JP2004032157A - Packet transmitting method, traffic conditioner, and priority control mechanism - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a packet transmission method which avoids deterioration at the same time. <P>SOLUTION: If the sum of the bands of a plurality of flows composed of packets with the same priority level exceeds an assured band shared by these flows, higher priority levels are given to packets belonging to flows having older communication start times than those belonging to newer communication start times in the order of arrival with respect to the communication start time of the flow. A classifier controls token thresholds and a priority control mechanism controls discarding thresholds every flow to utilize them for determining conditions. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a packet transmission method for guaranteeing quality of service (QoS) and related technology.
[0002]
[Prior art]
As a conventional QoS technology in an IP network, there is DiffServ (RFC2475) defined in the IETF (Internet Engineering Task Force).
[0003]
Next, the architecture of DiffServ will be described with reference to FIG. In DiffServ, QoS classes such as an EF (virtual leased line) class, an AF (minimum bandwidth guarantee) class, and a BE (best effort) class are defined as standards.
[0004]
In DiffServ, a range to which the same rule regarding quality assurance can be applied is referred to as “DS domain” 1. The DS domain 1 is used, for example, as an ISP (Internet Service Provider) network, a corporate network, or the like.
[0005]
The packet transmission devices (routers, switches, gateways, etc.) located at the boundary of DS domain 1 are called “edge nodes” 2, 3, and 4. The packet transmission devices located inside DS domain 1 are: Called "core nodes" 5,6.
[0006]
At the input interfaces of the edge nodes 2, 3, and 4, a traffic conditioner (Traffic Conditioner) 10 that monitors a packet 11 flowing into the DS domain 1 is provided (see FIG. 9).
[0007]
The traffic conditioner 10 has an MF (Multi-Field) classifier 12. The MF classifier 12 looks at the header information of the incoming packet 11, classifies which QoS class the packet belongs to, and measures the used bandwidth for each QoS class.
[0008]
Then, the MF classifier 12 sets a value in a DSCP (DiffServ Code Point: corresponding to priority) field of the incoming packet 11 according to the presence or absence of a contract breach.
[0009]
Also, as shown in FIG. 10, the core nodes 5 and 6 have a priority control mechanism including a BA (Behavior Aggregate) classifier 21. The classifier 21 performs classification by referring to only the DSCP value set by the edge nodes 2, 3, and 4, and performs a transfer process (priority control) of the packet 11 according to the DSCP value.
[0010]
[Problems to be solved by the invention]
(Issue 1) Traffic conditioner related
In the conventional DiffServ, traffic measurement, DSCP marking (priority setting), and packet discarding are performed using a mechanism called a token bucket.
[0011]
Hereinafter, for simplicity, a case where there are two classes, an AF class (high priority) and a class other than the AF class (low priority), will be described with reference to FIG.
[0012]
FIG. 9 is a block diagram of a conventional traffic conditioner. The traffic conditioner 10 is used as an input interface of the edge nodes 2, 3, and 4, and monitors the packet 11.
[0013]
Here, in the AF class, the minimum guaranteed band is determined as the contract band, and the MF classifier 12 checks whether the packet 11 is an AF class packet.
[0014]
If the packet 11 is other than the AF class, the MF classifier 12 sets a priority mark (DSCP3) of a class other than the AF class in the packet 11 and outputs the packet to the outside of the traffic conditioner 10.
[0015]
If the packet 11 is of the AF class, the MF classifier 12 outputs the packet 11 to the measurement priority setting unit 13 having a token bucket mechanism.
[0016]
The measurement priority setting unit 13 has a token buffer 14 in which tokens 15 are stored at a predetermined rate r. Upon input of the packet 11, the measurement priority setting unit 13 compares the packet length with the token amount stored in the token buffer 14 at the time of input.
[0017]
Here, if token amount ≧ packet length, the measurement priority setting unit 13 determines that the packet is within the guaranteed bandwidth, sets a high priority mark (DSCP1) on the packet 11, and sets the packet 11 to the traffic conditioner 10. Output to the outside of. At this time, the token in the token buffer 14 is reduced by the packet length of the packet 11.
[0018]
On the other hand, if the token amount <the packet length, the measurement priority setting unit 13 determines that the bandwidth exceeds the guaranteed bandwidth, sets a low priority mark (DSCP2) on the packet 11, and sets the packet 11 to the traffic conditioner 10. Output to the outside of.
[0019]
As described above, in the conventional traffic conditioner 10, even when there are a plurality of flows of the AF class, these flows are measured by the same measurement priority setting unit 13 without discrimination, and the flows of the flows having the same priority are measured. It is checked whether the total bandwidth is within the guaranteed bandwidth or exceeds the guaranteed bandwidth.
[0020]
Here, when the total bandwidth of the flows of the same priority exceeds the guaranteed bandwidth, the packets exceeding the guaranteed bandwidth among the flows of a plurality of AF classes (high-priority) are distinguished by the DSCP2 without discrimination between the flows. (Low priority).
[0021]
Since DSCP2 packets are not guaranteed for bandwidth, when congestion occurs, they are discarded. That is, there is a possibility that any AF class flow is uniformly discarded without distinction.
[0022]
This can be described phenomenally as follows. That is, when a plurality of users receive and view an image using packet communication of the AF class, if congestion occurs, the images viewed by all users are simultaneously disturbed. Will be.
[0023]
(Issue 2) Related to priority control mechanism
FIG. 10 is a block diagram of a conventional priority control mechanism. The priority control mechanism 20 shown in FIG. 10 is used as an output interface of the core nodes 5 and 6 and the edge nodes 2, 3 and 4 shown in FIG.
[0024]
The priority control mechanism 20 classifies the packet 11 based on the DSCP marked by the edge nodes 2, 3, and 4, and performs transfer processing (queuing and scheduling) according to the DSCP. Hereinafter, the case of two classes will be described for simplicity.
[0025]
The classifier 21 classifies which QoS class the packet 11 belongs to based on the DSCP.
[0026]
The classifier 21 inserts class 1 packets into the queue 22 and class 2 packets into the queue 23. However, when the queues 22 and 23 are full of packets, the packets are discarded without being inserted into the queues 22 and 23.
[0027]
The scheduler 24 determines the queues to be extracted and the amount of packets to be transmitted in accordance with an algorithm such as PQ (Priority Queueing) or WRR (Weighted Round Robin) for these queues 22 and 23. Then, in accordance with this determination, the packet is output to the outside of the priority control mechanism 20 from the queue 22 or the queue 23.
[0028]
Here, in the case of the AF (minimum bandwidth guarantee) class, the packet rate, that is, the usable bandwidth, which is serviced by the scheduler 24 differs between congestion and non-congestion.
[0029]
If the contracted bandwidth is exceeded at the time of congestion, DSCP2 packets, which should be given priority, may be discarded.
[0030]
The measurement of the packet and the setting of the priority (such as DSCP1 and DSCP2) are performed at the time of inputting the packet to the DS domain, and depend on the sender of the packet.
[0031]
However, there may be a case where it is desired to guarantee a band according to a band used by a receiver for a packet going out of the DS domain 1. In such a case, at the sender's base, even if a packet is marked as DSCP1 (high priority), if the sum of the packets exceeds the guaranteed bandwidth of the receiver, a situation occurs in which the packet is discarded.
[0032]
Therefore, an object of the present invention is to provide a packet transmission method and a related technique that can prevent the communication quality of all flows from being simultaneously degraded even when congestion occurs and the available bandwidth decreases.
[0033]
[Means for Solving the Problems]
In the packet transmission method according to the first aspect, when a plurality of flows composed of packets having the same priority share a guaranteed bandwidth, a priority of a packet belonging to at least one of the flows is determined; Among these flows, the priority of a packet belonging to a flow different from this one flow is handled with a difference.
[0034]
According to this configuration, the priority of the flow is made different, so that packets belonging to a flow that becomes relatively advantageous are not discarded, and the communication quality of this flow is maintained. As a result, even when congestion occurs and the available bandwidth decreases, it is possible to avoid a situation in which the communication quality of all flows is simultaneously degraded.
[0035]
In the packet transmission method according to the second aspect, the priority is made different according to the first-come-first-served order based on the communication start time of the flow.
[0036]
In the packet transmission method according to the third aspect, when the sum of the bandwidths of a plurality of flows composed of packets of the same priority exceeds the guaranteed bandwidth shared by these flows, the first-come-first-served basis based on the communication start time of the flows. Is handled so that the priority of the packet belonging to the flow whose communication start time is older is higher than the priority of the packet belonging to the flow whose communication start time is newer.
[0037]
In the packet transmission method according to claim 4, when the sum of the bandwidths of a plurality of flows composed of packets of the same priority exceeds the guaranteed bandwidth shared by these flows, the first-come-first-served basis based on the communication start time of the flows. , A higher priority is given to a packet belonging to a flow whose communication start time is older than a packet belonging to a flow whose communication start time is newer.
[0038]
In the packet transmission method according to the fifth aspect, when the sum of the bandwidths of a plurality of flows composed of packets having the same priority exceeds the guaranteed bandwidth shared by these flows, the first-come-first-served basis based on the communication start time of the flows. , Packets belonging to a flow whose communication start time is newer are discarded before packets belonging to a flow whose communication start time is older.
[0039]
With these configurations, a flow with an earlier communication start time can be advantageously handled, and the new flow can perform reasonable communication control without deteriorating the quality of the old flow.
[0040]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(Embodiment 1)
The present embodiment relates to a traffic conditioner that is used as an input interface of an edge node and monitors a packet flowing into a DS domain.
[0041]
Hereinafter, for simplicity, a case where there are two classes, an AF class (high priority) and a class other than the AF class (low priority), will be described as an example. In this embodiment, there are three or more classes. The same can be applied to the case.
[0042]
FIG. 1 is a block diagram of a traffic conditioner according to Embodiment 1 of the present invention. As shown in FIG. 1, the traffic conditioner 30 has the following elements.
[0043]
First, the measurement priority setting unit 32 measures a token and sets the priority of the packet 11, and has a token buffer 34 in which tokens 35 are accumulated at a predetermined rate r.
[0044]
The MF classifier 31 receives the packet 11, outputs a high-priority packet (AF class) to the measurement priority setting unit 32, and outputs a low-priority packet (other than the AF class) to the outside of the traffic conditioner 30. Output to
[0045]
As shown in FIG. 2, the flow management unit 33 stores a token parameter for each flow, a flow management table 40, searches for predetermined data from the table 40, and registers predetermined data in the table 40. And a flow search registration unit 41.
[0046]
In the present embodiment, this table 40 stores the flow No. , Header information, and token threshold, and manages data for each flow. This token threshold is a value corresponding to the token parameter. Other examples of the token parameter include a coefficient for multiplying the token amount.
[0047]
As shown in FIG. 1, when the MF classifier 31 receives the packet 11, the MF classifier 31 outputs the header information to the flow management unit 33 as flow information.
[0048]
When the header information is input from the MF classifier 33, the flow management unit 33 uses the flow search registration unit 41 and the flow management table 40 to set the token threshold A of the flow corresponding to the header information to the measurement priority setting unit 32. Output to
[0049]
When the packet 11 (AF packet) is input from the MF classifier 31, the measurement priority setting unit 32 compares the packet length with the token amount stored in the token buffer 14 at the time of input.
[0050]
The measurement priority setting unit 32 calculates a value obtained by subtracting the token threshold A obtained from the flow management unit 33 from the token amount in the token buffer 34 (the token is corrected by the token parameter input from the flow management unit 33), The packet length of the packet 11 is compared in size.
[0051]
That is, if (current token amount of token buffer 34−token threshold A) ≧ packet length, the measurement priority setting unit 32 assigns a high-priority mark (DSCP1) to this packet 11 within the guaranteed band. And output.
[0052]
Conversely, if (current token amount of token buffer 34-token threshold A) <packet length, the measurement priority setting unit 32 determines that the guaranteed bandwidth is exceeded and marks the packet 11 with a low priority. (DSCP2) and output.
[0053]
If (current token amount of token buffer 34-token threshold A) = packet length, regardless of the above, it is determined that the guaranteed bandwidth is exceeded, and a low priority mark (DSCP2) is added. The packet 11 may be output.
[0054]
Here, as illustrated in FIG. 2, the flow management unit 33 assigns a small token threshold in order from the flow started first.
[0055]
For example, when three flows are started in the order of flow A, flow B, and flow C, the flow management unit 33 sets the token thresholds Ta, Tb, and Tc for these flows to Ta <Tb <Tc. Assign values as follows.
[0056]
In the example of FIG. 1, No. 2, No. Since the communication is started in the order of 3, the token thresholds are “0”, “3000”, and “6000” in this order.
[0057]
In addition, the flow management unit 33 deletes an entry from the flow management table 40 for a flow in which a packet does not arrive for a predetermined time or more.
[0058]
By doing so, a token that is more likely to be acquired as the flow starts earlier is assigned DSCP1 (high-priority mark) preferentially from the flow that started earlier.
[0059]
As a result, even in a situation in which DSCP2 packets are discarded during network congestion, packet discard does not occur for flows that have started earlier and that fall within the guaranteed bandwidth, and packet congestion does not occur. Packet discarding over all flows can be prevented, and simultaneous degradation of all images in video communication can be suppressed.
[0060]
As described above, according to the traffic conditioner of the present embodiment, the following packet transmission can be realized.
[0061]
That is, in a network that guarantees QoS, when a plurality of flows (consisting of packets having the same priority) share a guaranteed bandwidth, the priority of a packet belonging to a certain flow and the priority of a packet belonging to another flow are determined. Is given a difference.
[0062]
More specifically, according to the first-come-first-served order based on the communication start time of a flow, a higher priority is given to a packet belonging to a flow whose communication start time is older than a packet belonging to a flow whose communication start time is newer.
[0063]
Therefore, a packet belonging to a flow whose communication start time is newer is discarded before a packet belonging to a flow whose communication start time is older.
[0064]
As shown in FIG. 1, the measurement priority setting unit 32 is shared for a plurality of flows. Here, it is conceivable to provide the measurement priority setting units 32 by the number of flows. However, this imposes a heavy burden on system resources.
[0065]
In this embodiment, in consideration of this point, only the value of the token threshold is separately provided for each flow as shown in FIG. 2, so that processing is different for each flow and the load on system resources is reduced. That is achieved at once.
[0066]
(Embodiment 2)
The present embodiment relates to a priority control mechanism used as an output interface of a core node or an edge node. This priority control mechanism classifies the packet 11 based on the DSCP marked by the edge node, and performs transfer processing (queuing and scheduling) according to the DSCP.
Hereinafter, for simplicity, a case of two classes will be described as in the first embodiment. However, in the present embodiment, even when there are three or more classes, the number of queues is increased and the like. Applicable.
[0067]
FIG. 3 is a block diagram of a priority control mechanism according to Embodiment 2 of the present invention. As shown in FIG. 3, the priority control mechanism 50 has the following elements.
[0068]
First, the priority control mechanism 50 has a plurality of queues 53 and 54 provided for each priority class (two classes in this example) of the packet 11. A class 1 packet is inserted into the queue 53, and a class 2 packet is inserted into the queue 54. Of course, if the discard condition described later is satisfied, the packet is discarded without being inserted into any queue.
[0069]
The classifier 51 receives the packet 11 and classifies the packet into class 1 and class 2 according to the priority.
[0070]
The queue management unit 52 inputs the packet classified by the classifier 51, and inserts the packet into one of the queues 53 and 54 unless the discard condition relating to the packet is satisfied.
[0071]
The scheduler 55 sequentially takes out packets from the queues 53 and 54 according to an algorithm such as PQ or WRR and outputs the packets to the outside of the priority control mechanism 50.
[0072]
The flow management unit 56 has a configuration very similar to the flow management unit 33 of the first embodiment. That is, as shown in FIG. 4, the flow management unit 56 searches the flow management tables 57 and 58 and predetermined data from these tables 57 and 58, and registers predetermined data in these tables 57 and 58. And a flow search registration unit 59. These tables 57 and 58 hold a discard threshold for each flow.
[0073]
Here, in the present embodiment, the flow management tables 57 and 58 are provided for each class. However, if care is taken so that information is not mixed, a single table may be used.
[0074]
In the present embodiment, these tables 57 and 58 store the flow No. , Header information, and a discard threshold, and manages data for each flow. This discard threshold is a value corresponding to the discard parameter. Other discard parameters include a coefficient by which the queue length is multiplied.
[0075]
When the header information (corresponding to the flow information) is input from the classifier 51, the flow management unit 56 uses the flow search / registration unit 59 and the flow management tables 57 and 58 to set the discard threshold of the flow corresponding to the header information, Output to the queue management unit 52.
[0076]
Using the current queue length of the class corresponding to the packet received from the classifier 51 and the discard threshold received from the flow manager 56, the queue management unit 52 inserts the packet into the queue under the following discard condition. Or discard it.
[0077]
That is, if (current queue length + packet length)> discard threshold, the packet is discarded, and if (current queue length + packet length) ≦ discard threshold, the packet is inserted into the queue of the corresponding class. .
[0078]
Here, as illustrated in FIG. 4, the flow management unit 56 assigns a large discard threshold in order from the flow started earlier.
[0079]
For example, when communication of three flows is started in the order of flow X, flow Y, and flow Z, the flow management unit sets the discard thresholds for these flows to Tx, Ty, and Tz such that Tx>Ty> Tz. There is a value to be assigned.
[0080]
Incidentally, in the example of FIG. 4, the flow No. 1 is the first to start the communication. 2, flow No. The order is 3. Therefore, in this order, the discard thresholds are sequentially reduced, such as “60000”, “55000”, and “50000”.
[0081]
In addition, the flow management unit 56 deletes the entry from the flow management tables 57 and 58 for a flow in which a packet does not arrive for a predetermined time or more.
[0082]
By doing so, the flow that started earlier is less likely to be discarded, and the flow that started earlier is preferentially inserted into the queue and can be serviced by the scheduler.
[0083]
As a result, the rate serviced by the scheduler during network congestion decreases, and even if packets accumulate in the queue, priority is given to packets of flows that started earlier and fall within the guaranteed bandwidth. The packet is inserted into the queue and the packet is not dropped.
[0084]
With this configuration, it is possible to prevent packet discarding over all flows at the time of congestion, and to suppress simultaneous deterioration of all images in the case of video communication.
[0085]
Note that, as shown in FIG. 3, the queue management unit 52 is shared by a plurality of flows. Here, it is conceivable to provide as many queue management units 52 as the number of flows. However, this imposes a heavy burden on system resources.
[0086]
In the present embodiment, taking this point into account, as shown in FIG. 4, only the value of the discard threshold is separately provided for each flow, so that processing is different for each flow and the load on system resources is reduced. That is achieved at once.
[0087]
(Modification 1)
Next, modifications of the first and second embodiments will be described. Modifications 1 to 3 below are modifications of the flow management unit 33 of FIG. 2 according to the first embodiment and the flow management unit 56 of the second embodiment. The same applies if the term is replaced with “discard threshold” in the second embodiment.
[0088]
Therefore, in order to avoid repetition, the following description will be made only along the first embodiment, and the description along the second embodiment will be omitted.
[0089]
As illustrated in FIG. 5, the flow management unit 60 according to the first modification sets the token threshold (discard threshold in the second embodiment) of a flow in which communication has continued for a certain period of time or more such that the flow is disadvantageous. Change to
[0090]
That is, a new “duration” field is added to the flow management table 62 in the flow management unit 60, and the duration from the start of the flow is measured by the flow management unit 60.
[0091]
For the flow whose duration exceeds a certain time, the token threshold set first is changed to a value larger than the largest token threshold in the current flow management table 62. As a result, this flow is disadvantageously treated.
[0092]
For example, the flow No. in FIG. The value 1 indicates that the token threshold is changed from the current value “0” to a value larger than the largest token threshold (eg, “9000”) because the duration is long.
[0093]
As a result, a packet of a flow after a lapse of a certain time is most likely to be marked as DSCP2 (low priority), and a flow protected preferentially on a first-come-first-served basis occupies a band for a long time. It can be prevented from continuing.
[0094]
(Modification 2)
In this example, the “duration” of the first modification is changed to “cumulative usage”.
[0095]
As illustrated in FIG. 6, the flow management unit 70 according to the second modification sets a token threshold (discard threshold in the second embodiment) of a flow in which the accumulated usage is equal to or more than a certain value so that the flow becomes disadvantageous. Change to
[0096]
That is, a field of “cumulative usage” is newly added to the flow management table 72 in the flow management unit 70, and the flow management unit 70 measures the cumulative usage since the start of the corresponding flow.
[0097]
For a flow in which the accumulated usage exceeds a certain value, the token threshold set first is changed to a value larger than the largest token threshold in the current flow management table 72. As a result, this flow is disadvantageously treated.
[0098]
For example, the flow No. in FIG. The value 1 indicates that the accumulated threshold value is large, and the token threshold is changed from the current value “0” to a value larger than the largest token threshold (eg, “9000”).
[0099]
As a result, a packet of a flow whose accumulated usage exceeds a certain value is more likely to be marked with the DSCP2 (low priority) mark, and a flow that is preferentially protected on a first-come, first-served basis for a long time. , It is possible to prevent the band from being continuously occupied.
[0100]
In this way, by limiting the amount, a flow requiring a larger bandwidth is limited in a shorter time.
[0101]
(Modification 3)
In this example, as shown in FIG. 7, a timer is provided for each flow, and compared with a method of determining the end, a smaller amount of hardware resources are required, and the flow management unit 80 can be easily implemented by hardware. Is devised in the configuration of FIG.
[0102]
That is, the flow management unit 80 according to the present example has counters 1, 2, and 3 in addition to the flow search and registration unit 81 and the flow management table 82. These counters 1, 2, and 3 operate as packet counters.
[0103]
When a packet arrives, the flow search registration unit 81 sets the packet counter of the flow to which the packet belongs to 0, and increments the packet counters of the other flows by one.
[0104]
As a result, the counter relating to the flow in which packets continuously arrive is reset to 0 each time a packet arrives, but the counter value of the flow in which no packet arrives increases.
[0105]
Then, the flow search registration unit 81 determines that the flow for which the packet counter has reached a certain value has been completed, and deletes the entry from the flow management table 82.
[0106]
Here, when the flow management is performed by software, a timer is provided for each flow, and a flow in which a packet has not arrived for a certain period of time or more may be deleted from the flow management table assuming that the flow has ended.
[0107]
However, this makes it difficult to implement hardware. This is because the number of flows can be quite large in some cases, and it is difficult to provide a large number of timers due to system resource limitations.
[0108]
In this embodiment, by providing the above-described counter, the load on the system resources is reduced, and processing equivalent to providing a large number of timers can be realized.
[0109]
【The invention's effect】
According to the present invention, the following effects can be obtained.
At the time of congestion, even if the usable bandwidth decreases, it is possible to prevent packets of all flows from being uniformly discarded, and to suppress deterioration of communication quality for all flows.
[0110]
Simultaneous image quality degradation of all images can be prevented even when congestion occurs and available bandwidth decreases when performing a plurality of video communications using a minimum bandwidth guarantee service in a DiffServ communication quality assurance network. .
[0111]
Regardless of whether the DiffServ method is used or not, in a general communication quality assurance method, when performing a plurality of video communications using a service in which a serviceable band fluctuates, congestion occurs. Even when the usable bandwidth is reduced, it is possible to prevent simultaneous image quality deterioration of all images.
[0112]
It is possible to prevent a flow that is preferentially protected on a first-come-first-served basis from continuously using a band for a long time.
[0113]
It is possible to prevent flows that are protected with priority on a first-come-first-served basis from continuing to use a large band.
[0114]
By limiting the amount, a flow requiring a larger bandwidth is limited in a shorter time.
[0115]
Compared with a method in which a timer is provided for each flow and the termination is determined, the number of hardware resources is reduced, and realization with hardware becomes easy.
[Brief description of the drawings]
FIG. 1 is a block diagram of a traffic conditioner according to Embodiment 1 of the present invention.
FIG. 2 is a block diagram of the flow management unit.
FIG. 3 is a block diagram of a priority control mechanism according to a second embodiment of the present invention;
FIG. 4 is a block diagram of the flow management unit.
FIG. 5 is a block diagram of a flow management unit according to a first modification of the present invention.
FIG. 6 is a block diagram of a flow management unit according to a second modification of the present invention.
FIG. 7 is a block diagram of a flow management unit according to a third modification of the present invention.
FIG. 8 is an explanatory diagram of a DiffServ architecture.
FIG. 9 is a block diagram of a conventional traffic conditioner.
FIG. 10 is a block diagram of a conventional priority control mechanism.
[Explanation of symbols]
11 packets
30 Traffic Conditioner
31 MF Classifier
32 Measurement priority setting section
33, 56, 60, 70, 80 Flow management unit
34 token buffer
35 tokens
40, 57, 58, 62, 72, 82 Flow management table
41, 59, 61, 71, 81 Flow search registration unit
50 Priority control mechanism
51 Classifier
52 Queue Management Unit
53, 54 queue
55 scheduler

Claims (19)

A packet transmission method for QoS guarantee,
When a plurality of flows composed of packets of the same priority share the guaranteed bandwidth, the priority of packets belonging to at least one of the flows and the one of the flows A packet transmission method that handles packets with different priorities belonging to flows different from the above. 2. The packet transmission method according to claim 1, wherein the priority is made different according to a first-come-first-served order based on a communication start time of the flow. If the sum of the bandwidths of a plurality of flows composed of packets of the same priority exceeds the guaranteed bandwidth shared by these flows, the flow whose communication start time is older according to the first-come-first-served order based on the flow communication start time The packet transmission method according to claim 1, wherein the priority of the packet to which the packet belongs is set to be higher than the priority of the packet belonging to the flow whose communication start time is newer. When the sum of the bandwidths of a plurality of flows composed of packets of the same priority exceeds the guaranteed bandwidth shared by these flows, the flow whose communication start time is older according to the first-come-first-served order based on the communication start time of the flow. 4. The packet transmission method according to claim 1, wherein a higher priority is given to a packet belonging to the packet than to a packet belonging to a flow having a newer communication start time. If the sum of the bandwidths of a plurality of flows composed of packets of the same priority exceeds the guaranteed bandwidth shared by these flows, the flow whose communication start time is newer according to the first-come-first-served order based on the flow's communication start time. The packet transmission method according to claim 1, wherein the belonging packet is discarded earlier than the packet belonging to the flow whose communication start time is older. A measurement priority setting unit that measures the token and sets the priority of the packet;
An MF classifier that inputs a packet, outputs a high-priority packet to the measurement priority setting unit, and outputs a low-priority packet to the outside;
A flow management unit that holds a token parameter for each flow,
The flow management unit inputs flow information of a packet from the MF classifier, and outputs a token parameter corresponding to the flow information to the measurement priority setting unit;
The measurement priority setting unit compares the size of a token corrected by a token parameter input from the flow management unit with the packet length of a packet, and sets the priority of the packet according to the comparison result. Conditioner. The traffic conditioner according to claim 6, wherein the measurement priority setting unit is shared for a plurality of flows. The traffic conditioner according to claim 6, wherein the token parameter is a token threshold value subtracted from the token in the modification of the measurement priority setting unit, and the flow information is packet header information. 9. The traffic conditioner according to claim 6, wherein the flow management unit sets a token parameter in accordance with a first-come-first-served order based on a communication start time of a flow so that a flow whose communication has started earlier has an advantage. The traffic conditioner according to claim 6, wherein the flow management unit changes a token parameter of a flow for which communication has continued for a predetermined time or more so that the flow is disadvantageous. The traffic conditioner according to claim 6, wherein the flow management unit changes a token parameter of a flow whose accumulated usage exceeds a predetermined amount so that the flow is disadvantageous. The flow management unit has a packet counter for each flow, sets the packet counter of the corresponding flow to zero each time a packet arrives, increments the packet counter of the other flows by one, and sets the packet counter of a flow exceeding a certain value. 12. The traffic conditioner according to claim 6, wherein the management of the traffic conditioner is ended. A plurality of queues provided for each class of packet priority;
A classifier that enters packets and classifies them according to priority;
A queue management unit that inputs a packet classified by the classifier and inserts the packet into any of the plurality of queues, unless a discard condition relating to the packet is satisfied.
A flow management unit that holds a discard parameter for each flow,
The flow management unit inputs flow information of a packet from the classifier, outputs a discard parameter corresponding to the flow information to the queue management unit,
A priority control mechanism, wherein the discard condition is determined based on a packet length of the packet, a queue length of a queue related to the packet, and a discard parameter of a flow related to the packet. 14. The priority control according to claim 13, wherein the discard condition indicates that when the sum of the packet length of the packet and the queue length of the queue related to the packet is larger than the discard parameter, the packet is discarded. mechanism. 15. The priority control mechanism according to claim 13, wherein the queue management unit is shared for a plurality of flows. 16. The priority control mechanism according to claim 13, wherein the flow management unit sets a discard parameter in accordance with a first-come-first-served basis based on a communication start time of a flow so that a flow whose communication has started earlier has an advantage. 17. The priority control mechanism according to claim 13, wherein the flow management unit changes a discard parameter of a flow for which communication has been continued for a predetermined time or more so that the flow is disadvantageous. 18. The priority control mechanism according to claim 13, wherein the flow management unit changes a discard parameter of a flow whose accumulated usage exceeds a certain amount so that the flow is disadvantageous. The flow management unit has a packet counter for each flow, sets the packet counter of the corresponding flow to zero each time a packet arrives, increments the packet counter of the other flows by one, and sets the packet counter of a flow exceeding a certain value. 19. The priority control mechanism according to claim 13, wherein management of the priority control is terminated.

Images