FI104670B - Packet routing in a data communication system - Google Patents

Description

104670

FIELD OF THE INVENTION The invention relates generally to packet transmission in a packet switched communication system, in particular to the transmission of IP packets (IP, Internet Protocol) over an ATM network.

BACKGROUND OF THE INVENTION 10 IP is the most popular of the current network layer (third layer of the OSI model) protocols, mainly due to the high popularity of the Internet. With the exponential increase in the number of host computers connected to the Internet, the performance of IP networks has become a bottleneck and new ways of transmitting IP traffic faster than ever are needed.

Figure 1 illustrates a typical structure of an IP network. In an office environment, PCs or similar terminals are connected to LANs LAN1 ... LAN3, which are typically Ethernet networks. The local area networks, in turn, are interconnected by a backbone network (WAN, Wide Area Network) with nodes (RT1 ... RT6). All computers that are on the same LAN have the same IP network address. When a data packet is sent from a PC connected to a LAN,

I> I

the IP stack of the hosts protocol stack checks if the IP destination address * t · is the same as its own IP network address. If the address is the same, no routers are needed, but the packet is sent over the LAN. for a computer with a destination address. If the destination's IP network address is different from the machine's own IP network address, the sending computer sends the packet to the router, which forwards the packet tl * · '' to the other network.

Transmission connections between routers can be implemented using PDH or SDH 9 9 9 - *; {· * technology or e.g. packet network technology (ATM, Frame Relay, X.25).

• · · '30 The router has two main functions: packet forwarding and updating routing tables. In principle, the packet relay process works so that the router first reads the destination's network address from the incoming IP packet. Thereafter: v. it retrieves the routing table in question. to the output port corresponding to the address and send the f * [• w ': packet in question. port to the next router. The packets are transmitted from the router f ·! 35 routers until one router discovers that the destination address is the same as its 104670 2 dedicated network address, whereby it sends the packet to the destination host machine.

As the transfer rate requirements increase, new technologies have been introduced. ATM technology is increasingly being used as backbone network technology because it allows high capacity backbone connections. The router nodes then have interfaces built into the ATM network; From the cells coming from the ATM network, the packets are first reconstructed, the packets are routed, and then the packets are re-chopped into ATM cells for transmission within the ATM network. The standard ATM adaptation layer (AAL, ATM Adaptation Layer) performs IP packet cleavage and reconstitution. This will be described in more detail below as a background to the following description of the invention.

When a workstation in the Ethemet LAN described above transmits data to a workstation in another LAN, the data application P1 formed by the workstation application is first encapsulated, e.g., in TCP packet P2 (assuming that the transport layer protocol is TCP, Transmission Control Protocol). The TCP packet is then encapsulated in the IP packet P3 and the IP packet is further encapsulated in the Ethernet frame P4, which is transmitted over the LAN to a router connected thereto, which also has an interface to the ATM network. This router strips off the Ethernet segment and splits the IP-20 packet into ATM cells on the ATM adapter layer. Note that encapsulation may include insertions both in front of and behind the packet (so-called trailer).

| I * # 'f. \ Figure 3 illustrates the structure of a single IP packet (or IP datagram) «· y \ 30. The minimum size of the packet header is 20 bytes divided into five% St * / four-byte "words", shown in the figure below. Each field name is followed by parentheses in the field in brackets. 4 «· *; !; * sa is the first 4-bit version field 31 to denote the current IP version of '' ', followed by a length field 32 (IHL, Internet Header Length), which indicates the length of the header in 4 bytes. i * · | · * quality and field including the 34 datagram length header.

'' J * 30 Authentication field 35 is used to identify the IP packet during packet reconstruction. The flags field 36 can be used to determine whether a split packet is present

IM

·; ··: The last piece of the original package or not. The Location field (Fragment

The offset) 37 content, in turn, tells you where the original ♦ · \ package belongs. The Lifetime field 38 indicates the longest time that a packet can be in a network-

I »I

* * 35 Sa. Each router through which the packet passes reduces the value of the field.

104670 3

The protocol field 39 indicates what is the upper-layer protocol followed by the data transmitted by the datagram (e.g., TCP). Field 40 contains the header checksum. Fields 41 and 42 are for source and destination addresses, that is, they represent the address of the sending and receiving host as 32-bit 5 addresses. The address fields are followed by an optional field 43, which is rarely used. The data being transported in this field is usually related to network testing or troubleshooting, for example, the data may define a particular path for the datagram to follow. If necessary, the field is completed with padding bits such that the number of bytes is divisible by four.

10 After the header described above, the actual data of the IP packet begins.

The length of the data field may vary, but its upper limit is limited by the length of the field 34, which limits the total packet length to 216 bytes.

As mentioned above, in the ATM network of Figure 1, IP packets are transmitted as ATM cells. Figure 4a shows the basic structure of a single cell transmitted in an ATM network. Each cell transmitted on the network consists of a payload of 48 bytes and a header of 5 bytes, but the exact structure of the header (header content) depends on what part of the ATM network is being used. Namely, the ATM network architecture comprises a set of interfaces that are precisely defined in the standards, and the header structure used in the ATM-20 depends on which interface (i.e. which part of the network) is involved.

<1 I

Figure 4b shows the structure of the cell header in the UNI boundary of the ATM network.

f I I

i '\ interface (User-to-Network Interface) which is the interface between the ATM terminal and the ATM-I · *' / node. Fig. 4c, in turn, shows the header structure at the ATM network (NNI), which has two

t · S

An interface between an ATM node, either within a network or between two different networks.

s:

• The cell header routing field consists of a Virtual Path Identifier (VPI) and a Virtual Channel Identifier (VCI)

= ·; * / (Virtual Channel Identifier). In the header structure of Fig. 4b, which • m · '· *' 30 is used only in the subscriber interface, a total of 24 bits are reserved for the routing field (VPIA / CI). In the header structure of Fig. 4c, which is used everywhere else in the ATM network, 28 bits are reserved for the routing field (VPIA / CI). As the name implies, the routing field serves as the basis for cell routing in the ATM network. Primarily, the inside of the network uses a virtual bus identifier 35 VPI, which in practice often determines which physical connection the cell needs to be routed. In contrast, the virtual channel identifier VCI is often used in routing only at the edge of the network. However, it should be noted that only the VPI and VCI together uniquely define the cell path.

Other fields specified in the specifications in the ATM cell header 5 are: - GFC (Generic Flow Control), a field for traffic control in a subscriber interface, not yet defined, - Payload Type Indicator (PTI), used mainly to distinguish network control cells from In addition, it is possible to differentiate according to whether or not congestion is detected on the way, - CLP (Cell Loss Priority), is used to prioritize cells over the likelihood of destruction (largely equivalent to the DE bit of the Frame Relay network), the checksum.

Of these other fields, the present invention mainly relates to the PTI-15 field, which can be used to monitor the inter-packet boundary. The last bit of the PTI field (bit number two in Figs. 4a ... 4c) indicates when a new upper layer packet (IP packet) will start. When the bit is set to number one, it is the last cell in the IP packet, whereupon the next packet starts from the next non-idle cell.

The ATM adaptation layer usually receives signals in different formats, and the function of the ATM adaptation layer is, on the one hand, to generate these signals in the standard format required by the ATM network before transmitting them to the ATM network and to reconstruct these signals from cells coming from ATM networks before as the signals are forwarded to user or control interfaces. Adaptation layers have been standardized for different types (AAL1 ... AAL5) for different service classes (A ... D). For example, AAL types 3, 4, and 5 provide migration services: for applications where there is no time dependence between source and destination.

Figure 5 illustrates AAL 5's IP packet splitting: to ATM cells and packet reconstruction from ATM cells as it

30 illustrates the operation of the ATM adaptation layer in the exemplary network of FIG. The ATM adaptation layer is generally divided into two sublays called SARs (Segmentation And Reassembly sublayer) and CS

«· (Convergence Sublayer). The latter performs encapsulation / deencapsulation of user data units «· *: V (e.g. IP packets) and control data * ·. '·: 35 (encapsulation / deencapsulation). The frame resulting from encapsulation by the CS sublayer 5104670 is called a Convergence Sublayer Protocol Data Unit (CS-PDU). AAL5 encapsulation is accomplished by adding a trailer section containing e.g. error checking part (CRC). The trailer is 8 bytes long. The entire length of the CS-PDU 5 corresponds to a multiple of 48 bytes, which is achieved by adding, if necessary, a fill pad PAD of 0 to 47 bytes between the trailer part and the payload part of the packet.

The SAR sublayer segments in the transmission direction each segment of the CS-PDU into 48-byte segments called SAR-PDU 10 (Segmentation And Reassembly Protocol Data Unit). In the receiving direction, CS-PDUs are formed by connecting the SAR-PDUs together.

The ATM layer below the ATM adaptation layer is responsible for adding the five byte header fields CH (Figs. 4b and 4c) to the SAR PDUs being transmitted to form ATM cells 50 which are transmitted to the ATM network. The ATM-15 network only handles the cell header, the 48 bytes payload part is neither processed nor even read on the ATM network. In the receiving direction, the ATM layer removes the headers from the cells and feeds the 48 byte payload components to the SAR sublayer for interconnection.

When the routers transmit IP packets in the network of Figure 1, they segment the packets into cells as described above and transmit the packets to the ATM connection. The router at the opposite end of the transmission (link) reconstructs the packet from the cells as described above, performs the routing head: .i: routinely based on the IP address, and re-segmentes the packet into cells for the next ATM link.

25 Normally, the packet routing decision is made by the software. When t ·.:. * · Software-implemented routing is combined with the packet segmentation and reconstruction described above, the ATM-based router network slows down considerably. This conventional routing head is also expensive to implement.

- In order to alleviate these drawbacks, a method has been developed called: T: 30 called IP switching. IP switching is based on a flowchart: a stream is a series of - IP packets that are (usually) traveling from the same source to the same destination. Thus, one stream includes (usually) those IP packets that have the same source and destination addresses. For example, a TCP connection is a flow: when the TCP connection I »»: V is opened, a series of packets are transmitted from source to destination. In the IP connection, V ·: 35 router nodes detect a leak and request routers on the edge of the network to provide packets for each flow with a unique leak identifier, e.g., a unique VPI / VCI identifier. When the packets belonging to the flow are provided with a unique VPI / VCI identifier, routers in the middle of the network can perform packet transmission at the cellular level using a standard ATM connection. Thus, the routing does not need to go up to the third layer (IP layer), but the routing can be done with the second layer (ATM layer).

However, the disadvantage of the latter known solution is that it requires a separate control protocol by which the nodes in the middle of the network negotiate with the incoming traffic nodes to assign a dedicated 10 VPIA / CI identifier to a particular stream. Such a solution makes the network more complex and generates additional traffic to the network, which overloads the network.

SUMMARY OF THE INVENTION The object of the invention is to overcome the drawbacks described above and to provide a solution that speeds up packet routing without requiring the introduction of any new flow control protocol.

This object is achieved by the solution defined in the independent claim.

The idea of the invention is to observe the packet butter (plural partitive word) in a node (which receives upper protocol layer packets) at the edge of the network. When a particular: .i .: Flow meets the criteria for a dedicated connection is given. For stream t · '·, * ·: a dedicated lower layer protocol connection identifier. When the next node in the traffic direction receives a data unit having a "new" connection identifier 9, it immediately knows that there is a flow assigned to the dedicated channel.

For the first packet, the node performs normal routing based on the IP destination address, but all of the following data units having the same: access identifier are switched forward based solely on the access identifier of the lower protocol layer. The nodes also have a timeout, which deletes. use of dedicated connections after a certain predetermined time • ·. no time has been received for this time. traffic covered by a goat.

«·. In a more preferred embodiment of the invention, the lower layer data units are ATM cells and the upper layer packets are IP packets, but the method may also be implemented in conjunction with other protocols or transport methods.

The solution according to the invention makes it possible to increase the performance of nodes by eliminating the need for segmentation and reconstruction of packets in nodes in the middle of the network. This is further achieved by eliminating the need for an additional flow control protocol in the network 5, whereby the router nodes also remain simpler and do not cause additional network traffic.

List of figures

In the following, the invention and preferred embodiments thereof will be described in more detail with reference to Figures 6 to 11 g in the accompanying drawings, in which Figure 1 illustrates the environment in which the invention is used; Figure 2 illustrates packet formation before being transmitted to the network; Figures 4a-4c illustrate the general structure of an ATM cell, Figure 5 illustrates packet cleavage by ATM cell 5 into ATM cells and packet reconstruction from ATM cells, Figure 6 illustrates the operation of the method of the invention in three consecutive network nodes 7, Fig. 8 is a flow diagram illustrating the operation of the method of the invention in the second and third nodes of Fig. 6, Fig. 9 illustrates the structure of a gateway node, * Fig. 10 hava Inspire the structure of the node in the middle of the network, and · Figures 11a. 11g show tables used in the node of Fig. 10.

DETAILED DESCRIPTION OF THE INVENTION In the solution of the invention, segmentation of IP packets is performed in a node on the edge of the network (either node RT1 or RT4 in Fig. 1) as described above and ATM cells are transmitted to the ATM. -siirtoyhteydelle.

6 shows three nodes N 1 to N 3 of the ATM network, which are V'i 35 in succession on the packet path and which implement the method according to the invention. Default inter-node channels are indicated in the figure by thicker tubes and inter-node dedicated channels by thinner tubes. The various steps of the method are indicated by circled numbers.

Initially, the ice system has only default channels and when traffic starts, dedicated channels are allocated to the connections as needed. Default channels are sent from the node in all directions and one default channel can carry packets with different source and / or destination addresses (even though they share the same VPI / VCI identifier). Thus, the default channel is a channel that uses ATM only as a bearer and where routing is performed at a higher level (IP-10 level) as normal. For this reason, traffic on default channels is unwanted traffic that is being tried to get rid of. However, the default channels have to be used because, in practice, a very large part of all traffic is so short-lived (perhaps only a few packets are sent over TCP) that it is not worth giving them their own VPI / VCI. Only for the part of the traffic that is longer-term it is worth creating a dedicated virtual connection.

The first node (N1) only receives IP packets, which it splits into cells as described above and forwards the cells to N2.

Node N1 examines the desired header fields (source and destination addresses and possibly other fields) of incoming IP packets in a manner known per se.

20 Upon detecting (circled number one) that a particular IP connection (those packets having the same source and / or destination addresses) fulfills the criteria for a second layer (OSI layer 2) dedicated to certain predefined criteria '·· ’: .v node N1 decides to provide a second-level connection dedicated to that stream • * V '! (also known as a connection identifier), which is a new VPI / VCI identifier for an ATM network.

25 The node may make the above decision, for example, as a result of its attention to «V:; country Having regular or heavy traffic on the IP connection (node v: counting incoming packets) or discovering that the IP connection needs certain quality requirements. For example, the node may notice that it is an FTP transfer that: requires fast service. The above criteria can be of many types.

30 When node N1 detects the first packet after the above decision, it selects the first packet. the packet's IP header, a free second-tier connection identifier (VPI / VCI).

For example, upon discovering for the tenth packet that the connection needs: V a dedicated second-level connection, the Node N1 begins at that. from the package to use

Vi 35 identifiers belonging to a dedicated second level connection in 104670 g bones (circled two). The first nine packets have gone forward (in chopped cells) on the default channel. The cells of each packet are transmitted sequentially on the default channel so that the different packets do not overlap.

5 Next, the next node (N2) in the traffic direction knows, after receiving a cell having a "new" (not one-flow) VPI / VCI identifier, that it is a flow that has a dedicated second level connection identifier in the uplink node. (circled third). From this cell, the node reads the IP destination address 10, which is mapped, as seen in Figure 3, in a segmentation such that it is always in a fixed location in bytes 13-17 of the cell payload. Based on the read destination address, the node searches the routing table for the destination. identifies the output port corresponding to the address and forwards the cell to that output port. All cells having the same VPI / VCI identifier on the same output port are then transmitted. Sol-15 only needs to look at the first IP address by which it performs normal routing (IP-level). It can then perform the switching at the ATM level directly based on the VPI / VCI identifiers and pass the cells again with the new VPI / VCI identifier to the next node (Nelonen circled).

Fig. 7 is a flowchart illustrating the operation of 20 nodes (gateway node N1) at the edge of the network. In the figure the reference marks Tn (n = 1 to 5) are shown in brackets at various stages, which refer to the tables used in the node • * '·' 'shown in Figure 9 below.

Initially, the node receives the IP packet, reads at least V * i source and destination addresses from its header (step 71), and updates the packet counter corresponding to the address pair (or destination address only) (step 72, Table T1).

The node then examines whether the · · f v: dedicated connection is set for that address pair (step 73). This is done by looking at the table to see if a dedicated channel corresponding to this address pair is set on (table T1). A dedicated second level connection identifier assigned to the flow. If so, a routing tag corresponding to the connection is retrieved from the routing table (Table T2), which is the internal identifier of the ATM switch • field, by which the ATM switch field performs the switch to the correct output port (step 74).

: V If a dedicated connection has not yet been provided, let us examine next: V ·: 35 whether the dedicated connection criteria are met. for the flow (step 75), i.e. whether the 10 104670 e.g. traffic has exceeded a certain value during the measurement period or are the traffic quality requirements of the flow (e.g. requiring high throughput probability). The node may utilize, for example, the connection type field 33 (Figure 3), which indicates the quality of service in terms of delay, throughput 5, and reliability. If the criteria are not met, a routing identifier for the default channel corresponding to the IP destination address in question is retrieved (step 74, table T2) and routing is performed accordingly.

If it is found in step 75 that the dedicated connection criteria are met, an output port 10 (Table T2) corresponding to the address pair or destination address in question and a free routing identifier corresponding to the output port (Table T3), i.e. a routing identifier corresponding to the IP address, are not retrieved. The dedicated channel corresponding to the IP address is then switched on (Table T1) and the routing identifier is updated. for the connection (Table T2). At the output port, 15 free VPIs / VCIs (Table T5) are retrieved per new routing identifier and updated to the routing identifier conversion table (Table T4).

Next, it is possible to proceed to step 78, where the IP packet is split into cells and the routing identifier as previously obtained is appended to the cells. The cells are then forwarded to the output port 20 corresponding to the routing identifier (step 79). Before transmitting cells to the next link, the routing identifier is replaced by its corresponding VPI / VCI identifier (Table T4).

* · '·' Then we proceed to receive the next IP packet.

v .. * In addition, the node has a separate aging logic that disables »· V'i dedicated connections and their associated contacts after .. '. Γ 25 incoming calls have not been received for a predetermined period of time.

A:.: Traffic covered by a goat. It is preferable for all nodes to use a similar mechanism, · · · v, whereby the dedicated connection is removed from the source node. If the node has had time to remove the dedicated connection and traffic flows, the routing decision will be made as described above: 'V: 30 for the' first 'packet.

. · * ·. Figure 8 illustrates the operation of a node in the middle of the network (e.g., N2 and N3). In the figure, in various steps, the brackets • * refer to Sn (n = 1 to 7), which refer to the tables used in the node, · · ·: which are shown below in Figs. 10 and 11a to 11g.

V ·: 35 Initially, the node receives the cell (step 80) and reads its header (step 104670! 11> 81). The traffic counter corresponding to the VPI / VCI identifier is then updated (Step 82, Table S1). Next, the node examines whether it is the default channel, i.e. whether the inbound VPI / VCI identifier is marked as the default channel (step 83, your table).

5 If it is not the default channel, the next step is to determine whether the stream with the dedicated channel is on (step 84, table S1), i.e. whether the inbound VPI / VCI is marked as a dedicated channel. If this is the case, the routing identifier corresponding to the incoming VPI / VCI identifier (step 86) is applied directly from Table S3.

If, on the other hand, a dedicated channel is not yet assigned to the Flow, proceed to step 88 to read from the cell an IP destination address that is mapped, as shown in Figure 3, in a segmentation such that it is always fixed in bytes 13-17 of the cell payload. Based on the read destination address, the corresponding output port (Table S4) and the corresponding 15 free routing identifiers (Table S5) are retrieved from the routing table. The output port looks for a free VPI / VCI identifier for the routing identifier (Table S7) and places it in the output port routing identifier conversion table (Table S6). At the same time, the connection is marked as dedicated and the data is updated in tables (Tables S1, S3, S4).

If a default channel is detected in step 83, the traffic counter corresponding to the IP address pair or destination address is updated (step 85, table S2) and the value of the dedicated connection (step 87) is examined based on the counter value. flow. If the criteria for a dedicated connection are not fulfilled, the cell is routed normally, i.e., it is searched for the IP destination address «· '* · *: the corresponding path (output port) and the corresponding routing identifier (step 89, Table S4). If, on the other hand, the criteria for a dedicated connection are met, then proceed as in setting up a new dedicated channel (step 88).

* ** v: Following the steps described above, the node has found the correct routing identifier and the cell can be routed to its correct output port (step 90). Before transmitting v v: cells, the routing identifier is removed from the following link and the header · *:: 30 is provided with an outgoing VPI / VCI identifier (Table S6).

, ···. As shown in the flowchart above, it is also possible to set up dedicated channels in a node in the middle of the network, e.g. based on the destination address, since traffic can be multiplexed from multiple sources or • · · =; · 'Parallel links to the same destination. If dedicated channels are multiplexed, it must do so while maintaining the order of the packets.

12 104670

Figure 9 illustrates a possible implementation of a gateway node as a functional block diagram.

For temporary storage of incoming packets, the node input buffer has a buffer 91. The measuring and monitoring block 92 reads the packet header and 5 uses a source / destination pair to retrieve from its table (T1) whether the dedicated channel parameter is set on.

The search block 93 obtains an IP address from the measurement and monitoring block and, based thereon, obtains a routing tag from the routing tables (T2), which the search block enters into a segmentation block 95 for segmenting cells and assigning a routing tag to the segmented cells.

In addition, there are accounting units 98 on the input side of the node that keep track of the free routing identifiers corresponding to each input port and the output port (Table T3).

With the routing identifier, the cell is fed from the segmentation unit ATM-15 to the switch 96 which switches the cell based on the cell identifier to the correct output port 97. Thus, the routing identifier is an internal identifier of the switch not transmitted on the network.

The measurement and monitoring block measures traffic for each stream and removes dedicated channel parameters, such as routing tag and VPI / VCI, when it has not detected that particular channel for 20 periods. traffic covered by a goat.

Figure 10 shows the principal structure of a node in the middle of the network. The node blocks are the same as above, only the node behavior · 'which tables differ from the gateway node. Therefore, the same reference numerals are used for the different V · blocks as for the corresponding blocks in Figure 9.

Tables S1 ... S7 used in the different blocks are shown separately in Figures 11a ... 11g.

• ·· V: Measurement and monitoring block 92 holds a counter for each VPI / VCI pair) and indicates whether or not the VPI / VCI pair is connected. default channel or dedicated channel (Table S1, Figure 11a). In addition, in order to transfer the default channel traffic to a dedicated channel: ded: T: 30, it maintains an IP address pair or destination address, ···, counter (Table S2, Figure 11b). In the middle of the network, the source / destination pair traffic may · · be multiplexed by, for example, parallel links or multiple sources • • 1 2 multiplexed traffic to a specific destination address for which a dedicated channel is formed.

The routing identifiers corresponding to the VPI / VCI for dedicated channels are 13,104,670 in their own table (S3, Fig. 11c) in use by the routing block. In addition, for all connections, there is a routing table with an IP address corresponding to the output port and routing identifier (Table S4, Fig. 11d). This table is required for all default channel routed packets and when creating a new dedicated channel.

The accounting unit 98 keeps records of the free routing identifiers corresponding to each input port and output port (Table S5, Figure 11e).

The ATM output port has a conversion table in which the incoming cell routing identifier is replaced by the outgoing cell VP1 / VCI (Table S6, Fig. 11f). In addition, the output port must keep track of free VPI / VCI identifiers (Table 10 S7, Figure 11 g).

While the invention has been described above with reference to the examples in the accompanying drawings, it is clear that the invention is not limited thereto but can be modified within the scope of the inventive idea set forth in the appended claims. For example, the flow may comprise, as described above, traffic between the same source / destination pair or traffic to the same destination.

«« • · <· · I «i« '(»· f · # I · m» i • · • · S1 · • ·

IM • M

Jl ·

• I

«• I l I I I • I · I a f · · • II ··· I · I a f · · • I · f I>>> f f f

Claims (5)

A method of transmitting packets in a packet switching data communication network in which the packets are transmitted from one of the network nodes to another on the base address of the packets, according to which method 5. the data packets (30) to be conveyed in a node (N1) in the edge of the network is segmented into data units (50) according to a lower protocol layer, - the data units are conveyed both across nominal channels and dedicated channels in the network, whereby a single nominal channel transmits data units belonging to packets with different message addresses and over a single dedicated channel. data packets belonging to the same flow, characterized by - in the node (N1) at the edge of the network, packets belonging to the same flow are taken into account and, to the availability of a flow, when it is found that it meets predetermined criteria, a dedicated connection identifier is stored by a lower protocol layer, - in a node (N2) in the middle of the network, pay attention to the connection of the incoming data packets e-identifier and, for a data unit, when a new identifier belonging to neither a nominal channel nor a dedicated channel is labeled, on the basis of the destination address of the packet, a new port and a new output connection identifier are determined by the lower protocol layer and the data: ':'; the output is controlled to the relevant output, - the vi determination in the node (N2), where in the data units that come in later, the same connection identifier is labeled by the lower protocol layer as said new connection identifier, is performed only on the basis of the connection identifier of the lower protocol layer, and In the nodes, dedicated connections are discontinued after, for a period of a certain length of time, no traffic has been marked in connection with a flow with dedicated connection identifiers. • · Method according to claim 1, characterized in that the same time control is used in the notes. Method according to claim 1, characterized in that in the node. (N2) in the middle of the network is also taken into account traffic over the nominal channels and that the dedicated connection identifiers are distributed to the lower 11 «: the protocol layer where it is observed that a certain flow on a nominal channel meets the« predetermined criteria. 17 104670 Method according to claim 1, characterized in that, as said predetermined criteria, the amount of traffic received during a certain measurement period is used. Method according to claim 1, characterized in that, as said predetermined criteria, the information contained in the data packets about the service quality they require is used. • ♦ · • · · II i • «· • · * * • · ·» t · m • • • »i« • · · · · · · · · · · · · · · · * • · • ♦ · • · · • • • ♦ ♦ • • · · ♦ ♦ · • · · · · · · · · · · · · · · · · · · t · ♦ ·, • · · • »· t * · • *» · · · ·