This invention relates to an improved method of operation of telecommunication systems comprising at least one network where data in appropriate form is sent between nodes via links. It has particular but not exclusive application to LAN's, WLAN's such as Ethernet

In some networks such as Ethernet, all the switches are transparent, network data are passed the destination element of a frame doesn't realise that the frame has passed through a number of switches. If an Ethernet network contains loops there is more than one way from one network element to another network element the transparency is lost since a frame may be duplicated i.e. has been sent of down multiple paths to the same destination, forking of from a network element. This problem is further exacerbated by the fact that other switches don't know that a frame is already duplicated and may further duplicate it themselves, causing what is known as a broadcast storm.

This is demonstrated in figure 1 where work station 1 send frame to workstation 2 and two frames would arrive since switch 3 would broadcast to all the port, except the one it received it from, since a still hasn't learned the location of the workstation 2. Two frames would then reach C which would then become confused since it would learn two possible locations for the origin of workstation 1 (it would only keep one of them thinking the location of 1 had changed. Worse than this, C would broadcast the frames received from port 2 to ports 1 and 3 and the frame it received from port 1 to ports 2 and 3. Two frames would reach workstation. Although this is obviously undesirable as it involves unnecessary duplication and thus unnecessary taking up of bandwidth it is usual to have more than one physical path from one network to another in order to have fault tolerance so that if a link is down data can still be passed between networks.

One known method of dealing with the problem so as to redeem the dilemma is to set up the system such that although there may be multiple

physical link there is only one logical link through which data can reach one particular unit in a network to another. In spanning tree protocol (STP), the switches exchange messages between themselves in order to guarantee that only one path effectively exists between one node and another. Based on these messaged the node will then block all the traffic from specific ports in order to avoid loops (but surely you mean switches). Such a system is shown in figure 2; by using STP switch A decided to block all traffic going through port 2. This effectively cuts the loop and guarantees for this network that only one paths exists from one node another.

In Ethernet networks the same problem exists. Where ever a frame does not know where to deliver a frame because it does not know the address it has to broadcast it to all the ports except the one where it received it from.

This problem has been partially solved in such system using the concept of Virtual Local Area Networks. In such systems an internal table is kept containing the VLANs and the ports where these VLANS are assigned to. In such a way a network element is able to broadcast the frames belonging to a given VLAN just to the ports where that VLAN is assigned to in order to support the concept of VLANs, the Ethernet frames had to be revises in order to include a VLAN tag which assists a receiving unit in knowing through which port (s) to send it further in through which it should not,. In this way whenever it receives a frame it knows which VLAN it belongs to by examining it and extracting the VLAN tag. This is shown in figure 3.

The number besides the ports represent the VLANS that are assigned to them. This means that only frame that belong to that VLAN are allowed to pass. All others are discarded. When a workstation belonging to VLAN 2 sends frames it marks them with VLAN tag 22 when a switch receives frames from VLAN 2 it broadcasts them to ports that have the VLAN 2 membership. A port may well have several VLAN memberships, in order to ensure connectivity between workstations in different switches but belonging to the same VLAN. Although this allow the network to be set up in an

efficient manner which multiple but at the same time restrict paths (loops) it is inflexible when there is variability in link usage.

Another similar way of reducing the problem is the use of the Multiple Spanning Tree Protocol (MSTP) tries to solve the problem of unbalanced Ethernet networks when using normal STP. This solution is described with reference to figures 4 and 5. This shows a typical Ethernet network where STP and VLAN's are used. Suppose that the VLAN range 1 to 1000 belongs to customer A and that VLAN range 1001 - 2000 belongs to customer B. The STP would non allow traffic to go through one of the links so as to avoid loops; it would do so fro all VLANS. One link would thus have no data traffic whatsoever. In order to avoid this, the efficiency is enhanced by Multiple Spanning Tree Instances which are entities that contain a VLAN range and a STO tree. This way its possible for the MSTP to have different STP tress for different VLANS thus balancing network traffic, this is illustrated in figure 4 b where there is one spanning tree for the VKL An range 1 -n 1000 and a different spanning tree for VLAN range 1001-2000 This way the network is much more balanced than the one using STP since all the links are being used. In this way the usage of the links is divided out between LANS, customers etc so one link doesn't get too overloaded but all links are blocked. In summary there fore SPT RSTP and MSTP are protocols which are used to avoid loops in the active network. These IEEE protocols are used and applied to physically meshed (wholly or partially) networks.

A problem is that if a node blocks data on a link, the adjacent node still takes up processing power and link bandwidth by sending data to the blocked port.

It is an object of the invention to overcome this problem.

The invention comprises in a communication system having a first and a second node where data is passed between said nodes via a link, a method of controlling data flow between the links comprising said second node

sending a request to said first node to block its proximal port so as to prevent said first node sending data to said second node.

Figure 6 shows the prior art system showing two nodes 1 and 2 with link comprising sub-links: a b c and d. These sub-links in the example correspond to data pertaining to The system has decided to block link (d). However the sending node 1 continues to send data (e.g. broadcasts and unknown unicasts) and does not know that the traffic is blocked on the other end of the link. All other links (which may be for particular are sending and receiving data packet. In other words all others VLANs are sending and receiving data packets. On node 2 the port is blocked just for just for VLAN (d), so that just packets received with green VLAN tag are dropped (discarded without any process). Traffic marked with different VLAN tag (a, b, c) continues to be processed normally.

Figure 7 shows an example of the invention to contrast to the above. As a result of too high bandwidth usage, node 2 requests to node 1 to block the sending port on node 1 for VLAN (d). In 7 b the node 1 accepts the request and does not send any packet from VLAN (d) neither does it process any packet that it receives with a VLAN (d) tag.

A request is made when the link bandwidth usage is above a threshold.

In alternative embodiments it may be if the bandwidth for a sub-link (i.e. sub- link (e) is above a certain threshold. The invention is not limited to these two examples and the request for both ports to be blocked may be when there is a certain relationship with between a certain sub-link (VLAN (e)) usage and total and link bandwidth usage, e.g. proportion. Or it may set up to be automatically so configured.

Figure 8 shows a flow diagram of the process in an embodiment of the general method.