by the originating station, and the declared value is The above-mentioned originating station preferentially restricts the above-mentioned call issuing traffic.

A congestion control method according to a sixth aspect is the congestion control method according to the fifth aspect, wherein during congestion, even if the input traffic at the originating station is within the declared value, a call that is above the reference value is being output. It is characterized in that the regulations are applied with priority.

According to the congestion control method of claim 7, when the congestion level is defined in two or more steps in the congestion control method of claim 5, input traffic of a declared value or more is applied to a call using the congestion level. For outgoing calls,
It is characterized in that at least one level of regulation is applied to calls that comply with the declared value.

[0013]

According to the congestion detection method of the first aspect, at the time of detection, the difference value (increase amount) from the buffer utilization rate at the time of the previous detection is calculated and compared with the reference value to distinguish normal data from burst data. , It is possible to prevent large congestion from occurring in a short time due to burst data.

According to the congestion detection method of the second aspect, at the time of detection, the difference value (increase amount) from the buffer usage rate at the time of the previous detection is calculated and compared with the reference value to distinguish the normal data and the burst data. Therefore, the buffer usage rate exceeding the burst data congestion detection buffer usage rate threshold value is not used as normal data.

According to the congestion detection method of the third aspect, congestion can be detected early when the buffer usage rate is high.

According to the congestion detection method of the fourth aspect, the congestion when the buffer usage rate is large can be detected earlier than the congestion detection method of the third aspect.

According to the congestion control method of the fifth aspect, since the input traffic is monitored for each call at the calling station, a call using a congestion route at the time of congestion is preferentially given to a call issuing traffic equal to or more than the declared value. The resources in the network can be effectively used, for example, the traffic can be regulated, and when the network has a margin, even traffic exceeding the declared value can be accepted.

According to the congestion detection method of the sixth aspect, even if the input traffic is within the declared value during congestion, the call issuing traffic exceeding the reference value is restricted.
Congestion control considering fair use of resources in the network is possible.

In the congestion control method of claim 7, since the congestion level is defined in two or more stages and the input traffic is monitored for each call at the originating station, at the time of congestion, the input traffic is checked for each call using the congestion route, For calls that generate more traffic than the declared value, at least one level more restrictive than calls that comply with the declared value. Therefore, if there is a fair use of resources in the network and there is room in the network, Since the traffic can be accepted, the resources in the network can be effectively used.

[0020]

【Example】

Example 1. Embodiment 1 of the present invention will be described below with reference to the drawings. In FIG. 1, reference numeral 81 is an exchange, and reference numeral 82 is a traffic monitor, which periodically monitors input traffic for each call. A state management device 83 monitors the buffer usage rate on the trunk side and detects congestion on the trunk side. 84 is a switch, and 11X (X = 0, 1, 2, ...) Is another exchange. 9X (X = 0, 1, 2, ...) Performs header processing by the user I / F and user frame processing such as user input traffic control. 10X (X = 0, 1, 2, ...) Is a terminal. In a network that handles burst data, the state management device 83 of FIG. 1 periodically monitors the buffer utilization rate on the trunk side in order to distinguish between normal data and burst data. The congestion detection method 1 of FIG. 2 is a graph of the buffer utilization rate on the trunk side, but the difference from the previous cycle is calculated for each cycle, and when the difference becomes larger than the reference value C ₁ , it is judged that burst data has flowed in. It is an example of a method of doing. That is, in FIG. 2, the difference 21 is smaller than the reference value C ₁ (21 <
C ₁ ), the normal data difference 22 is larger than the reference value C ₁ (22> C ₁ ), so that it becomes burst data. When it is determined that the burst data has flowed in, the state management device 83 of FIG. 1 determines the congestion by using the burst data congestion detection threshold value instead of the normal data congestion detection threshold value, so that the congestion can be detected earlier than the normal data. . That is, in FIG.
By using the burst data congestion detection threshold instead of the normal data congestion detection threshold when the burst data flows in, congestion can be detected at 23 instead of detecting congestion at 24, causing large congestion in a short time. Potential burst data congestion can be detected early.

Example 2. In the congestion detection method, when a plurality of congestion levels are defined, the state management device 83 of FIG. 1 periodically monitors the buffer utilization rate on the trunk side and uses the difference from the previous period to perform the same operation as in the above embodiment. The method determines the burst data inflow. At the time of inflow of burst data, congestion is judged using a burst data congestion detection threshold value which is smaller than the normal data congestion detection threshold value by the same width for each level, and congestion at each level is detected earlier than normal data.
FIG. 3 is a graph defining two congestion states, light congestion and heavy congestion, but the difference 31 is larger than the reference value C ₁ (31>
C ₁ ), it is judged that the burst data has flowed in, and the burst data light congestion detection threshold value is not the normal data light congestion detection threshold value, and the burst data heavy congestion detection threshold value is not the normal data heavy congestion detection threshold value. Is used to judge congestion (at this time, the difference 36 between the normal data light congestion detection threshold and the burst data light congestion detection threshold, the normal data heavy congestion detection threshold and the burst data heavy congestion detection threshold are detected. The difference from the threshold value 37 is the same). Therefore, in the example of FIG. 3, without distinction between normal data and burst data, light congestion detected at 33 can be detected at 32 and heavy congestion detected at 35 can be detected at 34, and burst data congestion can be detected early. Can be detected. Even if you define three or more congestion levels,
Similar to the method of FIG. 2, the difference between the normal data congestion detection threshold value and the burst data congestion detection threshold value can be set to detect the burst data congestion at an early stage.

By changing the sizes of 37 and 36 in FIG. 3 and increasing 37, it is possible to detect the heavy congestion earlier, and if heavy congestion causes frame discarding, etc.,
It is possible to prevent it by the congestion control function etc. by early detection. Further, at some levels, there is a method in which the difference between the normal data congestion detection threshold and the burst data congestion detection threshold is set to 0 and normal data and burst data are not distinguished. That is, in FIG. 2, assuming that the normal data light congestion threshold value is sufficiently small, it is not necessary to distinguish between the normal data light congestion threshold value and the burst data light congestion threshold value. It is also possible to set 0 to 0 and make the threshold value the same.

Example 3. In the congestion detection method, the last two times (3 times, 4 times ...
・ ・) There is a method to judge by using the sum of the difference values. The upper part of Fig. 4 shows the buffer utilization on the trunk side.
4, 46 and 47 are the differences from the previous cycle of each cycle. Figure 4
The above figure is a graph showing when the difference decreases for a certain period when the burst data flows in. In the method of judging whether it is normal data or burst data only by the difference from the previous cycle in the above-mentioned Embodiments 1 and 2, the differences 42, 46 and 47 are larger than the reference value, so it is judged as burst data and burst data congestion is detected. Since the congestion is detected using the threshold value, it is in the congestion state. However, since the difference 44 is smaller than the reference value,
Since congestion is detected using the normal data congestion detection threshold value, the congestion state disappears at point 45. The lower diagram of FIG. 4 shows a method in which a change in the congestion state due to a momentary decrease in the difference during burst data inflow is suppressed and early detection of burst data is possible. Figure 4
The lower diagram is a graph showing the sum of the differences of the past two times. For example, the sum of the differences 41 and 42 is shown in the second cycle, and the sum of the differences 42 and 44 is shown in the third cycle. The state management device 83 in FIG. 1 determines the inflow of burst data by comparing this sum of differences with the burst data detection threshold value.
That is, since the sum of differences (41 + 42) in the second cycle exceeds the burst data detection threshold, burst data,
Since the sum of differences (42 + 44) in the third period also exceeds the burst data detection threshold value, it becomes burst data. Therefore, even in the third period, it is determined as burst data, and the state management device 83 in FIG. 1 detects the congestion using the burst data congestion detection threshold value, so that the point 45 in FIG. 4 is also in the congestion state. In this way, in the method of discriminating between normal data and burst data by using the difference sum of the past two times (or three times, four times ...), the state of the next cycle can be predicted from the past state. It is possible to detect burst data congestion at an early stage.

Example 4. FIG. 5 shows the state management device 8 of FIG.
3 is a graph in which a difference is added to the buffer usage rate on the trunk side monitored by 3. The state management device 83 of FIG. 1 compares the value obtained by adding the difference to the buffer usage rate with the normal data congestion detection threshold value, and the value obtained by adding the difference to the buffer usage rate is shown as point 51 in FIG. When it becomes larger than the normal data congestion detection threshold value, the state management device 83 determines that it is congestion.
With this congestion detection method, when the buffer usage rate is small, congestion does not occur unless the difference is large, but when the buffer usage rate is large, congestion occurs even if the difference is small.
Therefore, with this congestion detection method, congestion can be detected early when the buffer usage rate is high.

In FIG. 5, there is also a method of multiplying the difference value by an arbitrary coefficient to obtain the sum of the buffer usage rate and comparing the sum with the normal data detection threshold value. With this method, it is possible to detect congestion earlier when the buffer usage rate is higher than the above method.

Example 5. Although FIG. 6 shows an overall image of congestion control, A terminal, B terminal, and C terminal have 64, 63, and 62 calls, respectively, and communicate with D terminal and E terminal. Such a route can occur when there is a failure in the 61 trunk line, but it is now assumed that the B terminal issues traffic exceeding the declared value and the data of A calls, B calls, and C calls are concentrated at the exchange A. To do. Then, in the exchange A, traffic exceeding the allowable amount may be concentrated and congestion may occur on the transmission side of the exchange A. At this time, the exchange A checks the calling station of the call using the route 67 and notifies the calling station of the congestion. In FIG. 6, the exchange A notifies the exchange B of the congestion. The call originating station (switch A and switch B) using the trunk line 67 examines the traffic of each call, and restricts the discarding of frames and the like for the call that has the input traffic exceeding the declared value at the originating station. Call. In addition, even during traffic during congestion, even if the traffic is within the declared value, a call such as a frame discard is applied to a call that outputs the input traffic above the reference value. This is explained in FIG. 7, and when the exchange A in FIG. 6 becomes congested, the traffic increase amount 71 in the period immediately before arrival of the congestion notification of the B call is larger than the declared value C ₂ , so the B call is issued at the same time as the congestion. Regulations such as frame discards will be applied at the station, and traffic will no longer increase. Also, for A calls, the input traffic is output in compliance with the declared value.
If the congestion is released when the B call is restricted, data can be output with the input traffic as it is, but if the congestion is not released, it is compared with the reference value C ₃ and 72> C.
If it is ₃ , at time 75 as shown in FIG. 6, restrictions such as frame discard are applied, and the traffic cannot be further increased. In the case of C call, if 73 <C ₃ , the current traffic can be maintained without restriction even during congestion as shown in FIG. 6b. Therefore, the calling station regulates the call using the congestion route according to the state in the network, so that the effective use of the resource in the network and the early termination of the congestion can be achieved.

An example of congestion control processing in the exchange will be described with reference to the block diagram of FIG. 1. The state management device 83 monitors the buffer utilization rate on the trunk side and congestion notifications from other exchanges. When the congestion notification (congestion notification to the calling station of the call using the congestion route) is received from another exchange, the state management device 8
3 examines the call using the route and examines the input traffic of the call from the traffic monitoring device 82. When the input traffic violates the declared value, the state management device 83 gives an instruction to the user I / F 9X, and the user I / F which has received the instruction puts restrictions such as frame discard.

Example 6. In the above congestion control, when the congestion level is defined in two or more stages, for example, two types are defined, that is, a level for alerting the user to the congestion and a level for discarding the frame. It is assumed that the B call and C call comply with the declared value.
When the congestion of the degree of user warning occurs in the exchange A, the exchange A notifies the originating station of the call using the trunk line 67 of the congestion. Each calling station examines the input traffic, and for B calls and C calls, the congestion level that is notified as observing the declared value, in this case, a warning is given to the user. However, since the input traffic exceeding the declared value is being output for the A call, the frame discarding one step higher is regulated. This makes it possible to regulate calls that are more likely to cause congestion than other calls.

[0029]

According to the congestion detection method of the first aspect of the present invention, in a network that handles burst data, means for calculating a difference value (increase amount) from the buffer usage rate at the time of the previous detection at the time of detection, and the difference value and the reference value. Means for comparing the values, and if the difference value exceeds the reference value, it is determined that the burst data inflow, the buffer utilization rate and burst data congestion detection buffer utilization rate threshold value is compared. Since it is detected, the difference value (increase amount) from the buffer usage rate at the time of the previous detection is calculated at the time of detection, and by comparing with the reference value, normal data and burst data are discriminated. Prevent congestion.

The congestion detection method according to claim 2 is the congestion detection method according to claim 1, wherein the judgment of the burst data inflow is made by at least the sum of the difference values of the past two times.
At the time of detection, the difference value (increase amount) from the buffer usage rate at the time of the previous detection is calculated, and normal data and burst data are discriminated by comparing with the reference value, so the burst data congestion detection buffer usage rate threshold The buffer usage rate that exceeds the value is not used as normal data.

According to the congestion detection method of the third aspect, congestion can be detected at an early stage when the buffer usage rate is high.

A congestion detection method according to a fourth aspect is the congestion detection method according to the third aspect, wherein a sum of a value obtained by multiplying a difference value by an arbitrary coefficient and a buffer usage rate is a normal data congestion detection buffer usage rate threshold value. Since the congestion is judged by comparing with the above, the congestion when the buffer usage rate is large can be detected earlier than the congestion detection method of the third aspect.

In the congestion control method of claim 5, in congestion control when congestion is detected, means for notifying the originating station of the state in the network, means for monitoring the input traffic at the originating station, and the input Means for comparing the traffic and a traffic value (declared value) reported in advance by the originating station, and when the congestion in the network occurs, the input traffic of the call is examined by the originating station, and the declared value is Since the call originating from the above traffic is preferentially restricted by the calling station, the resources in the network can be effectively used.

A congestion control system according to a sixth aspect is the congestion control system according to the fifth aspect, wherein during congestion, even if the input traffic at the originating station is within the declared value, a call that is above the reference value is being output. For example, even if the input traffic is within the declared value during congestion, it restricts the calls that have traffic exceeding the reference value. Congestion control that considers usage is possible.

According to the congestion control method of claim 7, in the congestion control method of claim 5, when the congestion level is defined in two or more stages, input traffic of a declared value or more is applied to a call using the congestion level. For outgoing calls,
Since at least one level of restrictions are imposed on calls that comply with the declared value, resources within the network can be effectively used.

[Brief description of drawings]

FIG. 1 is a block diagram of an inside of an exchange showing congestion control according to a first embodiment of the present invention.

FIG. 2 is a buffer utilization rate graph showing a congestion detection method according to the first embodiment of the present invention.

FIG. 3 is a buffer utilization rate graph showing a congestion detection method according to a second embodiment of the present invention.

FIG. 4 is a buffer utilization graph showing a congestion detection method according to a third embodiment of the present invention.

FIG. 5 is a buffer utilization graph showing a congestion detection method according to a fourth embodiment of the present invention.

FIG. 6 is an overall view of congestion control according to a fifth embodiment of the present invention.

FIG. 7 is an input traffic graph diagram showing congestion control according to the fifth embodiment of the present invention.

FIG. 8 is a block diagram in a switching system showing conventional congestion control.

[Explanation of symbols]

21, 22 Buffer usage rate difference value from previous cycle 23, 24 Congestion detection point 31 Buffer usage rate difference value from previous cycle 32 to 35 Congestion detection point 36, 37 Normal data congestion detection threshold value and burst data detection threshold Difference with value 41-42 Buffer usage rate difference value with previous cycle 43 Congestion detection point 44 Buffer usage rate difference value with previous cycle 45 Congestion detection point 46-47 Buffer usage rate difference value with previous cycle 51 Congestion detection point 61 Trunk line 62 C call route 63 B call route 64 A call route 65 B call regulation 66 A call regulation 67 Congestion trunk line 71-73 Input traffic for each monitoring period 74, 75 Regulation start point 81 Switch 82 Traffic monitoring device 83 State Management device 84 Switch 85 Other switch 9X User I / F 10X Terminal 200 ATM switch 21X A M terminal 23X input monitoring device 240 switches 250 call processor

Claims (7)

[Claims] 1. A network for handling burst data, comprising means for calculating a difference value (increase amount) from a buffer usage rate at the time of previous detection at the time of detection, and means for comparing the difference value with a reference value. Then, if the difference value exceeds the reference value, it is determined that burst data has flowed in, and congestion detection is performed by comparing the buffer usage rate and burst data congestion detection buffer usage rate threshold value to detect congestion. method. 2. The congestion detection method according to claim 1, wherein the judgment of the burst data inflow is made based on the sum of difference values at least twice in the past. 3. At the time of detection, the sum of the difference value (increase amount) from the buffer usage rate at the time of the previous detection and the buffer usage rate,
A congestion detection method characterized by determining congestion by comparing with a threshold value for the normal data congestion detection buffer usage rate. 4. The congestion is determined by comparing the sum of the difference value multiplied by an arbitrary coefficient and the buffer utilization rate with a normal data congestion detection buffer utilization rate threshold value. Congestion detection method. 5. In congestion control when congestion is detected, means for notifying the originating station of the state in the network, means for monitoring the input traffic at the originating station, and the input traffic and the originating station Means for comparing the reported traffic value (declared value), and when congestion in the network occurs, the call input traffic is examined by the originating station, and traffic exceeding the declared value is output. A congestion control method in which a call is preferentially restricted by the calling station. 6. During congestion, even if the input traffic at the originating station is within the declared value, calls with traffic exceeding the reference value are preferentially restricted. The congestion control method according to Item 5. 7. When the congestion level is defined in two or more levels, a call using the congestion level has a higher input traffic than a declared value than a call that keeps the declared value. 6. The congestion control method according to claim 5, wherein the restriction is applied at least one step higher.