GB2342825A - Controlling traffic flow in a local area network partitioned into virtual networks - Google Patents

Abstract

In a method of operating a local area network 1 which is composed of end stations 11,12, 21-24 and switches 1-7, the switches each having a multiplicity of ports, each end station being connected to a port and at least some of the ports of each switch being connected to a port of another switch, means is provided for adjustably limiting the traffic flow capacity of at least those ports (of switches 2,5,6,7) which are connected to end stations. The method comprises identifying a plurality of virtual local area networks 11,12 and 21-24 within the local area network; specifying for each virtual local area network a traffic flow limit; causing at least one switch to store an indication of the traffic flow limit for each virtual local area network identified in the local area network; monitoring in said at least one switch the traffic flow and detecting whether the traffic flow (bandwidth) for any specified virtual local area network exceeds the respective limit; and adjusting the traffic flow capacity of at least one port connected to an end station which is a member of the respective virtual local area network whereby to reduce through the said at least one switch the traffic flow attributable to the said respective virtual local area network.

Description

METHOD OF CONTROLLING TRAFFIC FLOW IN A LOCAL AREA NETWORK PARTITIONED INTO VIRTUAL NETWORKS Field of the Invention This invention relates to packet-switched data communication networks of the kind known as local area networks, particularly those operating according to an'Ethernet'protocol. It particularly concerns the avoidance of unnecessary congestion caused by excessive traffic flow between some of the members (i. e. end stations) in a network.

Background to the Invention Local area networks (LANs) are in ever increasing use for enabling users having access to'end stations'to communicate with each other selectively, to share information from a common store or to receive broadcast messages originated inside or outside the network, and for a variety of other purposes. For several reasons, for example to avoid excessive occupation of the available'bandwidth' (information carrying capacity) of the network by administrative multicasts or other messages which may be sent to at least a large proportion of the stations on the network, it has become customary to partition a local area network under software control into a plurality of'virtual'local area networks (VLANs). In general, these virtual local area networks correspond to a group of end stations or devices and may be identified by an appropriate field in the header part of packets originating from or directed to a station in the virtual network.

The use of virtual local area networks has substantially alleviated the problem of congestion arising from unnecessary traffic flow arising from multicast messages and other causes which need not be described here.

Even so, it will be obvious that any switch or other device which provides interconnection of VLANs the area network may experience congestion arising from the fact that it must physically handle traffic (packet-based data) arising from more than one virtual local area network.

Summary of the Invention It is possible to adjust the'bandwidth'i. e. traffic flow limit on a single port of a switch to less than the maximum or ordinary'bandwidth'which can be handled by the switch. A system for so doing, including a counter configure as a'leaky bucket'is disclosed in our prior copending application No. 9807264.8 filed 3 April 1998. However, the present invention concerns an automated extension of that idea, so that a network administrator can automatically configure ports in the network so as to monitor the usage of the physical network by the VLANs into which it is partitioned and to exert limitation of the traffic flow rate through at least one port which conveys traffic for a VLAN of which the usage of the network exceeds a specified threshold. Broadly, a user or administrator may specify a traffic flow limit (in practice some specified value in megabits per second) for each VLAN of the network. Such a value may be communicated to each device, i. e. switch, in the network Each device or switch can maintain, by examination of packets passing through it or samples of those packets the utilization of the information carrying capacity. Each device will need to identify which ports are connected to end stations. If a specified threshold for a particular VLAN is exceeded, then the limitation of the allocated bandwidth to that VLAN by at least one of those ports may be made operative. Various ways in which this may be achieved are discussed in detail later.

Brief Description of the Draina The accompanying drawing illustrates a local area network partitioned into two virtual local area networks.

Detailed Description The drawing illustrates in much simplified form a local area network which, apart from the interconnecting transmission medium, is composed of end stations 11-12 and 21-24 and switches 2-7. The term'switch'is used herein to denote a network device which has a multiplicity of ports and is capable of conveying signals received at any one of the ports at any selected one of its other ports. As will be appreciated by those skilled in the art, network devices have generally been divided into'bridges', and'routers'but the term switch is intended to embrace both of these as well as hybrid devices capable of performing both bridging and routing. It is presumed that the local area network is a packet-switched network in which information is contained by means of packets in a predetermined format, such as one of the specified Ethernet formats, the packets each including a header which has fields identifying source and destination addresses and, among other things, the number of a VLAN which may be defined within the local area network. In the IEEE 802.1 Q standard, an Ethernet packet includes a 12-bit field for this purpose.

The drawing illustrates two groups of end stations. The first group consists of the end station 11 which is coupled to one ingress/egress port of the switch 2 and end station 12 which is connected to a port in switch 5. A second group consists of end station 21 connected to a port in switch 7, end stations 22 and 23 which are connected to respective ports in switch 5 and end station 24 which is connected to a port in switch 6.

Those may be seen, switches 2 and 7 are connected to respective ports in switch 3 which is connected to switch 4 which is itself connected by a respective port to switches 5 and 6.

The first group of end stations and the relevant ports is defined as a first VLAN within the network whereas the second group of stations 21 to 24 is defined as the second VLAN within the network. Such definition may be made by known techniques by a network administrator.

As is known, the general purpose of partitioning the network into virtual local area networks is to group together those end stations, i. sources and destination of data-packets, which communicate with each other frequently and in particular generate multicast traffic within the group but which communicate with members of other groups less frequently.

The main objection of the invention is to provide limitation of the allocation of'bandwidth', i. e. traffic flow capacity to the various virtual local area networks so as to minimise the danger of excessive use or utilization of the traffic flow capacity of the switches, particularly the switches that must handle traffic from more than one VLAN in the network.

A first step in the administration of the network for this purpose is to provide switches which at least the ports which are to be connected to end stations have some adjustable bandwidth control by means of which the maximum traffic flow rate (i. e. a specified number of bits per unit time) may be controlled. Such devices are known from for example the above-mentioned co-pending application ; it is not intended to limit the invention to any particular device for bandwidth limitation.

Further, each of the switches 2 to 6 requires, either in hardware or software or a combination of the two, means for monitoring the traffic passing through it so as to gauge the current bandwidth occupied by traffic associated with each VLAN. Again, methods of monitoring such traffic are known and it is not desired to constrain the invention to any particular type of monitoring, the device may monitor packets to determine from the VLAN numbers of the packets the traffic flows over some suitable period of measurement.

Further, each device needs to be able to identify which ports are connected to end stations; such identification is an ordinary facility in currently available switches.

One method of administering a network of the kind described is to allot some predetermined or maximum desired bandwidth for each VLAN defined in the network. For example, the first VLAN composed of end stations 11 and 12 may be allocated a 50 Mbps and the second VLAN composed of stations 21 to 24 may be allocated 30 Mbps. Each device (2-7) in the network will be caused to learn these values via such a protocol as GARP. Each of the devices will monitor the VLAN usage as indicated and will therefore detect when in any particular device the bandwidth used by any VLAN exceeds the prescribed limit.

Each port which is connected to an end station is allocated a specified bandwidth or traffic flow capacity which may but need not be directly related to the bandwidth allocated to the VLAN It should be mentioned that in general a port may belong to more than one VLAN so that the actual bandwidth setting may be a combination of settings each one specified in accordance with the bandwidth allocated to the associated VLAN.

One method of operation comprises, when the bandwidth utilized by a particular VLAN in one of the devices exceeds a threshold, adjustment of the ports in the edge devices for that VLAN so as to decrement the allocated bandwidth associated with that port. For example, if in the specific example the bandwidth utilization for the first VLAN is detected by switch 5 to exceed the allocation, the ports in switch 5 connected to end stations 11 and 12 may have their specific bandwidth allocations decremented. Similarly, if the utilization of bandwidth by the respective VLAN drops below some recovery threshold (which may or may not be the same as the threshold previously mentioned) the bandwidth allocations for the ports in the edge device (switch) 5 may be incremented back up to the originally specified setting.

Various schemes may be employed to adjust the ports connected to end stations in order to achieve the general bandwidth control policy specified for the whole network. As mentioned above, one scheme is for every end station port which is a member of a controlled VLAN to have its bandwidth allocation decremented similarly. However, if the monitoring of the packets for the purpose of measuring bandwidth utilization by the individual VLANs includes a detection of the relevant end station port, then other schemes are possible. In particular, by establishing the traffic flow usage in terms of the individual ports, it is possible then to control the bandwidth allocated to that particular port only or even to decrement the relevant ports in proportion to their contribution to the VLAN traffic.

The scheme described above may be extended to the operation of a switch (such as switch 4) which is not directly connected to end stations. It may also employ a scheme, such as that described in our co-pending application No. 9812065.2 filed 5 June 1998, to reduce the bandwidth occupied by a specific VLAN.

Claims (3)

1. CLAIMS 1. A method of operating a local area network which is composed of end stations and switches, the switches each having a multiplicity of ports, each end station being connected to a port and at least some of the ports of each switch being connected to a port of another switch, means being provided for adjustably limiting the traffic flow capacity of at least those ports which are connected to end stations, the method comprising : identifying a plurality of virtual local area networks within the local area network ; specifying for each virtual local area network a traffic flow limit ; causing at least one switch to store an indication of the traffic flow limit for each virtual local area network identified in the local area network ; monitoring in said at least one switch the traffic flow and detecting whether the traffic flow for any specified virtual local area network exceeds the respective limit ; and adjusting the traffic flow capacity of at least one port connected to an end station which is a member of the respective virtual local area network whereby to reduce through the said at least one switch the traffic flow attributable to the said respective virtual local area network.
2. 2. A method according to claim 1 wherein each of a plurality of said switches is caused to store an indication of the traffic flow limit for each virtual local area network identified in the local area network.
3. 3. A method according to claim 2 wherein said traffic flow is monitored in each of said plurality of switches.