FI104602B - Flow control in a telecommunications network - Google Patents

Description

1 104602

Flow control in a telecommunications network

Field of the Invention

The invention relates generally to flow control in a telecommunications network. More specifically, the invention relates to congestion control in a packet switched communication network, particularly in a network where TCP (Transmission Control Protocol) is used as a transport layer protocol.

BACKGROUND OF THE INVENTION As is well known, TCP is the most popular protocol used for transport layer communication. It provides reliable data communication between two hosts. (A host refers to a network-connected computer or any system that can be network-connected to provide services to another host connected to the same network.) 15 TCP uses a number of procedures to maximize transmission performance by monitoring various variables associated with the transmission. TCP includes, for example, an internal algorithm to avoid congestion.

ATM (Asynchronous Transfer Mode), on the other hand, is a (newer) connectivity packet switching technology that has been selected by the international telecommunications industry standardization organization ITU-T as the basic solution for Broadband Digital Multiple Service (B-ISDN). In the ATM network, problems with traditional packet networks have been eliminated by using short, fixed-length (53 bytes) packets, called cells. ATM networks are rapidly being adopted as backbone networks in various parts of TCP / IP networks (such as the Internet).

25 Although ATM is designed to provide end-to-end transport layer-level service, it is very likely that in the future networks will be implemented such that (a) TCP / IP remains the de-facto standard for networks and (b) only part of the end-to-end . Thus, while ATM utilization continues, TCP will still need 7:30 to provide end-to-end transmission services.

Implementing ATM also means that implementations must be able to support a large number of existing data applications where TCP is commonly used as a transport layer protocol. Several approaches for congestion management of ATM networks 35 have been developed in the past to transfer existing protocols of the upper layers to ATM networks.

2,104,602

Congestion control is a common problem in managing traffic on packet networks. Congestion refers to a situation where the number of transfer requests at a given time exceeds the capacity of a given point in the network (called a bottleneck resource). Congestion usually results in over-5 load conditions. As a result, buffers, for example, leak, causing the network or subscriber to retransmit packets. Generally, congestion occurs when the amount of traffic to a given link exceeds the link's output capacity. The main function of congestion control is to ensure good throughput and delay performance while maintaining a fair allocation of network resources to users. For TCP traffic, which often has bursting traffic patterns, congestion control poses a challenging problem. It is known that packet loss leads to a significant loss of TCP throughput. Therefore, for best throughput, the number of lost packets should be kept to a minimum.

The present invention relates to congestion management in packet networks. For the reasons stated above, most such networks are, and will be in the foreseeable future, TCP networks or TCP / ATM networks (i.e., networks where TCP provides end-to-end transmission services and ATM provides the "bit-pipes" below). The following is a brief description of the congestion control mechanisms of these networks.

ATM Forum has defined five different service classes that link traffic characteristics and Quality of Service (QoS) to network behavior. These service categories are: Constant Bit Rate (CBR), Real-time Variable Bit Rate (rt-VBR), Available Bit Rate (ABR) and Unspecified Bit Rate (UBR). Rate). These service categories divide traffic into guaranteed traffic and so-called. "Best effort" traffic, which is the traffic that fills up the remaining bandwidth after the service is provided to the guaranteed traffic.

'' '30 One possible solution for "best effort" traffic is to use ABR flow control. The main idea in ABR flow control is to use special cells, so-called RM cells (Resource Management), to control the speeds of the sources. factors such as available bandwidth, congestion status, and impending congestion) by sending RM cells between data cells .RM cells are bounced back to the target and sent back to the source .At the time ATM switches can write the traffic information to these RM cells. the returned RM cells may increase or decrease their rate or maintain their current rate, depending on the kind of information the cells bring.

In TCP / ATM networks, the source and destination are connected through 5 IP / ATM / IP subnets. Figure 1 illustrates a connection between a TCP source A and a TCP destination B in a network where the communication path passes through an ATM network using ABR flow control. When congestion is detected on the ATM network, ABR speed monitoring begins to take effect, forcing the edge router R1 to reduce its transmission speed in the direction of the ATM network. Thus, the purpose of the ABR-10 monitoring loop is to command network ATM sources to reduce their transmission speed. If the congestion is not alleviated, the router's buffer will reach its maximum capacity. As a result, the router begins to drop packets, resulting in a reduction of the TCP congestion window (the congestion window concept will be explained in more detail below.) 15 For congestion control, the network of Figure 1 comprises two independent control loops: an ABR control loop and a TCP control loop. However, such congestion control based on two congestion management systems on different protocol layers can have an unexpected and unwanted impact on network performance. More specifically, the inner control loop-20 loop (ABR loop) can cause unexpected delays in the outer control loop (TCP loop).

An alternative approach to “best effort” traffic is to use the UBR service with sufficiently large buffers and allow upper-layer protocols such as TCP to handle congestion or congestion. Figure 25 2 illustrates such a network, i.e. a TCP / UBR network. The nodes of such a network comprise packet rejection mechanisms that reject packets or cells when congestion occurs. When a packet is dropped somewhere in the network, the corresponding TCP source does not receive acknowledgment. As a result, the TCP source reduces its transmission rate.

30 The UBR Service does not use flow control and does not rely on numerical guarantees for quality of service; this is why it is also the cheapest service to offer. However, due to its simplicity, UBR alone without sufficient buffer size gives poor performance in congested networks.

More sophisticated congestion control mechanisms have been proposed to overcome this disadvantage. One such is the early packet discard (EPD) system. In the EPD system, the ATM switch destroys entire packets before 4 104602 buffer overflows. In this way, the throughput of the TCP / ATM network can be significantly improved, since ATM switches do not have to send cells of a packet containing corrupt cells, i.e. cells belonging to packets of which at least one cell has been rejected (these packets would be discarded Another advantage of the EPD system is that it is relatively inexpensive to implement in an ATM switch. Those interested in the subject will find a detailed description of the EPD method in, e.g., A. Romanow and S. Floyd, Dynamics of TCP Traffic over ATM Networks, Proc. ACM SIGCOMM '94, p. 79-88, August 1994.

However, the EPD method treats users unfairly.

This is because the EPD destroys entire packets on all connections, regardless of their current speeds or their relative proportions in the buffer, i.e., their relative impact on the congestion situation. To overcome this disadvantage, several variants have been proposed for selective packet rejection. One of these is described in Röhit Goyal, Perfomnance of TCP / IP over UBR +, ATM_Forum / 96-1269. This method uses the FIFO buffer in the switch and performs virtual connection accounting to keep track of the share of each virtual connection in the buffer. In this way, only cells of overloaded connections can be discarded, while underloaded connections can increase their throughput.

In spite of all the above improvements, prior art congestion control methods still have the disadvantage of not being able to provide early warning to a traffic source when overloading is detected on the network. In other words, the traffic source is not quickly informed of the overload so that the source can reduce its transmission speed.

BRIEF SUMMARY OF THE INVENTION

The object of the invention is to eliminate the above disadvantage and to provide a method which makes it possible to use a simple implementation; 30, inform the traffic source at a very early stage that the network is • becoming congested and request the source to reduce its transmission speed.

It is also intended that the method allow efficient collaboration between the TCP and ATM flow control mechanisms.

This objective can be achieved by using the solution presented in the independent claims.

5, 104602

The basic idea of the invention is to delay the acknowledgments that are being transferred from the destination to the sender. This can be done at the same point in the network where the congestion is detected, or alternatively, the network point that detects the congestion or overload can control another point 5 of the network to delay acknowledgments. Thus, in the invention, congestion control is performed on the return path of the connection, whereas known systems control traffic along the route. Rather than rejecting packets or cells along the route, the network according to the invention delays acknowledgments on the return path and thus causes the TCP source to reduce its transmission rate.

The invention provides a low-cost solution to provide an early warning to a TCP source that an overload or congestion is threatening the network. It is also important to note that there is no need to change the TCP protocol. A congestion control algorithm must be introduced to implement the invention in the network, but many existing algorithms related to TCP / UBR networks can be used for this purpose with only minor changes.

In addition, the present invention can smooth out the output rate of a TCP source, which in turn leads to more efficient bandwidth utilization. In addition, buffer capacity requirements are reduced as (rate) variation is reduced.

According to a preferred embodiment of the invention, load level information is transmitted from an ATM network point in RM cells to a node providing access to the ATM network and acknowledgments are delayed in said access node based on the information contained in the RM * cells. In this way, TCP and ATM flow control mechanisms can be made interdependent so that they work effectively together.

With the invention, connection performance can be significantly improved, especially in large latency networks.

List of patterns,; Fig. 1 illustrates a TCP communication path through an ABR-based ATM subnet, Fig. 2 illustrates a TCP connection through a UBR-based ATM subnet, Fig. 3 illustrates a flow control loop according to the invention 35 in a TCP / ATM network, Figure 4b illustrates one possible embodiment of a new method in an IP switch, Figure 4b is a time diagram showing significant time points in the implementation of Figure 4a, 5 Figure 5 is a flow diagram illustrating a method for determining delay values, Figure 6a illustrates another embodiment of delay delaysb an alternative embodiment, Fig. 7a illustrates an embodiment of the method in an IP network, Fig. 7b illustrates another embodiment of the method in an IP network, Fig. 8a illustrates an embodiment of the method in an ATM network, Fig. 8b illustrates a method another embodiment of the ATM network, FIG. 9 illustrates an interaction of TCP and ATM flow control loops 20 according to a preferred embodiment of the invention, FIG. 10 illustrates an example packet transfer between a traffic source and a destination in a known TCP network, and FIG. 11 illustrates an example between a subject in a TCP network using the method of the invention.

Detailed Description of the Invention

Figure 3 illustrates the basic principle of the invention by showing a connection between two user terminals (A and B) on a TCP / ATM network in which; 30 terminals use TCP as a transport layer protocol. In addition to the access nodes (AN1 and AN2) of the terminals, the figure shows only one intermediate node (N1) and the transmission lines (TL1 and TL2) connecting the nodes.

The connection between hosts A and B, like any other TCP connection, begins by negotiating between hosts to open the connection. Initial negotiation is called a three-phase handshake because during this handshake three opening segments are moved. The term "segment" 7,104,0602 refers to an information unit transmitted by TCP to the Internet Protocol (IP). The IP headers are associated with these TCP segments to form IP datagrams, i.e., the TCP segments are transferred to a receiver within the IP datagrams. During the initial handshake process, the host computers inform each other, for example, of the maximum segment size they accept, in order to avoid fragmentation of TCP segments as this would significantly reduce the performance of the TCP connection.

After the initial handshake is complete, the hosts start transmitting data using TCP segments. Each error-free TCP segment 10 is acknowledged, including handshake segments. To illustrate the basic idea of the invention, suppose that host A transmits one TCP segment to host B, on the network layer host A adds an IP header to this TCP segment to form an IP datagram. The telegram is transformed into standard ATM cells at the access node AN1 located at the edge of the ANM of the ATM-15 network. The cells of the telegram are then routed through the ATM network to the access node AN2 of the host B. This access node reconstructs the original IP datagram from incoming cells and sends it to host B. Host B removes the IP header to expose the TCP segment. If the segment is received correctly, host B sends an acknowledgment TCP-20 segment ACK1 back to host A. So far, the network has operated in a known manner.

The network load is monitored at the access node AN1, e.g. by monitoring the fill rate of one or more buffers that buffer traffic to the ATM network. If an overload is detected, i.e., if the buffer fill rate exceeds 25 predetermined limits, a congestion message CM is sent inside the node to delay the acknowledgments currently passing through the switch towards the traffic sources. Thus, the reset (ACK1) we use as an example is also delayed as it passes through the access node AN1, provided that the node AN1 is overloaded during that period.

·: 30 TCP is one of the few transport protocols that has an internal congestion control mechanism. The solution according to the invention is dependent on this known TCP monitoring mechanism, no other monitoring mechanisms are needed at the source or at the destination. In the following, this mechanism is briefly described.

TCP congestion control is based on two variables: the declared window (Wrcvr) of the receiver 35 and the congestion window (CNWD). The advertised window of the receiver is maintained at the receiver as a measure of the buffering capacity of the receiver, and the congestion window is maintained at the transmit end as a measure of network capacity. A TCP source can never send more segments than the minimum specified in the receiver's declared window and congestion window.

5 The TCP congestion control method consists of two steps: slow start and congestion avoidance. The SSTHRES (slow start threshold) variable is maintained at source to distinguish between the two steps. The source starts the transmission according to the slow start by transmitting one TCP segment, i.e. the value of CWND is set to 1 at the beginning. When the source receives the acknowledgment, it increments the CWND by one and, as a result, sends two segments. In this way, the value of the CWND is doubled in the slow start phase during each time period corresponding to the round trip as the destination terminal acknowledges each segment. The slow start phase ends and the congestion phase begins when the CWND 15 reaches the value SSTHRES.

If the packet is lost on a TCP connection, the source will not receive an acknowledgment and its timeout will be triggered. The source sets the SSTHRES variable to half the value of the CWND variable at the time of packet loss. More specifically, SSTHRES is set to max {2, min {CWND / 2, 20 Wrcvr}} and CWND is set to one. As a result, the source enters a congestion phase during which it increases the CWND by 1 / CWND each time the segment is acknowledged.

Since the invention does not in any way alter the known TCP congestion control mechanisms described above, it will not be described further here. 25 Interested persons can find more detailed information in several books in the field. (See, e.g., W. Richard Stevens, TCP / IP Illustrated Volume 1, The Protocols, Addison-Wesley, 1994, ISBN 0-201-63346-9).

According to the invention, one or more acknowledgments are delayed when overload or congestion is detected at a network point. In this way, the TCP source, ·: 30, which operates as described above, will automatically begin to reduce its transmission rate, or at least it will not increase its transmission rate as fast as it otherwise would. This is because the delay reduces the rate at which the source increases the size of its congestion window.

Figure 4a illustrates this principle by providing an example in which acknowledgments 35 are delayed at the output port OP of the IP switch. The load measurement unit LMU measures the load level of the switch by measuring the fill levels of the buffers which buffer the outgoing traffic through the switch. It should be noted, however, that the load level can be determined in any known manner.

The IP data telegrams passing through the switch in the reverse direction are first routed to the correct output port. Data telegrams received at this port are stored in a FIFO-type output buffer OB.

The traffic splitter TS reads the stored buffers from the output buffer, packet by packet from the first buffer location ML1. The traffic splitter works as follows.

10 If the congestion signal CS from the load measurement unit indicates that the load on the switch is below a predetermined level, the traffic branch transmits all data telegrams (packets) directly to the outgoing transmission link OL, whether or not they contain acknowledgments.

On the other hand, if the congestion signal CS indicates that the load level has reached a predetermined level, the traffic branch begins to read the acknowledgment bit of the TCP header inside each IP telegram. If this bit is valid, that is, if the data telegram contains an acknowledgment, the traffic splitter forwards the packet to the acknowledgment buffer AB. If the bit is not valid, the traffic splitter will forward the packet directly to the outgoing transmission OL. Thus, only packets containing acknowledgment 20 are delayed.

In the acknowledgment buffer, each data telegram is delayed for a certain period of time. The length of the period is preferably directly proportional to the current load level measured by the LMU. After each outgoing. when the acknowledgment packet delay time has elapsed, the packet is transmitted to the outgoing transmission link.

If ACKTi represents the time at which the packet containing the acknowledgment is transferred from the traffic branch to the acknowledgment buffer and if ACKTo represents the moment at which the packet is transferred from the acknowledgment buffer to the transmission link, ACKTo may be represented as:. j 30 ACKTo (j) = ACKTi (j) + dj (j = 1,2, ...

where j is the sequence number of the packet and dj is the value of the delay associated with the packet having the sequence number j.

Figure 4b illustrates the times when packets leave the traffic splitter and the acknowledgment buffer. Suppose that an excessive load is detected after a moment ACKTo (7) (before this, the acknowledgments have not been delayed). If the congestion signal received by the delay control unit DCU indicates that the load level 10410602 has exceeded a predetermined value, the delay control unit performs an algorithm that determines how long the next packet to be transmitted to the transmission link should be delayed. The calculated value may depend on one or more parameters, such as the current traffic speed, the current buffer load rate, or the previous value of the delay (dj,). As seen in Figure 4b, the delay value may vary from packet to packet.

FIG. 5 is a flowchart illustrating an example of an algorithm performed by a delay control unit for each packet read out from acknowledgment buffer AB.

10 If congestion is detected, the delay value dj of the packet currently read out from the acknowledgment buffer (i.e., the amount of time the current packet is delayed in the buffer) is calculated by the following formula: dj = adH + (1-a) dM (1) where dj. , dM is the measured delay value and a 15 is the smoothing factor (preferably α <0.5 <1). The measured delay is the actual delay measured from the moment the packet is received in the acknowledgment buffer to the moment the packet is read out from the acknowledgment buffer. This delay can be measured as an average over a given time period or as an average over a certain number of packets. The delay control unit can perform this measurement.

20 If congestion is detected and if dj, = 0 and dM = 0, that is, if there was no delay in the previous packet and if there were no packets in the acknowledgment buffer AB for a predetermined predetermined time period, the delay value dj of the current packet gets value, that is, df = djnff,.

When the switch returns from a congestion or overload condition, the delay control unit calculates a delay value dj using the formula: dj = adj., - (1-a) dM (2).

The second term is intended to steadily increase the delay in formula (1) when congestion is detected, and in formula (2) to steadily reduce the delay when the network returns from congestion.

·: 30 Figure 6a illustrates the solution of Figure 4a in a distributed buffer switch architecture. In the embodiment of Figure 6a, all packets in the shared buffer SB are buffered before each packet is routed to the switch's right output port OP ;. In another respect, the embodiment of Figure 6a corresponds to the embodiment of Figure 4a. Traffic junctions TS; (i = 1 ... n) may also form one unit which reads the packet at a time from the shared buffer and 11,104,602 delivers the packet to the correct port. The delay control unit DCU (not shown in Fig. 6a) may also be implemented as a unit common to all output ports.

In the embodiments of Figures 4a and 6a, the acknowledgment buffer includes multi-link packets and all packets are delayed according to the same delay algorithms-5 min. Alternatively, packets can be stored on each output port per connection, i.e., data packets for each IP connection (or TCP connection) can be stored in a separate buffer. In these cases, each buffer can be a FIFO-type buffer, since packets in a single queue do not have to be rearranged even if different connections are delayed differently. The relative proportion of each of the 10 connections in the uplink buffer can also be determined by load level measurement and the connections can be delayed based on the measured values. In this way, it is possible to delay more acknowledgments of connections that load more heavily on the network. Figure 6b illustrates this alternative embodiment, wherein the output port has a buffer unit 15 BFU that includes separate queues for at least some of the connections.

If dedicated buffers are not used and different connections are delayed differently, the buffers may be, for example, shift register type memories, which allow packet reordering so that packets of less busy connections can override packets of overloaded connections.

As mentioned earlier, the congestion control method of the invention can be utilized in packet networks. This means that there are user terminals in the network, network access points that are running the interface, and switches.

User terminals act as traffic sources, i.e., points that send and receive data. The switches may be packet switches or ATM switches. The access point may be, for example, a router or the access point may perform packet reassembly / reassembly, routing or switching. The delay of acknowledgment packets is preferably performed at the access points, but it can also be performed at the switches in the network, as later. "30 will be described.

Figures 7a and 7b show two different embodiments of the invention in an IP network. In the embodiment of Figure 7a, congestion detection and acknowledgment delays are performed at the access switch IPS1 providing the access to the IP network. In the embodiment of Figure 7b, congestion detection is performed at the subscriber node 35, while acknowledgment delay is implemented in the TCP / IP protocol stack of the UT. Congestion messages CS are transmitted to the user terminal 104 1042, where packets containing acknowledgments are delayed as described above before being transmitted to the TCP source.

Figures 8a and 8b show two different embodiments of the invention in connection with an ATM network. In the embodiment of Figure 8a, congestion detection and acknowledgment delay are performed at the access node AN. The access node can be divided into an interface card unit ICU and an ATM switch ASW. The interface card unit includes ATM Adaptation Layer (AAL) functions for cleaving and reassembling IP datagrams. Congestion is monitored in the node's ATM switch side, for example, by monitoring the fill rates of the buffers that push the subscriber's traffic to the network. Congestion messages are transmitted to the interface card unit, whereby the redesigned IP packets are delayed as described above. In the embodiment of Figure 8b, congestion is monitored on the ASW switch, acknowledgment packets are instead delayed on the TCP / IP protocol stack of the user terminal.

Embodiments of Figures 7a and 8a are more advantageous because it is much more economical to implement acknowledgment delay in a single access node than in a plurality of terminals located in users' premises. Furthermore, it is of course more advantageous that the user terminals do not have to be modified in any way when the invention is implemented.

As mentioned earlier, one network element on the connection path may command another network element on the same path to perform the delay. FIG. 9 illustrates this principle in a TCP / ATM network by illustrating a connection between two terminals (A and B) using TCP as a transport layer protocol. In addition to the user terminal access nodes 25 (ANS and AND), the figure shows only one intermediate ATM node (N1) and transmission lines connecting the nodes. Assume that network nodes have channels in two directions; the return channel and the return channel. To simplify the description, we assume that data packets are transmitted from terminal A to terminal B via access node ANS, one or more ATM switches and access node AND (downstream) and acknowledgments are returned from terminal B to terminal A via access node AND, one or more ATM switches and access ). As mentioned above, the access nodes can be divided into an interface card unit ICU and an ATM switch ASW. The interface card unit includes the ATM Adaptation Layer 35 (AAL) functions for cleaving and rebuilding IP datagrams. As in the example of Fig. 8a, acknowledgment delays are performed on the interface unit "10460210". However, in this case, the congestion is not monitored in the ATM switch section of the access node, but in the ATM switch further away in the ATM network. The ATM switch mentioned in Figure 9 which commands the access node to delay acknowledgments is Node N1.

The network of Figure 9 has ABR flow control between the transmitting extreme system (ANS) and the receiving extreme system (AND). With respect to the RM cell stream on this two-way connection, each endpoint is both a transmitting and a receiving extreme system. As shown in Fig. 9, for the flow of forward information from the access node ANS to the access node 10 AND, the monitoring loop consists of two RM cell streams, one outgoing and the other reverse. The access node ANS generates forward RM cells which are inverted by the access node AND which are sent back to the access node ANS as reverse RM cells. These reverse RM cells carry feedback information provided by network nodes and / or access node AND 15. A network node in an ATM network, such as node N1, can: - add feedback information directly to the RM cells as they pass through the node in the upstream or downstream direction, - inform the source indirectly from congestion by setting Explicit Forward Congestion Indication (EFCI) ) in headings that go in the direction of going. In this case, the access node AND updates the reverse RM cells according to this congestion information, - generate the reverse RM cells.

Thus, there are at least three different ways to control the acknowledgment delay at the access node 25 from the network.

In RM cells, the congestion information can be added, for example, to a 45 octets long field called "Function Specific Fields" or to a subsequent Reserved part which is 6 bits long. The traffic parameters transmitted to the ABR-capable user in RM cells are described in ITU-T Specification 1.371, paragraph 5.5.6.3, and the structure of the RM cell, in paragraph 7.1 of the same specification, from which an interested reader can find a more detailed description of RM cells.

The EFCI bit is the middle bit in the 3-bit PTI (Payload Type Indicator) field of the ATM cell header.

In accordance with this preferred embodiment of the invention, the corresponding access node 35 receives reverse RM cells containing congestion information when an overload or congestion is detected in an ATM network node. Based on this 14,104,602 information, the ATM switch portion of the access node adjusts its transmission rate in the direction of the ATM network and the flow control mechanism delays acknowledgments to the traffic source on the reverse channel. In this way, the TCP source automatically begins to lower its transmission rate, or at least it does not increase its transmission rate as fast as it otherwise would. As mentioned earlier, this is because the delay reduces the rate at which the source increases its congestion window size.

End-to-end ABR flow control can be performed as described above without changing the cooperating TCP protocol. In other words, the ATM and TCP flow control loops can be implemented in an economically advantageous way.

Figures 10 and 11 are timelines illustrating the interchange of segments between a TCP source and a TCP destination. The source is shown on the left and the subject on the right. There are 15 sending and receiving events marked with numbers starting with three.

FIG. 10 is an example of how a source and a destination behave in a conventional network, i.e., a network that does not employ the method of the invention on the reverse link path. Initially, the source is in the slow start phase. Let us assume that the network load gradually increases, resulting in the loss of packet P10 at the congested point of the network, which is transmitted at 21. After that, the source still sends packets because the acknowledgments it receives are in order. At 37, the source finally discovers that the received acknowledgment number was not in the correct order, and thus stops sending.

? 25 At number 41, the timer is triggered and the source performs retransmission of packet P10. At the same time, the source goes into the congestion phase.

Figure 11 illustrates an example of data exchange when a network uses the present invention. In this case, an overload is detected after the object has sent the seventh acknowledgment (ACK7). As a result, * 30 this acknowledgment and subsequent acknowledgments (ACK8 ... ACK11) are delayed on the network.

As can be seen in the figure, the source begins to decrease its transmission rate already at 24, continuing at the slow start phase. As shown in the figures, the traditional network behaves more unevenly; first 35 sources send a lot of packets and when the traffic is detected no packets are sent at all. The network using the invention, on the other hand, behaves much more steadily, since delaying the acknowledgments prevents the source from increasing its peak traffic as fast as in the known network. This allows the buffering capacity of the access network to be reduced.

Although the invention has been described above with reference to the examples in the accompanying figures, it is to be understood that the invention is not limited thereto, but can be modified in many ways within the scope of the appended claims. The following is a brief description of some variations.

As stated above, a condition for the user terminal is that it acknowledges correctly received (error-free) data units. As a result, the 10 ideas can in principle be used in conjunction with any other protocol that transmits acknowledgments and reduces its transmission rate if acknowledgments are delayed. The formula used to calculate the absolute value of the delay can also vary in many ways. The measurement unit can inform the load level in many ways; More than one bit may be used as ON / OFF type information or to indicate the value of the measured load. The signal (CS) which informs of the load level may also contain information about the connections affected by the delay of acknowledgments. User terminals may also have a wireless interface to the network.

«

• I

Claims (15)

104602 1. A method of monitoring congestion in a packet switched network comprising traffic sources (A), traffic destinations (B), and network nodes (AN, N1), comprising: 5. transmitting information units from a traffic source to a destination, - sending an acknowledgment from the source to the destination; - measuring the load level at at least one node of the network, characterized in that the acknowledgments per 10 sources are delayed if the measured load level exceeds a predetermined value. Method according to claim 1, characterized in that the acknowledgments are delayed at the same node of the network where the load level is measured. A method according to claim 1, characterized in that the acknowledgments are delayed at a different node of the network than the one in which the load level is measured. A method according to claim 3, characterized in that the acknowledgments are delayed in the access node (AN, ANS, AND), which provides access to traffic sources and destinations in the network, and the load level is measured in at least one node (N1) in the middle of the network. The method of claim 4, wherein the network between the access nodes is an ATM network, characterized in that - load level information is transported to the access node in RM cells, and the acknowledgments are delayed based on the information contained in the RM cells. A method according to claim 1, characterized in that the acknowledgments are delayed in at least one node of the network by: - storing at least a portion of data packets passing through the node in a first direction ·: * 30 to the first buffer, - reading the data packets from the first buffer such that containing an acknowledgment are transferred to another buffer, and (b) packets without acknowledgment are forwarded directly to the outgoing transport link (OL), - defining a delay value for each packet in the second buffer, and • · 17 104602 - reading the packet from the second buffer for the outgoing transport link. (OL) when the delay specified for said packet has elapsed. Method according to claim 6, characterized in that the first and second buffers are used at each output port in the first direction. The method of claim 6, characterized in that the delay value is determined using the same determination rule for all packets in the second buffer. Method according to claim 8, characterized in that the packet delay value is determined on the basis of the delay value of the preceding packet and the delay value measured over a certain preceding period. A method according to claim 1, characterized in that only acknowledgments belonging to selected connections are delayed if the measured load level exceeds a predetermined value. A packet switched communication network comprising: - nodes interconnected by transmission links (TL1, TL2), - user terminals (UTs) connected to nodes, which act as traffic sources that send data packets and traffic objects that receive data packets, and 20. measuring means (LMU) for measuring the current load level in the node, characterized in that the network further comprises: - delay means (AB, DCU) operatively coupled to the measuring means (LCU) for delaying data packets carrying from the destination to the source acknowledgments. The network according to claim 11, characterized in that one node of the network comprises both measuring elements and delaying elements. 13. The network of claim 12, wherein the at least one node is an access node that connects at least one user head. : 30 telecommunication network. The IP network according to claim 12, wherein the network nodes transmit IP packets, characterized in that said at least one node can be any one or more network nodes. The TCP / ATM network according to claim 11, characterized in that the delay elements are connected to the measuring elements by means of a RM cell current, which ... RM cells carry load level information. 104602