GB2575383A

GB2575383A - Selection of radio bearers for scheduling in a mobile communications network

Info

Publication number: GB2575383A
Application number: GB1914885.7A
Authority: GB
Inventors: Kwan Raymond; Bendle Ian; Davies Simon
Original assignee: Cisco Technology Inc
Current assignee: Cisco Technology Inc
Priority date: 2013-04-26
Filing date: 2013-04-26
Publication date: 2020-01-08
Anticipated expiration: 2033-04-26
Also published as: GB2575383B; GB201914885D0

Abstract

A method for setting a reference bit rate for a radio bearer. The method comprises establishing a reference bit rate for the bearer (310); comparing an estimate of delay performance for the bearer to a required delay performance for the bearer (320); and adjusting the reference bit rate for the bearer according to the comparison between estimated and required delay performance (330). The estimate of delay may be a ratio of data transferred on the bearer to an associated bit rate, which may be the lower of the reference bit rate and an achieved average rate. The comparing may comprise calculating a metric representative of the relation between the estimated delay and required delay. The bit rate may be incremented when the metric falls outside a threshold range. The method may allow a non-guaranteed bit rate bearer to have a reference bit rate that satisfies a delay budget. Also disclosed is a method for selecting bearers for scheduling based on a comparison of predicted delay and delay budget.

Description

SELECTION OF RADIO BEARERS FOR SCHEDULING IN A MOBILE COMMUNCIATIONS NETWORK

The present invention relates to a method for selecting radio bearers for scheduling in a mobile communications network. The invention also relates to a method for setting a reference bit rate for a radio bearer in a mobile communications network and to a basestation configured to operate in accordance with the above methods.

In mobile communications networks, it is necessary to schedule access to radio resources in order to allow users of the network to access their subscribed services. Radio resources are allocated to individual user equipment devices (UEs), allowing them to receive data via downlink and transmit data via uplink. The scheduling challenge is to allocate radio resources in such a way as to satisfy quality of service (QoS) requirements and to maximise overall system performance. System performance may be represented for example by an objective function relating to system throughput or may by some other metric aimed for example at improving fairness in the allocation of resources between users.

In the Long Term Evolution (LTE) telecommunications standard, the scheduling process is performed by the MAC scheduler within the eNodeB MAC sublayer. The following discussion explores the scheduling problem within the example of the LTE standard, but it will be appreciated that the issues discussed are equally relevant to other telecommunications standards including for example Wideband Code Division Multiple Access (WCDMA).

The LTE air interface uses a transmission time of 1ms, which corresponds to a unit of one subframe. Each subframe is divided in the time domain into two slots of 0.5ms in length. Each slot is divided in the frequency domain into a number of resource blocks. The task of the MAC scheduler is to allocate resource blocks to radio bearers carrying data to or from UEs.

In order to reduce complexity, most MAC schedulers operate in two phases as shown in Figure 1: Time Domain Packet Scheduling (TD-PS) 2 followed by Frequency Domain Packet Scheduling (FD-PS) 4. Radio bearers for scheduling are held in data buffers 6 before entering TD-PS 2 and FD-PS 4. The TD-PS 2 creates a Scheduling Candidate Set (SCS) which is a list of radio bearers which may be allocated resources in the current scheduling period. The FD-PS 4 receives the SCS and determines the actual allocation of resource blocks to users of the selected bearers in the SCS. The creation of the SCS is typically based on metrics calculated for each radio bearer to be scheduled, and on the basis of which the bearers may be ranked in a list. A higher metric value indicates a greater level of importance for the associated radio bearer, and so results in a higher priority for that bearer in the list. There is typically a hard limit N on the number of bearers that can be scheduled at any one time, and the SCS is assembled from the N highest ranked bearers in the list.

The separation of schedulers into time domain and frequency domain scheduling components provides scheduling flexibility, allowing each domain to be independently configured, possibly using different metrics in order to provide the scheduling characteristics required. For example, different circumstances or network operators may dictate the prioritising of high throughput, fairness, low packet drop rate etc, and the flexibility provided by individually configurable scheduling components assists in the delivery of these scheduling priorities. In addition, the separation of time and frequency domain scheduling allows the structure of the frequency domain packet scheduler to be simplified, so decreasing overall component complexity.

Owing to the central role of the scheduler in determining overall system performance, considerable development effort has been devoted to the improvement of scheduling performance. However to date, limited support has been provided for the simultaneous scheduling of multiple users. It is also commonly assumed that a clear distinction may be made between real time and non-real time traffic for scheduling, and that these traffic types will correspond to particular QoS requirements. However, the boundary between real time and non real time traffic is not always clear cut, particularly in LTE, and the assumption that real time and non real time traffic flows correspond to particular QoS requirements also does not hold true for all situations.

According to an aspect of the present invention, there is provided a method for selecting radio bearers for scheduling in a mobile communications network, comprising:

comparing predicted delay and bearer delay budget for at least some of the bearers;

selecting bearers for scheduling based on the comparison between predicted delay and delay budget; and retaining remaining bearers for future scheduling.

According to another aspect of the present invention, there is provided a method for setting a reference bit rate for a radio bearer in a mobile communications network, comprising:

establishing a reference bit rate for the bearer;

comparing a representation of actual delay performance for the bearer to a required delay performance for the bearer; and adjusting the reference bit rate for the bearer according to the comparison between representation of actual and required delay performance.

According to another aspect of the present invention, there is provided a basestation adapted to operate in accordance with the method of the first or second aspects of the invention.

For a better understanding of the present invention, and to show more clearly how it may be put into effect, reference will now be made, by way of example, to the accompanying drawings, in which:Figure 1 illustrates operation of a scheduler;

Figure 2 is a flow chart illustrating steps in a method for selecting radio bearers for scheduling;

Figure 3 illustrates an example of operation of a scheduler according to the method of Figure 4;

Figure 4 is a flow chart illustrating steps in a method according to that of Figure 2;

Figures 5 and 6 illustrate a worked example of the method of Figure 4;

Figure 7 is a flow chart illustrating steps in another method according to that of Figure 2;

Figure 8 illustrates an example of operation of a scheduler according to the method of Figure 7;

Figures 9 and 10 illustrate worked examples of the method of Figure 7;

Figure 11 is a flow chart illustrating steps in a method for selecting a reference bit rate for a radio bearer; and

Figure 12 is a flow chart illustrating a worked example of a method according to that of Figure 11.

The following description illustrates aspects of the present invention with reference to a network operating on LTE. However the methods and apparatus provided by the present invention may be applied in networks operating according to other standards including for example WCDMA.

The following discussion examines scheduling procedures in the downlink system. However, the principles of the present invention may also be applied in the uplink system.

As discussed above, the MAC scheduler within the eNodeB MAC sublayer allocates resource blocks to radio bearers carrying data to or from UEs. In the downlink system, data to be transmitted on radio bearers is stored in the Radio Link Controller (RLC) buffer until radio resources are allocated to the relevant radio bearer. Typically, there is a hard limit on the number of bearers that can be supported at any one time, with radio resources allocated to the bearers based on a metric indicating the importance of the data they have to transmit. Some radio bearers may have only a small amount of data to be scheduled, requiring only a small number of resource blocks to transmit. When too many such bearers with small amounts of important data to be scheduled attempt to access the system simultaneously this situation can result in an inefficient use of radio resources. With each radio bearer only requiring a small number of resource blocks to transmit its associated data, the hard limit on the number of bearers to be scheduled may be reached with many resource blocks still unused. These unused resource blocks represent wasted radio resource, which could have been used to transmit data relating to services of a lower priority.

According to aspects of the present invention, this situation is avoided through the incorporation of delay associated with the data to be scheduled into the scheduling consideration. Aspects of the present invention consider the delay budget associated with a particular radio bearer and compare the predicted delay of the data to be scheduled for the bearer with the bearer delay budget. Bearers are then selected for scheduling on the basis of this comparison. The predicted delay may for example be assessed by considering the amount of data in the buffer and the rate at which that data is likely to be transmitted, as well as the head of line (HOL) delay within the buffer. By taking account of the delay associated with the data to be scheduled, important but less urgent bearers may be delayed to allow for a more efficient use of available radio resources.

Figure 2 is a flow chart illustrating a process by which radio bearers may be selected for scheduling. The process may for example be performed within the MAC scheduler of a basestation.

Referring to Figure 2, in a first step 10, the scheduler compares predicted delay and bearer delay budget for bearers having data to be scheduled. In step 20, the scheduler then selects bearers for scheduling based upon the comparison between predicted delay and delay budget. Finally in step 30 the scheduler retains remaining bearers for scheduling in a future scheduling period. Each of these steps is discussed in further detail below.

According to examples of the method, the step of comparing predicted delay and delay budget may comprise calculating a metric representing a relation between the predicted delay for data to be scheduled and the delay budget of the radio bearer carrying the data. The predicted delay may be considered as a combination of the existing head of line delay in the data buffer and the likely delay arising from transmission of the data.

Let be the amount of data in the RLC buffer for radio bearer i, R, be the average bit rate achieved for radio bearer i, and 5,. be a safety factor that provides a level of protection, rendering the metric more or less conservative in its selection of bearers for scheduling. Let be the packet delay budget for radio bearer i, and be the time stamp of the head of line (HOL) packet delay within the RLC buffer for the radio bearer i.

The HOL packet delay ^^!)can be obtained by subtracting 20ms between the PCEF (Policy and Charging Enforcement Function) and the eNB (evolved NodeB or basestation) from the PDB (packet delay budget) of the bearer. ZTand ^^!)are defined in 3GPP TS 23.203 V9.6.0. Both ZTand can be sent to the scheduler via the SCHED_DL_RLC_BUFFER_REQ parameter defined in the LTE MAC Scheduler Interface Specification v1.11 12-10-2010.

The packet delay budget for the radio bearer can be obtained via the associated GCI (Guality of Service Class Identifier) value which can be found in the LogicalChannelConfigListElement within the CSCHED_LC_CONFIG_REG parameter. As suggested in the LTE MAC Scheduler Interface Specification v1.11 12-10-2010, this parameter is sent from the MAC layer to the scheduler. The average bit rate achieved for radio bearer /?, can be measured at the scheduler internally, although confirmation of this value may also be obtained later.

The comparison of predicted delay and delay budget may be framed as a filtering condition for the delay based selection of bearers for scheduling as indicated in equation (1) below:

-=- + t⁽o^l)+0i >t^ (1)

Λ

It is noted that the average bit rate achieved for a radio bearer may vary considerably, and a very high average throughput R. may potentially be achieved if the channel quality of the UE in question is very good. Thus, the above equation may be modified to allow for the use of a realistically assumed bit rate based on the guaranteed bit rate (GBR) T^.when the average throughput is significantly high. By working from a realistically assumed bit rate under such circumstances, the chance of scheduling the data associated with the bearer too early is reduced. Conversely, if the average achieved bit rate is below the GBR, a more realistic estimate of the likely delay for the data can be obtained by using the currently achieved bit rate in the above equation. The value of the GBR for a bearer can be obtained in the LogicalChannelConfigListElement as described above. If the bearer is a non-GBR bearer without an assigned guaranteed bit rate (as may for example be the case with best effort services such as internet browsing and file download), then a pre-selected value R. may be used. The pre-selected value may be a nominal value selected by a network operator, or may be calculated according to parameters including a delay requirement, source traffic rate and traffic patterns for the bearer. Thus, the delay based filtering condition is modified to be

(2) where if non GBR if GBR (3)

The first term of equation (2) represents a realistic estimate of the likely transmission delay for the data to be scheduled, this delay being represented as the ratio of the amount of data to be scheduled to the likely rate at which it will be transmitted. The second term represents the current HOL delay for the data in the buffer. Finally, the third term introduces a margin of safety. The combined transmission delay, HOL delay and safety factor are then compared to the transmission budget for the bearer to establish the urgency with which the bearer should be scheduled.

Underlying equation (2) is the understanding that at a given bit rate, the lower of the average achieved bit rate and the GBR, a higher value of //.requires a longer time to be cleared. Thus, if the transmission delay plus the HOL delay for a particular radio bearer is much smaller than the bearer delay budget offset by the safety factor, then there is little incentive for this radio bearer to be scheduled. Immediate scheduling of the bearer would result in the data arriving well before the expiry of the relevant delay budget, owing to the relatively small buffer size with respect to the transmission rate. Even if the data contained in the buffer is highly important, scheduling of the bearer can safely be delayed without risking exceeding the delay budget for the bearer. Another bearer having a greater amount of data to be scheduled, and a smaller margin with respect to its delay budget, can be scheduled in place of the low data bearer, thus making more efficient use of the available resource blocks in the current scheduling period.

The low data, important bearer may be delayed and held in the data buffer until a later scheduling period. During this delay, the HOL delay of the low data bearer will increase, and additional data may arrive in the buffer, thus increasing the values of tf¹and D,, and consequently increasing the chance of the bearer being scheduled in the next scheduling period. By delaying bearers to more closely match data delivery with delay budget, the overall number of low data bearers to be scheduled is reduced, as bearers that can be delayed are held back, allowing more data to build up in the data buffer. In addition, the simultaneous scheduling of such buffers is also reduced. Both these factors allow for an increase in the efficiency with which radio resources are used, reducing the number of unused resource blocks in any single scheduling period.

In order to assist with selection of bearers for scheduling based on the above discussed comparison between predicted delay and delay budget, the filtering criterion (2) may be expressed as a delay based metric B:

+ <% (4)

Where/?, is the delay based metric for bearer i. The value of B_: indicates the relation between predicted delay and delay budget, with negative values of B_t indicating data that can still be delivered on time, B_: = 0 indicating data that should be scheduled as soon as possible to arrive within the delay budget, and positive values of B, indicating data that may potentially be out of date unless a very high bit rate becomes available and the data can be scheduled very quickly. Use of the delay based metric B_: assists in selection of bearers for scheduling, by allowing for the definition of an urgency criterion, and for the ranking of bearers according to the urgency with which they should be selected for scheduling, as discussed in further detail below.

Referring again to Figure 2, following the comparison between predicted delay and delay budget for radio bearers, the scheduler proceeds in step 20 to select bearers for scheduling based on the comparison, before retaining in step 30 those bearers not selected so that they may be scheduled in a future scheduling period, as discussed above.

The step 20 of selecting bearers for scheduling may be implemented in different ways, and at different stages with reference to the TD-PS and FD-PS processes. The result of the selection process is to present bearers for FD-PS scheduling, however the delay based filtering of radio bearers may be incorporated before or after or in combination with the application of the existing TD-PS metric.

The stage at which the delay based metric is introduced into the existing processes may also influence the way in which the step of selecting bearers for scheduling is implemented. In one example, the step of selecting bearers may comprise ranking bearers according to their metric B, and selecting for example the top N ranked bearers, where N is a limiting number which may the hard limit for the number of bearers that may be simultaneously supported, or may be some other limiting threshold. In other examples, an urgency criterion may be defined, and all bearers satisfying the urgency criterion may be selected. In still further examples both ranking and the application of an urgency criterion may be used in a combined manner.

An urgency criterion may for example be defined with respect to a limiting value of B, above which bearers are considered to be urgent. As discussed above, one approach would be to consider all bearers having a positive vale of B_: to be urgent, as a positive B_: indicates data that may potentially be out of date. However, in order to introduce a level of flexibility and a margin of safety into the criterion, an urgency value η may be defined, where η > 0. The urgency criterion may then be defined as a value of £>, satisfying B_: > -η .

Figures 4 and 7 are flow charts illustrating steps in two methods 100, 200, according to which the steps of the method of Figure 2 are implemented in different ways. The methods may be conducted in a MAC scheduler of a basestation. In the method of Figure 4, the selecting step 20 is conducted following application of an existing TD-PS metric. In the method of Figure 7, the selecting step is conducted before application of the existing TD-PS metric. Each of the methods of Figures 4 and 7 incorporates the use of an urgency criterion, which criterion represents the overriding condition for bearer consideration. Bearers fulfilling the urgency criterion are given priority over bearers that do not fulfil the urgency criterion according to both methods. However, the methods deal differently with situations involving a large number of urgent bearers for scheduling. In such cases, where a selection must be made form among many bearers all satisfying the urgency criterion, the method of Figure 4 relies upon the already applied TD-PS metric. In contrast, the method of Figure 7 relies upon the delay based metric. Each of the methods 100, 200 is discussed in further detail below with reference to Figures 3 to 10.

According to the method of Figure 4, the selecting of bearers according to a delay comparison may be incorporated after the application of an existing TD-PS metric. In this manner, the delay based selection acts as a post filtering step with reference to the TD-PS. This implementation is illustrated in Figure 3, where the selecting step can be seen at 8, illustrated as a post TD-PS filter.

Referring to Figure 4, in a first step 102 of the method 100, the scheduler computes a TD-PS metric M_t for each of the bearers to be scheduled. The metric M_t forms part of a TD-PS scheduling criterion which may be selected according to the scheduling characteristics required by a network operator, and may for example maximise throughput or apply some fairness based criterion. The scheduler then ranks the bearers according to their M_t values in step 104. The scheduler then proceeds to compute the delay based metric B_t for all bearers at step 106, and at step 108, the scheduler assesses whether the number N’ of urgent bearers is less than the number N of bearers to be selected for FD-PS. Urgent bearers are defined as bearers satisfying the urgency criterion of B_t >-η, and N is the hard limit imposed on the number of bearers to be included in the SCS. If the number N’ of urgent bearers is greater than or equal to the limit N on the number of bearers for FD-PS scheduling (No at step 108), then the scheduler retrieves the M_t ranking of the N’ urgent bearers and selects from among the N’ urgent bearers the N highest ranked bearers according to theirA/, value for inclusion in the SCS at step 110. If the number N’ of urgent bearers is equal to the hard limit N on the number of bearers for FD-PS scheduling, step 110 has the effect of selecting all N’ urgent bearers as the SCS. If, however, the number N’ of urgent bearers is less than the hard limit N (Yes at step 108), the scheduler proceeds to select the N’ urgent bearers for inclusion in the SCS at step 112. The scheduler then fills the remaining available places in the SCS with the N - N’ non urgent bearers having the highest ranking.

A worked example of the method of Figure 4 is illustrated in Figures 5 and 6. Figure 5 illustrates a situation in which many urgent bearers are presented for scheduling, while Figure 6 illustrates a situation in which relatively few urgent bearers are presented for scheduling. According to the examples of both Figures 5 and 6, the scheduler is presented with 6 bearers for scheduling, and must select 4 out of the 6 bearers for FDPS. At step 102, the scheduler computes the metric M_t for each bearer / according to the required criterion, resulting in the set 102a. In step 104, all bearers are ranked according to the calculated metric, resulting in the set 104a of ranked metrics representing the metric of bearer i as the /-th largest among all bearers. In step 106, the scheduler computes the delay based metric B_t for all bearers and in step 108, the scheduler assesses whether the number of urgent bearers (satisfying B_t > -η) is less than the hard limit N. .

In the example of Figure 5, the scheduler finds that bearers 5, 6, 1, 3 and 4 all satisfy the urgency criterion, B_t >-//,/= 5,6,1,3,4 and the scheduler therefore proceeds at step 110 to select from among the urgent bearers the 4 bearers having the highest M_tranking. This results in an SCS 110a comprising bearers 5, 6, 1 and 3.

According to the example of Figure 6, only bearers 2 and 1 are found to have a B_tmetric satisfying the urgency criterion and thus to be classed as urgent, B_t > -η,ΐ - 2,1. The scheduler therefore proceeds to select bearers 2 and 1 for the SCS and to fill the remaining 2 places in the SCS with the non urgent bearers having the highest ranking. This results in an SCS 114a comprising bearers 2, 1, 5 and 6.

The above worked examples demonstrate how the method of Figure 4 prioritises the TD-PS packet scheduling metric while applying the urgency criterion as a post scheduling filter, to ensure that regardless of the TD-PS scheduling results, a non urgent bearer is not scheduled at the expense of an urgent bearer. Thus important but non urgent bearers may be delayed to ensure that more urgent bearers can be scheduled in time to meet their delay budgets. By applying the urgency criterion as an overriding condition, the scheduler matches data delivery to data budgets, and so enables a more efficient use of radio resources.

Figure 7 illustrates another example method, in which the selecting of bearers according to a delay comparison may be incorporated before the application of an existing TD-PS metric. The delay based selection thus acts as a pre filtering step with reference to the TD-PS, applying the TD-PS ranking metric only after delay based assessment has been made. This implementation is illustrated in Figure 8, where the selecting step can be seen at 8, illustrated as a pre TD-PS filter.

Referring to Figure 7, in a first step 202 of the method 200, the scheduler computes the delay based metric B_t for each of the bearers to be scheduled, and then in step 204 the scheduler ranks the bearers according their values of B_t. In step 206, the scheduler assesses whether the number N’ of bearers having a B_t value that satisfies the urgency criterion of B_t > -η is less than the hard limit N for the number of bearers to be passed for FD-PS. If the number N’ of bearers satisfying the urgency criterion is greater than or equal to the limit N (No at step 206), the scheduler selects the N bearers having the highest B_t values as the SCS at step 208 and forwards these bearers for FD scheduling. In the situation where the number N’ of urgent bearers equals the hard limit N, the effect of step 208 is to select all N’ urgent bearers for the SCS. However, if the number of urgent bearers is less than N (Yes at step 206), then the urgent bearers are selected for the SCS at step 210 and the scheduler proceeds to compute the TD-PS metric M_t for the remaining bearers at step 212. The remaining bearers are ranked according to the metric M_t at step 214 and the SCS is completed up to the limit number N with the highest ranked bearers according to the metric M_t in step 216.

In the pre TD-PS example of Figure 7, the value of the urgency criterion factor η impacts upon whether or not the TD-PS metric is actually applied. In a very conservative arrangement, in which η is set to a very large positive value, the number of bearers satisfying the urgency criterion will be large, and is likely to exceed the number of bearers to be selected for FD-PS. In such an arrangement, the highest B_t value bearers are selected for FD-PS and the TD-PS is conducted entirely based on the delay metric B, with the exiting TD-PS metric not performed. A less conservative, smaller value of η will result in fewer urgent bearers and a greater likelihood of computing the existing TD-PS metric M_t.

A worked example of the method of Figure 7 is illustrated in Figures 9 and 10, showing the effects of a lower value η in Figure 9 and a higher value η in Figure 10. According to the example of Figures 9 and 10, the scheduler is presented with 6 bearers in the RLC buffer for scheduling, and must select 4 out of the 6 bearers for FD-PS. At step 202, the scheduler computes the metric B_t to generate the set 202a. The bearers are then ranked at step 204 according to the B, values to generate the ranked set 204a. In the case of a lower value η as illustrated in Figure 9, at step 206, the scheduler discovers that out of the six bearers, only bearers 5 and 6 have a metric B_t satisfying the urgency criterion: B_t >-7/,/ = 5,6. Bearers 5 and 6 are therefore selected for the SCS at step 210. The scheduler then proceeds at step 212 to calculate M_t for the remaining bearers and to rank the remaining bearers according to in step 214. The scheduler discovers that M₃ > >M₃ and thus bearers 3 and 4 are selected at step 216 to complete the SCS 216a, which comprises bearers 5, 6, 3 and 4.

Figure 10 illustrates how the selection procedure progresses in the event of a high value η. In this case, at step 206, the scheduler determines that all of the bearers satisfy the urgency criterion, meaning the top 4 ranked bearers are selected as the SCS for FD-PS at step 208 and no further ranking based on the metric is performed.

Each of the pre and post TD-PS methods described above with reference to Figures 3 to 10 has the effect of more closely matching data delivery to delay budgets. In cases of relatively few urgent bearers, up to the hard limit N, each approach selects the urgent bearers for the SCS and completes the SCS with the highest ranked bearers according to the TD-PS metric M_t. In cases involving many urgent bearers, the post filter approach of Figure 4 selects from among the urgent bearers using the TD-PS metric M_t, while the pre filter approach selects from among the urgent bearers based on their urgency, using the delay based metric B_: ranking.

In some cases, the post filtering approach of Figure 4 may offer a more efficient treatment of bearers associated with very poor channel quality. Such bearers are likely to have a combination of low M_t ranking and high B_t ranking, both explained by the poor channel quality. Channel quality is often an important factor in the TD-PS metric M_t, and the low channel quality is likely to translate to a high urgency owing to the low transmission rate associated with the bearer. It can be more resource efficient for the overall network not to prioritise such low channel quality bearers except when resource availability is very good. While the pre filter approach of Figure 7 is likely to schedule such low quality bearers regardless of other conditions, the post filter approach of Figure 4 is likely to have a different result depending upon the total number of urgent bearers. According to the method of Figure 4, if there are few urgent bearers to be scheduled, indicating good resource availability, then the low channel quality bearer will be scheduled along with all urgent bearers. However, if many urgent bearers are presented, indicating high resource usage, then the scheduler selects from among the urgent bearers according to their M_t ranking, which as discussed above will be low for a low channel quality bearer, meaning the low channel quality bearer is unlikely to be scheduled. Thus during periods of high resource usage, the method of Figure 4 filters out the low channel quality bearer in favour of more resource efficient bearer selection.

According to a further example of the present method, the delay based metric and existing TD-PS metric may be combined to create a single ranking metric:

= M_iB_i (5)

This example offers the advantage that only one level of ranking of M\ ,X/i is needed, removing the need for the two stage processes described above. However, the price of this increased calculation efficiency is that the effects of the two ranking metrics become blurred. For example, even if a bearer had little data to send, if the user associated with the bearer had exceptionally good channel quality, which is typically an important component of M_t, it is possible for the channel quality to dominate the overall metric, suppressing the effect of B_t. Thus, while the combined metric may offer advantages in certain circumstances, the two stage approach discussed above with reference to Figures 3 to 10 may be of particular use when it is desired to decouple the effects of the delay based metric from the existing TD-PS metric.

Additional control over the combined metric may be achieved by introducing exponents a and β as follows:

' ' ' (5a) where a and //are positive real numbers. In the modified combined metric (5a) the level of contribution of M, and // can be adjusted separately through the exponents a and β. For example, when a = 0, the combined metric returns a value based solely on the delay based metric B_r Conversely, when // = 0, the combined metric returns a value based solely on the original metricM_t. Various combinations of M_tand B_t between these extremes may be achieved through appropriate selection of values for a and β.

According to further examples, the delay based metric B_t used in the above described methods may be adapted such that it is normalised with respect to the delay budget of the radio bearer. As discussed above, the delay based metric B_t indicates a relation between the predicted delay for data associated with a radio bearer and the delay budget of the radio bearer. That relation is described according to equation (4) as a time difference between the predicted arrival time and the delay budget. This difference represents the absolute difference between the two quantities. However, a difference of the same magnitude may have varying implications for bearers with different delay budgets. For example, if delay budget Y of one bearer is 10 times larger than the delay budget X of another bearer, the same absolute difference between predicted delay and delay budget may have a different impact between the two bearers. In order to reflect the relative importance of the absolute difference with respect to the delay budget, the equation (4) for the delay based metric B_t may be modified as follows:

(6)

The urgency criterion for determining selection of bearers for FD scheduling can be re cast as B_t >η_ί. In this case, the safety factor 5_i need not be used, and the level of conservativeness desired in the scheduling selection may be embedded in the threshold η_:.

As discussed previously, the value of the threshold η_: in the above urgency criterion determines the threshold for delaying or scheduling of radio bearers. A higher threshold value potentially retains, or “gates” more bearers before they can be scheduled than a lower value. By gating more bearers and so reducing the number of bearers having small amounts of data to be scheduled, the available radio resources can be used more efficiently, allowing for an overall improvement in delay performance. Studies indicate that for values of 7_; of 0 and 0.65 the average number of bearers to be scheduled is 3.0 and 1.3 respectively, indicating how the number of bearers for simultaneous scheduling decreases with higher values of 7,.

Further modifications may be made to the delay based metric, for example to introduce weighting factors, allowing the selective weighting of the delay component caused by the HOL delay or the predicted transmission delay. For example, the following equation may be used:

λ \ D_t (1 - a)----/_ ~ \ + a · t _n ’ min/?.,/?., ^B< =------------ ω max

In equation (7), the factor a enables selective emphasis of the transmission delay or HOL delay as desired.

In some circumstances the level of urgency associated with scheduling a radio bearer may have a non linear relationship to the size of predicted delay relative to the delay budget. For example, for the same predicted transmission delay, the actual level of urgency associated with a HOL delay of 80% of the delay budget may be more than two times greater than the urgency associated with a HOL delay of 40% of the delay budget. The same non linear relation applies to the size of the buffer, and consequently the predicted transmission delay. Weighting functions may thus be introduced to reflect the changing importance of the HOL and transmission delay factors:

Where7iVJand -UWare functions placing different non linear emphasis on the respective input quantity such that a higher input value gives rise to a higher emphasis and vice versa.

The methods described above illustrate how methods according to the present disclosure may enable more efficient use of radio resources. By delaying the scheduling of bearers having little data to send, radio resources are feed up for more urgent bearers, which by the nature of their urgency are likely to have more data to send and so make greater use of the available resources. In addition, the total number of simultaneous bearers to be scheduled is reduced. These methods are thus particularly advantageous where hardware places a limit on the number of bearers that may be simultaneously scheduled.

Radio resources are frequently limited on communication network data channels and particularly on control channels. By enabling more efficient use of these resources, the above described methods allow more data to be transmitted in any one scheduling period, thus improving the performance of the networks.

The above discussed methods make use of the guaranteed bit rate associated with a bearer as a possible representative bit rate for the bearer in question. The GBR may be used as the representative bit rate if it is lower than the average achieved bit rate for the bearer.

For bearers having a GBR, (typically bearers associated with real time services), the GBR plays an important role in how the cell load is measured. For example, GBR is used directly in estimating the load contribution of a bearer, and the combined load from all bearers determines the load of a cell. Thus, if the GBR is set unnecessarily high for example, the cell load estimate becomes too high, giving rise to a higher likelihood of blocking entry to the cell and preventing efficient use of cell resources.

GBR is also considered as an important component of Quality of Service (QoS), although the ultimate measure of quality is the delay experienced from the users’ point of view. If the GBR is set too high, packets are delivered well before the delay budget at the expense of radio resources. This advance does not translate to an appreciable improvement for the user, and thus the additionally used radio resource is not contributing to user experience and so not being efficiently deployed. In contrast, if the GBR is set too low, delay budgets are not met; an impact that is felt by the user and negatively affects the user experience. In practice, the source bit rate for data, transmission bit rate, and delay budget are not always well matched. In the event of a mis-match between the actual source bit rate and the specified GBR, it is more resource efficient to adjust the GBR value to better match the source rate, subject to other QoS constraints such as the delay budget. An intelligent adjustment of the GBR, such as that provided in the following disclosure, may provide a more optimal balance between the radio resource consumption and user experiences.

It has previously been discussed that some radio bearers do not have an associated GBR, typically those bearers associated with best effort services such as file download and internet browsing. These bearers may nonetheless have a delay budget requirement, and thus it is expected that packets of such bearers will arrive within the delay budget and yet no bit rate is guaranteed to the bearer. In order to resolve this apparent contradiction, the present disclosure proposes a method by which a network may make use of an internally generated and adjusted reference bit rate for the bearer, which bit rate is matched to the delay performance of the bearer.

Figure 11 illustrates a method for setting a reference bit rate for a radio bearer. The reference bit rate may be a guaranteed bit rate, or may be a representative bit rate, in the case of bearers for which no guaranteed bit rate is provided. The method of Figure 11 may be conducted for example in a basestation of a mobile communications network, under the control of a processor.

Referring to Figure 11, in a first step 310, the basestation establishes a reference bit rate for the bearer. The basestation then compares an estimate of delay performance for the bearer to a required delay performance for the bearer in a step 320. Finally, the basestation adjusts the reference bit rate for the bearer according to the comparison between the estimate of delay performance and the required delay performance.

The step of establishing a reference bit rate for the bearer may be implemented differently depending upon whether the bearer in question is a GBR bearer or a non GBR bearer. In the case of a GBR bearer, establishing a reference bit rate for the bearer may simply comprise retrieving the GBR for the bearer. In the case of a non GBR bearer, establishing the reference bit rate may comprise retrieving a representative, preselected bit rate for the bearer. This pre-selected bit rate may for example be a nominal bit rate selected by a network operator. Alternatively, the preselected bit rate may be calculated so as to be representative of requirements and conditions for the bearer. For example, the pre-selected bit rate may be calculated based upon the delay requirement of the bearer, the traffic source rate and traffic patterns including how bursty the traffic is. In some examples, establishing the representative bit rate may comprise retrieving the necessary data and calculating the pre-selected bit rate.

The comparison between an estimate of delay performance and required delay performance may be conducted using the following equation:

Where Q/R J is the comparison metric.

D

The term ----/_ ~ \ provides an estimate of the delay due to the transmission of min(7?_;, R,) data. The value of D_: indicates the amount of data in the buffer to be transferred, while R_t and R, represent the average achieved bit rate and the established reference bit rate for the bearer, as previously defined. represents the head-of-line (HOL) delay, and represents the delay budget and so Q;(r,J provides a normalised comparison of an estimate of delay performance with required delay performance.

It will be appreciated that the above equation (9) for the comparison metric corresponds to equation (6) for the delay based metric used in the above described scheduling selection procedures. The modifications to equation (6) discussed with reference to equations (7) and (8) may equally be applied to equation (9) for the comparison metric, introducing weighting factors or functions to selectively weight the importance of the HOL and transmission delay in the comparison metric.

The step of adjusting the reference bit rate based on the comparison performed in step 320 may be implemented in various ways, two examples of which are discussed below. In a first example, illustrated in Figure 12, the adjustment step comprises comparing the computed metric 0,(//) , or an averaged value of (λ (//) over time, to a threshold range and incrementing the reference bit rate when the computed metric falls outside the threshold range. Let the term be the averaged value of ^O^over either a block of time instances (for example block averaging, weighted block averaging, etc.), or an exponentially averaged value as:

0(z + l) = (l-a)g.(z)+«<2'”'>(4) (10) where ^u < ^a - ¹is a filter coefficient, and V ^! 'is the instantaneous measured value of ^^)at time instance ^ + 1. Let h represent a threshold value for with which the network is satisfied, and let η represent a hysteresis. The relation of to the threshold range h < < h+η may be checked every bit rate modification period T.

Referring to Figure 12, at t = 0, Q^t) is updated in step 402 and the basestation waits until the bit rate modification time period T has expired at step 404. If the time period has not expired, the time counter is incremented and continues to be updated.

When the time period T has expired, the basestation first checks whether or not is less than the threshold value h at step 406. If is less than the threshold value, the reference bit rate is reduced by an increment Δ/?, in step 408 and the time counter is reset in step 410 before returning to step 402 to update If Q/ti is greater than h, the basestation then checks at step 412 if is greater than the value of h added to the hysteresis η. If this is the case, the reference bit rate is increased by an increment Δ/ζ in step 414 before resetting the time counter in step 410 and returning to step 402 to update If the value of ^' ^falls between h and h + η, (No at step 412) then no increment of the reference bit rate is required and the time counter is merely reset before returning to step 402 to update

In another example, the adjustment step comprises computing a probability that the calculated metric will fall below a threshold value, and incrementing the reference bit rate when the computed probability falls outside a threshold range. The threshold value may be h, where h represents a minimum value of with which the network is satisfied. Thus all values of (2_;(z?_;)>h are considered to be acceptable, and the probability of being below this acceptability threshold is calculated. The threshold range comprises a probability threshold range of ε < Pr(X) < ε+η. Where Pr(X) is the probability of X, ε is a probability threshold value between 0 and 1, and η is a hysteresis as before.

The adjustment process proceeds according to the following algorithm:

If Pr(g_;(r_;)<h)>s

R_t R_t - &R_t else if Pr((Z ) 7

R_t <r- R_t + Δ/7 end

The reference bit rate is again incremented by a modification step size Δ/?, . As in the above example, an overly small value of q/r,J indicates that the bearer is ahead of schedule, and 7?, may be reduced, while an overly low high value of indicates that the bearer is behind schedule, and R, may be increased.

The above algorithm obtains a probability by keeping a count of the number of times Q_t (/ti )is smaller than h . If the probability is greater than a certain threshold, meaning that the system is doing very well, (i.e. the delay is often well below the delay budget), the reference bit rate (GBR or pre-selected rate) can be reduced. By reducing the reference bit rate, the scheduler may potentially lower the priority of this bearer. In addition, as the reference bit rate is lowered, the load of the bearer is lowered, which eventually reduces the overall cell load. By reducing the overall cell load the system opens up the potential to admit more bearers, thus increasing the system use.

The method described above thus allows for adjustment of a reference bit rate which adjustment improves consistency across bit rate and delay requirements. The result is a more efficient use of radio resources. The reference bit rate bay be the guaranteed bit rate, for GBR bearers, or may be a representative, pre-selected bit rate for non GBR bearers. The pre-selected bit rate may be a nominal value or may be calculated as discussed above.

The above methods assume the availability of a good estimate of the average bit rate R_: achieved for a bearer, which satisfactorily reflects the QoS. For example, if the traffic pattern of a bearer is very bursty, the estimation of R, may exclude the appropriate non-activity period such that R_: is meaningful with respect to its QoS. Depending on the nature of the traffic, a peak rate may be used to estimateR,. The precise method of estimating R, is beyond the scope of this disclosure.

The methods described above may be used in any scheduler implementation operating according to any mobile communications standard, the methods being independent of the standard and precise implementation of the scheduler.

Examples ofthe disclosure are set out in the following list of numbered clauses.

1. A method for selecting radio bearers for scheduling in a mobile communications network, comprising:

2. A method according to clause 1, wherein comparing comprises calculating a relation between predicted delay and bearer delay budget, and wherein selecting according to the comparison comprises selecting for scheduling those bearers for which the calculated relation satisfies an urgency criterion.

3. A method according to clause 2, wherein the urgency criterion comprises a relation indicating the predicted delay is approaching or greater than the delay budget.

4. A method according to clause 2 or 3, wherein the relation comprises a time difference between the predicted delay and the delay budget.

5. A method according to clause 4, wherein the time difference is normalised with respect to the delay budget.

6. A method according to any one of the preceding clauses, wherein the predicted delay comprises a predicted transmission delay.

7. A method according to clause 6, wherein the predicted delay further comprises a head of line delay.

8. A method according to clause 6 or 7, wherein the predicted delay further comprises a safety factor.

9. A method according to any one ofthe preceding clauses, wherein the predicted delay comprises a weighted combination of a predicted transmission delay and a head of line delay.

10. A method according to any one of clauses 6 to 9, wherein the predicted transmission delay comprises a ratio of data associated with the radio bearer to a bit rate associated with the radio bearer.

11. A method according to clause 10, wherein the bit rate associated with the radio bearer comprises the lower of a reference bit rate for the bearer and an estimate of average bit rate achieved for the bearer.

12. A method according to clause 11, wherein, if the bearer is associated with a guaranteed bit rate, the reference bit rate for the bearer comprises the guaranteed bit rate for the bearer, otherwise the reference bit rate comprises a pre-selected bit rate for the bearer.

13. A method according to any one of clauses 3 to 12, further comprising recalculating the relation for retained bearers on expiry of a current scheduling period.

14. A method according to any one of the preceding clauses, wherein comparing comprises calculating a metric representative of a relation between predicted delay and delay budget, and wherein selecting for scheduling comprises ranking the bearers according to the calculated metric and selecting those bearers for which the metric exceeds a threshold value representative of an urgency criterion.

15. A method according to any one of the preceding clauses, wherein a threshold number N of bearers may be selected for scheduling, and wherein the method further comprises:

comparing the number of bearers selected for scheduling to the threshold number N; and;

if the number of bearers selected for scheduling is less than the threshold number, ranking the retained bearers according to a scheduling metric; and selecting for scheduling the top X ranked bearers, wherein X comprises the difference between the number of bearers already selected for scheduling and the threshold number N.

16. A method according to any one of the preceding clauses, wherein a threshold number N of bearers may be selected for scheduling, and wherein the method further comprises:

if the number of bearers selected for scheduling is greater than the threshold number N, ranking the selected bearers according to a metric; and selecting the top N ranked bearers, wherein the metric comprises one of a metric representative of a relation between predicted delay and delay budget and a scheduling metric.

17. A method according to any one of clauses 1 to 14, wherein comparing comprises calculating a metric representative of a relation between predicted delay and delay budget, and wherein selecting for scheduling comprises:

calculating a scheduling metric for at least some of the bearers; combining the calculated metrics;

ranking the bearers according to the combined metrics; and selecting the top X ranked bearers for scheduling, wherein X comprises a threshold number of bearers may be selected for scheduling.

18. A method for setting a reference bit rate for a radio bearer in a mobile communications network, comprising:

establishing a reference bit rate for the bearer;

comparing an estimate of delay performance for the bearer to a required delay performance for the bearer; and adjusting the reference bit rate for the bearer according to the comparison between estimated and required delay performance.

19. A method according to clause 18, wherein the estimate of delay performance comprises a ratio of data transferred on the bearer to a bit rate associated with the bearer.

20. A method according to clause 19, wherein the bit rate associated with the bearer comprises the lower of the reference bit rate and an estimated average bit rate achieved for the bearer.

21. A method according to any one of clauses 18 to 20, wherein comparing comprises computing a metric representative of a relation between the estimate of delay performance for the bearer to the required delay performance for the bearer.

22. A method according to clause 21, wherein adjusting according to the comparison comprises comparing the computed metric to a threshold range and incrementing the reference bit rate when the computed metric falls outside the threshold range.

23. A method according to clause 21, wherein adjusting according to the comparison comprises computing a probability that the calculated metric will fall below a threshold value, and incrementing the reference bit rate when the computed probability falls outside a threshold range.

24. A method according to any one of clauses 18 to 23, wherein the reference bit rate comprises a guaranteed bit rate for the bearer, and wherein establishing the reference bit rate comprises retrieving the guaranteed bit rate associated with the bearer.

25. A method according to any one of clauses 18 t o 23, wherein the reference bit rate comprises a pre-selected bit rate for the bearer, and establishing the reference bit rate comprises retrieving the pre-selected bit rate for the bearer.

26. A basestation configured to select radio bearers for scheduling in accordance with a method according to any one of clauses 1 to 17.

27. A basestation configured to set a reference bit rate for a radio bearer in accordance with a method according to any one of clauses 18 to 25.

Claims

1. A method for setting a reference bit rate for a radio bearer in a mobile communications network, comprising:

establishing a reference bit rate for the bearer;

2. A method as claimed in claim 1, wherein the estimate of delay performance comprises a ratio of data transferred on the bearer to a bit rate associated with the bearer.

3. A method as claimed in claim 2, wherein the bit rate associated with the bearer comprises the lower of the reference bit rate and an estimated average bit rate achieved for the bearer.

4. A method as claimed in any one of claims 1 to 3, wherein comparing comprises computing a metric representative of a relation between the estimate of delay performance for the bearer to the required delay performance for the bearer.

5. A method as claimed in claim 4, wherein adjusting according to the comparison comprises comparing the computed metric to a threshold range and incrementing the reference bit rate when the computed metric falls outside the threshold range.

6. A method as claimed in claim 4, wherein adjusting according to the comparison comprises computing a probability that the calculated metric will fall below a threshold value, and incrementing the reference bit rate when the computed probability falls outside a threshold range.

7. A method as claimed in any one of claims 1 to 6, wherein the reference bit rate comprises a guaranteed bit rate for the bearer, and wherein establishing the reference bit rate comprises retrieving the guaranteed bit rate associated with the bearer.

8. A method as claimed in any one of claims 1 to 6, wherein the reference bit rate comprises a pre-selected bit rate for the bearer, and establishing the reference bit rate comprises retrieving the pre-selected bit rate for the bearer.

9. A method for selecting radio bearers for scheduling in a mobile communications network, comprising:

10. A method as claimed in claim 9, wherein comparing comprises calculating a relation between predicted delay and bearer delay budget, and wherein selecting according to the comparison comprises selecting for scheduling those bearers for which the calculated relation satisfies an urgency criterion.

11. A method as claimed in claim 10, wherein the urgency criterion comprises a relation indicating the predicted delay is approaching or greater than the delay budget.

12. A method as claimed in claim 10 or 11, wherein the relation comprises a time difference between the predicted delay and the delay budget.

13. A method as claimed in claim 12, wherein the time difference is normalised with respect to the delay budget.

14. A method as claimed in any one of the preceding claims, wherein the predicted delay comprises a predicted transmission delay.

15. A method as claimed in claim 14, wherein the predicted delay further comprises a head of line delay.

16. A method as claimed in claim 14 or 15, wherein the predicted delay further comprises a safety factor.

17. A method as claimed in any one of the preceding claims, wherein the predicted delay comprises a weighted combination of a predicted transmission delay and a head of line delay.

18. A method as claimed in any one of claims 14 to 17, wherein the predicted transmission delay comprises a ratio of data associated with the radio bearer to a bit rate associated with the radio bearer.

19. A method as claimed in claim 18, wherein the bit rate associated with the radio bearer comprises the lower of a reference bit rate for the bearer and an estimate of average bit rate achieved for the bearer.

20. A method as claimed in claim 19, wherein, if the bearer is associated with a guaranteed bit rate, the reference bit rate for the bearer comprises the guaranteed bit rate for the bearer, otherwise the reference bit rate comprises a pre-selected bit rate for the bearer.

21. A method as claimed in any one of claims 11 to 20, further comprising recalculating the relation for retained bearers on expiry of a current scheduling period.

22. A method as claimed in any one of the preceding claims, wherein comparing comprises calculating a metric representative of a relation between predicted delay and delay budget, and wherein selecting for scheduling comprises ranking the bearers according to the calculated metric and selecting those bearers for which the metric exceeds a threshold value representative of an urgency criterion.

23. A method as claimed in any one of the preceding claims, wherein a threshold number N of bearers may be selected for scheduling, and wherein the method further comprises:

24. A method as claimed in any one of the preceding claims, wherein a threshold number N of bearers may be selected for scheduling, and wherein the method further comprises:

25. A method as claimed in any one of claims 9 to 22, wherein comparing comprises calculating a metric representative of a relation between predicted delay and delay budget, and wherein selecting for scheduling comprises:

26. A basestation configured to set a reference bit rate for a radio bearer in accordance with a method as claimed in any one of claims 1 to 8.

27. A basestation configured to select radio bearers for scheduling in accordance with a method as claimed in any one of claims 9 to 25.