GB2399989A - Packet control in cellular communications - Google Patents

Abstract

A method of, and apparatus for, controlling data packets routed via a packet control means (14) comprising a buffer (16) to a subscriber unit (34) in a cellular communications system (1). The method comprises determining a probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34); and dependent upon the probability, performing a packet control process for reducing the number of data packets in the buffer (16). The step of performing the packet control process may be further dependent upon the number of data packets in the buffer (16). The packet control process may comprise deleting data packets from the buffer (16) and/or providing for the rate or number of data packets being sent to the buffer (16) to be reduced. The interruption may be due to the subscriber unit (34) being handed over between cells of the cellular communications system (1).

Description

PACKET CONTROL IN CELLULAR COMMUNICATIONS

Field of the Invention

The present invention relates to control of data packets, in particular transmission and storage of packets, in packet switched communication systems incorporating one or more cellular communication systems. The present invention relates in particular, but not exclusively, to the following cellular communication systems: Universal Mobile Telecommunications System (UMTS), General Packet Radio Service (GPRS) and Global System for Mobile Telecommunication (GSM) systems.

Background of the Invention

Packet switched communication systems incorporating cellular communication systems are known. For example, a cellular communication system according to the well known standards General Packet Radio Service (GPRS) and Global System for Mobile Telecommunication (GSM) may be connected to the Internet and communicate therewith using the Internet Protocol (IP). In this situation Internet connection may be provided to an end user using a subscriber unit such as a mobile telephone, usually referred to as a mobile station (MS) in GSM/GPRS terminology.

By way of another example, a cellular communication system according to the well known standard Universal Mobile Telecommunications System - 2 (UMTS) may be connected to the Internet and communicate therewith using the Internet Protocol (IP). In UMTS terminology, a subscriber unit is usually referred to as user equipment (UE).

Under IP, network components providing the Internet connections, including components of the relevant cellular communication system and external components, follow the Transmission Control Protocol (TCP) of the IP.

This is, broadly speaking, an end-to-end protocol that controls transmission and storage of data packets between different nodes providing a transmission route l O in the Internet. TCP controls how many packets are sent and at what rate from a network element over a given route. Under TCP acknowledgement signals (ACKs) are sent back to this network element. Under TCP, the network element adjusts how many packets are sent and at what rate, depending on the ACKs it receives back.

TCP interprets gaps in service (i.e. a break in receipt of ACKs) as congestion, and, in response, it reduces both the rate at which it sends packets and the number of packets that it will send before receiving an acknowledgement. TCP triggers its re-transmission time out and begins a slow start on a new route. That is only a few packets are transmitted before waiting for an ACK from the far end of the route, which considerably reduces throughput especially when the round trip time on the route is long.

When the end of the route is a subscriber unit of a cellular communication system, packet control functionality is required in the cellular communication - 3 system. In GPRS systems, provision of packet control functionality usually includes a packet control unit (PCU) included in or added to a GSM-type base station controller (BSC). In UMTS systems, provision of packet control functionality usually includes providing such functionality as part of a radio network controller (RNC). For convenience, hereinafter such provision in a UMTS system will also be referred to as being provided by a discrete PCU.

The PCU controls receipt and storage of the packets at the node it is located at. Storage of packets is usually in a buffer, hereinafter termed the PCU buffer. The PCU further controls transmission of the packets to the radio elements of its cellular communication system.

Operation of cellular communication systems includes the well known operation of handover or cell re-selection, i.e. the subscriber unit moves from one cell to another and correspondingly service is handed over from one cell to another. In present systems, under operation of TCP over a route that includes a cellular mobile radio component, gaps in service caused by cell re-selection (cell handover in the cellular communication system) are treated in the same way as other gaps in service. In effect, the TCP treats handover as if it were congestion.

This leads to unnecessary or inappropriate actions being taken under TCP. This can result in major performance degradation.

For a typical cell dwell time of the order of 30-60 seconds and a cell reselection time of the order of 5 seconds, the resulting drop in throughput and - 4 slow regain in throughput can have catastrophic results on performance with packets sent and deleted multiple times.

Another disadvantageous effect is that billing (charging) anomalies may arise. For example, packets may be billed for multiple times because of re- transmissions. More particularly, in a GPRS system, packet-billing records are generated when packets are sent from a serving GPRS support node (SGSN) to a PCU. These packets are buffered at the PCU until they can be sent over the air interface to the subscriber unit. However, if the subscriber unit is transferred to another cell, and these packets are not redirected to the new cell, these packets will be deleted and ultimately will be re-transmitted by the TCP and will be billed for again.

Certain processes for controlling packets are known, but they do not adequately deal with the above identified problems.

One known process is called per-mobile flow control. This is conducted by the SGSN and PCU. However, this does not deal with an expected reduction in transmission over the air interface caused by a cell re-selection. Adapting this to assist more with the above identified problems would probably require inconvenient and costly use of larger than usual buffers at the SGSN, for

example.

Another known process is called random early detection (RED). RED is a process used in IP networks to avoid buffers carrying TCP information - 5 overflowing, deleting the last packets sent, and so triggering multiple TCP sessions to simultaneously attempt retransmissions making the overload situation worse. The process works by determining the amount by which a receiving data buffer is occupied above a first threshold, and in response to that amount, deleting packets. The deleted packets are selected at random. When these packets are found to be missing by the TCP layer, the transmission rate is reduced, so reducing the probability of simultaneous re-transmissions occurring in the network. However, RED is not responsive to predicted gaps in transmission, and as currently employed and understood in the art does not l O function to reduce the throughput on a link which is not experiencing a high buffer load condition.

Summary of the Invention

l 5 In a first aspect, the present invention provides a method of controlling data packets, as claimed in claim 1.

In a further aspect, the present invention provides a storage medium storing processor-implementable instructions, as claimed in claim 12.

In a further aspect, the present invention provides apparatus for controlling data packets, as claimed in claim 13. - 6

The present invention tends to alleviate or resolve packet control problems, such as those described above, caused by cell handover and other communications interruptions.

In particular, the present inventors have realised that it may be beneficial to reduce the amount of data in a PCU buffer when a cell handover, or other break in cellular communication, is impending, predicted, or suspected to occur.

In one preferred implementation, when it is determined that a cell handover, or other break in cellular communication, is impending, predicted or suspected to occur, the flow of packets to the PCU buffer is reduced.

In another preferred implementation, when it is determined that a cell handover, or other break in cellular communication, is impending, predicted or suspected to occur, packets are deleted from the PCU buffer. The packets may be deleted in some random fashion. The decision to delete packets, and/or the way in which they are randomly selected for deletion, may preferably be implemented using the same or similar techniques as are used in the RED process.

In another preferred implementation, when it is determined that a cell handover, or other break in cellular communication, is impending, predicted or suspected to occur, the flow of packets to the PCU buffer is reduced and packets are deleted from the PCU buffer. The packets may be deleted in some random fashion. The decision to delete packets, and/or the way in which they are - 7 randomly selected for deletion, may preferably be implemented using the same or similar techniques as are used in the RED process.

Brief Description of the Drawings

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which: FIG. 1 is a schematic illustration of part of a GPRS cellular communication system 1 connected with the Internet; FIG. 2 is a process flowchart showing certain steps carried out in a packet control method of a first embodiment of the invention; FIG. 3 is a process flowchart showing certain steps carried out in a packet control method of a second embodiment of the invention; and FIG. 4 is a process flowchart showing certain steps carried out in a packet control method of a third embodiment of the invention. - 8

Description of Preferred Embodiments

In the first embodiment, the invention is applied to a cellular communications system compliant with, and containing network elements of, GSM and GPRS. The cellular communication system is connected to the Internet.

However, it is to be appreciated the invention can be applied to other types of cellular system (e.g. UNITS). It is also to be appreciated that the invention can be applied to packet switching of packets being received or transmitted from networks other than the Internet, or indeed being switched internally within a cellular communication system.

FIG. 1 is a schematic illustration of part of a GPRS cellular communication system 1 connected with the Internet.

The GPRS cellular communication system 1 comprises a Gateway GPRS Support Node (GGSN) 2, which is arranged to provide a gateway connection with the Internet 4.

The system 1 further comprises a Serving GPRS Support Node (SGSN) 6 which is coupled to the GGSN 2. The SGSN 6 performs high level switching, including determining the location of a particular MS by means of accessing location registers (not shown).

The SGSN 6 comprises a buffer, hereinafter referred to as the SGSN buffer 8. The SGSN buffer 8 stores data packets which are en-route to MSs. - 9 -

The system 1 further comprises base station controllers (BSCs) 10, 12, and base transceiver stations (BTSs) 22, 24, 26. The BTSs transmit and receive radio signals to and from MSs. In this example, BTS 24 is transmitting and receiving radio signals to and from an MS 34.

BSCslO, 12 are coupled to and control the BTSs. In this example, BSC 10 is coupled to, and controls, BTS 22 and BTS 24; BSC 12 is coupled to, and controls, BTS 26.

BSC 10 comprises a Packet Control Unit (PCU) 14, which operates to control the receipt, storage and transmission of data packets. PCU 14 comprises a buffer, hereinafter referred to as PCU buffer 16, for storing the data packets.

Likewise, BSC 12 comprises a PCU 18, the PCU 18 comprising a PCU buffer 20.

The geographical area covered, i.e. served, by a respective BTS 22, 24, 26 forms a respective cell of the cellular radio communication system 1. In this example, the BTS 22 serves a cell 28, the BTS 24 serves a cell 30, and the BTS 26 serves a cell 32. Thus MS 34 is presently in cell 30.

In operation, data packets being routed to MS 34 from the Internet 4 are routed to MS 34 over the following route: Internet 4 - GGSN 2 - SGSN 6 BSC10 -BTS 24 - MS 34. During this, the data packets are buffered at the SGSN buffer 8 and the PCU buffer 16. -

The system 1 as described above corresponds to a typical conventional arrangement and operates in conventional fashion, except as will now be described below in relation to embodiments of the present invention.

In the embodiments described below, PCU 14 has been adapted, by provision of a selective packet reduction module, to offer, and provide for, an improved packet control process involving selective reduction of packets in and/or arriving at PCU buffer 16, as will be described in more detail below.

l 0 However, this adaptation may be implemented in any suitable manner to provide suitable apparatus. The module may consist of a single discrete entity added to a conventional PCU, or may alternatively be formed by adapting existing parts of a conventional PCU, for example by reprogramming of one or more processors therein. As such the required adaptation may be implemented l S in the form of processor-implementable instructions stored on a storage medium, such as a floppy disk, hard disk, PROM, RAM or any combination of these or other storage media. Furthermore, whether a separate entity or an adaptation of existing parts or a combination of these, the module may be implemented in the form of hardware, firmware, software, or any combination of these.

It is also within the contemplation of the invention that the process to be described below may alternatively be controlled, implemented in full or implemented in part by a module added to or formed by adaptation of any other suitable part of the communication system 1. For example, operation of the PCU - 11 14 (and e.g. PCU 18) in this way may be controlled remotely by an adapted form of the SGSN 6.

Further, in the case of other network infrastructures, implementation may be at any appropriate switching node such as any other appropriate type of base station, base station controller etc. For example, in a UMTS system, the processes described later below may be carried out as part of the packet control functionality provided as part of a radio network controller (RNC), and/or as part of a Node-B, since some aspects of RNC functionality may be delegated to l O this network element.

Another possibility is that the various steps involved in determining and carrying out such adaptation (as will be described in more detail below) can be carried out by various components distributed at different locations or entities l 5 within any suitable network or system.

The process steps carried out under the present embodiment by PCU 14 will now be described with reference to the process flowchart of FIG. 2.

At step s2, PCU 14 monitors the radio link between BTS 24 and MS 34.

Any suitable parameter of the radio link may be monitored (including any combination of plural parameters), where the parameter or combination of parameters may usefully indicate or predict that a break in communication, particularly a cell handover, is likely to take place soon. Possible parameters include Received Signal Level (Rvlev), either uplink, downlink or both; Frame - 12 Erasure Ratio (FER); Received Signal Quality (Rxqual), either uplink, downlink or both; neighbour cell measurement; and so on.

At step s4, PCU 14 predicts the probability of cell handover occurring.

This is performed by processing the parameter or parameters monitored at step s2 according to a suitable algorithm. Any suitable algorithm may be used. This will be determined by the skilled person according to the requirements and characteristics of the particular system under consideration. More particularly, the algorithm is written and implemented using knowledge of previous l O behaviour of the system with respect to the parameter or parameters, and related to how such behaviour has previously been a suitable indicator of likely cell handover. Another possibility is for such an algorithm, developed for one system or part of a system, to be used in a further system, on the assumption that similar behaviour is to be expected. Another possibility is to compare the radio l 5 link quality to MS 34 with measurements reported from other cells. In this case, such measurements may be carried out as an additional part of earlier described step s2.

At step s6, PCU 14 compares the predicted probability of cell handover occurring with a predetermined threshold. The predetermined threshold is set according to the requirements of the particular system under consideration. If the handover probability is not greater than the threshold, the process moves to step s16, which will be described later below. However, if the handover probability is greater than the threshold, then the process moves to step s8, as follows. - 13

At step s8, PCU 14 determines a packet processing factor based upon the predicted probability of cell handover. Any suitable algorithm may be employed, according to the requirements of the particular system under consideration. For example, the packet processing factor may be calculated as a function of the value of the predicted probability of cell handover or as a function of the difference between the value of the predicted probability of cell handover and the threshold used in step s6. The function may include one or more separate communication system parameters, such as volume of data traffic, for example.

lO In this embodiment, step s8 (and steps s10 and s12 to be described below) is implemented in a corresponding way to which processing is performed in the earlier mentioned RED process. The packet processing factor is determined as a function of the difference between the value of the predicted probability of cell handover and the threshold used in step s6, such that the packet processing l 5 factor increases in proportion to the predicted likelihood of cell handover.

At step s10, PCU 14 compares the packet processing factor with a predetermined threshold. This predetermined threshold is set according to the requirements of the particular system under consideration. In effect, step s10 is implemented in a corresponding way to which processing is performed in the earlier mentioned RED process. If the packet processing factor is not greater than the threshold, the process moves to step s16, which will be described later below.

However, if the packet processing factor is greater than the threshold, then the packet processing action to be performed in this embodiment will take place, and hence the process moves to step s12, as follows. - 14

At step s12, PCU 14 carries out a packet processing action in response to the process so far having determined this is appropriate in view of the likelihood of cell handover. In this embodiment, the packet processing operation performed by PCU 14 is deletion of some packets from its PCU buffer 16, i.e. at step s12, PCU 14 deletes some packets from PCU buffer 16. The process for determining how many packets, and which packets, to delete may be arranged as desired according to the requirements of the particular system under consideration. Also, the packet deletion may be implemented in a variable fashion dependent upon the occupancy of the PCU buffer 16, i.e. more packets are deleted the more full the PCU buffer 16 is. In this embodiment, packets are deleted in a pseudo random fashion implemented in a corresponding way to which processing is performed in the earlier mentioned RED process. In a preferred embodiment, the number pf packets deleted is such that TCP is triggered to reduce the rate of l 5 packet delivery (i.e. in response to ACKs not being received from the deleted packets) but not to trigger radio link time-out (RTO).

After deletion of the packets, the process moves to step s16. Step s16 is also reached when either of decision steps s6 or s10 has resulted in a negative outcome. At step s16, PCU 14 determines whether packet flow is continuing (i.e. in effect whether the data session is still in process and, if applicable, that handover has not actually happened yet). If packet flow is indeed continuing, then the process returns to step s2 to be repeated. However, if packet flow is no longer continuing, the process is ended. - 15

Thus, in this embodiment, described above with reference to FIG. 2, the size of the PCU buffer 16 is reduced, using an algorithm that is self contained within the PCU 14, ahead of a predicted cell re-selection event. Thus the number of packets that may be lost during a re-selection event is reduced. The mechanism employed in this embodiment is an extended version of the RED algorithm that increases the packet deletion probability for a link in proportion to the predicted likelihood of a cell re-selection event (or other break in communication). A cell re- selection event is predicted from analysis of the radio link quality, for example, from the signal level or FER on the serving link, or l O from comparison of signal level measurements reported by the user terminal to the network.

The process steps carried out by PCU 14 under a second main embodiment will now be described with reference to the process flowchart of l 5 FIG. 3. In this embodiment, the process comprises the same steps s2 to s10 as were carried out in the first embodiment.

However, in this second main embodiment, if, at step s10, the packet processing factor is found to be greater than the threshold, then the packet processing action to be performed in this embodiment is implemented at a new step s13, as follows. In this embodiment, the packet processing operation performed by PCU 14 is the sending back through part or all of the packet flow route (BSC 10 - SGSN 6 - GGSN 2) of a flow control message which will lead to reduced flow of packets to the PCU 14. This embodiment is implemented in a GPRS system, and hence this comprises a message reducing the "maximum - 16 bucket size" information it sends to the SGSN under the GPRS standard. Thus, at step s13, PCU 14 sends a flow control message to the SGSN 6, more particularly a message instructing the SGSN 6 to reduce the maximum bucket size for the PCU 14. (In other types of system, any appropriate flow control message may be employed. For example, a flow rate reduction instruction may be sent to the source node in the Internet of the data packets being received at PCU 14. ) This results in less data packets being received and buffered at PCU 14 of BSC 10 just prior to the predicted handover, thus avoiding the need for them to be sent twice, i.e. once before any such handover and again after the handover. If lO handover does not materialize, the data flow can be returned to the previous rate, by virtue of no flow reduction message being sent the next time the process reaches step s13.

As in the first embodiment, after the packet processing step, which in this l 5 second main embodiment is the flow control message step s13, the process moves to step s16. Step s16 is also reached when either of decision steps s6 or s10 has resulted in a negative outcome. At step s16, PCU 14 determines whether packet flow is continuing (i.e. in effect whether the data session is still in process and, if applicable, that handover has not actually happened yet). If packet flow is indeed continuing, then the process returns to step s2 to be repeated. However, if packet flow is no longer continuing, the process is ended.

Thus, in this second main embodiment, the process performs per MS flow control dependent on the predicted likelihood of cell re-selection (i.e. the flow control has been performed in response to specific conditions relating to the data - 17 transmission to the specific MS 34, in particular the likelihood that this MS will be changing cells. The size of the PCU buffer 16 is reduced, using an algorithm that is self contained within the PCU 14, ahead of a predicted cell re-selection event so that the number of packets that may be lost during a re-selection event is reduced. The cell re-selection event is predicted from analysis of the radio link quality, for example, from the signal level or FER on the serving link, or from comparison of signal level measurements reported by the MS to the cellular communication system. Depending of the predicted likelihood of a cell- reselection event, the per mobile flow control mechanism is initiated to reduce l O the amount of data buffered at the PCU 14 for transmission to that mobile.

A further advantage of this embodiment relates to billing (i.e. charging).

Billing records are typically generated when packets are sent from the SGSN 6 to the PCU 14. In this case, the reduction of data flow implemented in this second l 5 main embodiment reduces the number of data packets that are first sent to PCU 14 then need to be re-sent to the new PCU 18 after handover to the new cell 32.

These data packets are billed for twice, thus double-billing is advantageously reduced in this second main embodiment.

The process steps carried out by PCU 14 under a third main embodiment will now be described with reference to the process flowchart of FIG. 4. In this embodiment, the process comprises the same steps s2 to s10 as were carried out in the first and second main embodiments. - 18

However, in this third main embodiment, if, at step s10, the packet processing factor is found to be greater than the threshold, then the packet processing action to be performed in this embodiment is implemented at a new step s14, as follows. In this embodiment, the packet processing operation performed by PCU 14 comprises deletion of some packets from its PCU buffer 16, and the sending back of a flow control message through the packet flow route (BSC 10- SGSN 6 - GGSN 2), in this example to the SGSN 6. Thus, at step sly, PCU 14 deletes some packets from PCU buffer 16 and also sends a flow control message to the SGSN 6. Details of the deletion of some packets are as in the first lO main embodiment, i.e. as described above for step s12 of FIG. 2. Details of the sending of a flow control message are as in the second main embodiment, i.e. as described above for step s13 of FIG. 3.

As in the first and second main embodiments, after the packet processing l 5 step, which in this third main embodiment is the combined packet deletion and flow control message step s14, the process moves to step s16. Step s16 is also reached when either of decision steps s6 or s10 has resulted in a negative outcome. At step s16, PCU 14 determines whether packet flow is continuing (i.e. in effect whether the data session is still in process and, if applicable, that handover has not actually happened yet). If packet flow is indeed continuing, then the process returns to step s2 to be repeated. However, if packet flow is no longer continuing, the process is ended.

Thus the various benefits of the first and second main embodiments may be achieved in combination in this third main embodiment, with the added - 19 advantage that these increased benefits may be achieved with only a small increase in processing as many of the steps of the overall processes are shared.

In the above described embodiments, the decision processes, thresholds etc. for determining whether the likelihood of a cell handover event are as described. It will be appreciated that in other embodiments, other decision processes, thresholds and so on may be employed. For example, operating parameters may be monitored and analysed so as to recognise patterns of behaviours indicative of impending handover. Another possibility is that the location of the MS may be monitored, and this used to predict that handover is likely.

In the above described embodiments, the potential communications interruption whose prediction leads to packet processing (comprising packet l 5 deletion and/or flow rate reduction) is cell handover of the MS. However, the present invention further encompasses applying the packet processing in response to predicting other forms of communications interruption. The present invention may be applied to any type of communications interruption where the packet processing actions would be beneficial compared to allowing the system to respond as if the resulting delay in ACKs was due to congestion. Examples of other types of communication interruption are entering a region where radio signal quality is poor, e.g. in shadow of buildings or tunnels or excessive body loss, entering a region where interference is excessive, entering a regionwhere mobile speed is excessive, entering a region where transmission is prohibited, at a time when mobile battery power is limited (e.g. run down battery or swapping - 20 of batteries), in a communications system comprising a number of ad-hoc links between mobile terminals where movement of terminals leads to a prediction that the communications path may be compromised or break down, perhaps temporarily. - 21

Claims (25)

1. A method of controlling data packets routed via a packet control means (14) comprising a buffer (16) to a subscriber unit (34) in a cellular communications system (1); the method comprising: determining a probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34); and dependent upon the probability, performing a packet control process for reducing the number of data packets in the buffer (16).

2. A method according to claim 1, wherein the step of performing the packet control process is further dependent upon the number of data packets in the buffer (16).

l 5

3. A method according to claim 1 or 2, wherein the packet control process comprises deleting data packets from the buffer (16).

4. A method according to claim 1 or 2, wherein the packet control process comprises providing for the rate or number of data packets being sent to the buffer (16) to be reduced.

5. A method according to claim 1 or 2, wherein the packet control process comprises deleting data packets from the buffer (16) and providing for the rate or number of data packets being sent to the buffer (16) to be reduced. - 22

6. A method according to claim 3 or 5, wherein the packets which are deleted are selected in a random or pseudo-random way.

7. A method according to claim 6, wherein the packets are deleted using a random early detection congestion control algorithm.

8. A method according to claim 4 or 5, wherein the process of providing for the rate or number of data packets being sent to the buffer (16) to be reduced comprises sending a flow control message.

9. A method according to any of claims 1 to 8, wherein the interruption in the communication route comprises an interruption due to the subscriber unit (34) being handed over from a first cell of the cellular communications system (1) for which the packet control means (14) is responsible to a second cell of the cellular communications system (1) for which the packet control means (14) is not responsible.

10. A method according to any of claims 1 to 9, wherein the step of determining the probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34) comprises processing one or more parameters related to a radio link part of the communication route.

11. A method according to any of claims 1 to 10, wherein the step of determining the probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34) comprises comparing - 23 parameters related to a radio link part of the communication route with corresponding parameters reported from other cells of the cellular communications system (1).

12. A storage medium storing processor-implementable instructions for controlling a processor to carry out the method of any of claims 1 to 11.

13. Apparatus for controlling data packets routed via a packet control means (14) comprising a buffer (16) to a subscriber unit (34) in a cellular communications system (1); the apparatus comprising: means for determining a probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34); and means for performing, dependent upon the probability, a packet control process for reducing the number of data packets in the buffer (16).

14. Apparatus according to claim 13, wherein the means for performing the packet control process is further dependent upon the number of data packets in the buffer (16).

15. Apparatus according to claim 13 or 14, wherein the packet control process comprises deleting data packets from the buffer (16). - 24

16. Apparatus according to claim 13 or 14, wherein the packet control process comprises providing for the rate or number of data packets being sent to the buffer (16) to be reduced.

17. Apparatus according to claim 13 or 14, wherein the packet control process comprises deleting data packets from the buffer (16) and providing for the rate or number of data packets being sent to the buffer (16) to be reduced.

18. Apparatus according to claim 15 or 17, wherein the packets which are lO deleted are selected in a random or pseudo-random way.

19. Apparatus according to claim 18, wherein the packets are deleted using a random early detection congestion control algorithm.

l 5

20. Apparatus according to claim 16 or 17, wherein the process of providing for the rate or number of data packets being sent to the buffer (16) to be reduced comprises sending a flow control message.

21. Apparatus according to any of claims 13 to 20, wherein the interruption in the communication route comprises an interruption due to the subscriber unit (34) being handed over from a first cell of the cellular communications system (1) for which the packet control means (14) is responsible to a second cell of the cellular communications system (1) for which the packet control means (14) is not responsible. -

22. Apparatus according to any of claims 13 to 21, wherein the means for determining the probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34) comprises means for processing one or more parameters related to a radio link part of the communication route.

23. Apparatus according to any of claims 13 to 22, wherein the means for determining the probability of an interruption in a communication route between the packet control means (14) and the subscriber unit (34) comprises means for lO comparing parameters related to a radio link part of the communication route with corresponding parameters reported from other cells of the cellular communications system (1).

24. A method of controlling data packets substantially as hereinbefore 1 S described with reference to the accompanying drawings.

25. Apparatus for controlling data packets substantially as hereinbefore described with reference to the accompanying drawings.