DESCRIPTION
COMMUNICATION APPARATUS, METHOD, AND COMPUTER IMPLEMENTABLE
PRODUCT FOR ADMISSION CONTROL
Technical Field:
The present invention relates to admission control of traffic into a network, particularly
but not exclusively a cellular or wireless telecommunications network. The invention has
particular but not exclusive relevance to auto-tuning of an admission control threshold for use in
admission control decisions.
Background Art:
In cellular communication networks, it is usual to include some kind of admission control
function, which regulates the admission of arriving calls into the communication network based
on the current load of the network. Typically a call from a new user is admitted if the following
condition is satisfied:
p + Ap p l (1)
where p is the current load of the cell, Ap is the estimated load increase that will be
caused by admission of the new user's call, and p l is a threshold parameter. Various methods
are known for measuring or estimating the load value p for use in call admission decisions.
The threshold p lh represents a trade-off between blocking incoming call requests and the
protection of the Quality of Service (QoS) offered to users already connected to the cell. A
higher threshold results in fewer blocked calls, but may result in more users sharing the available
capacity of the cell, thus reducing the QoS experienced by each connected user.
Typically the threshold p l is set 'manually' by the network operator, which may be on a
per-cell basis. However, this is often challenging in practice because the relationship between
the load threshold p t , user QoS and blocking is not simple to predict. An estimate of a suitable
threshold value can be determined from analysis and computer simulations of network
performance at a base station, but the actual performance is very difficult to predict accurately
because it may depend on many factors such as traffic mix, cell size and local radio environment.
Therefore, a method of automatically adjusting p l would be of considerable benefit,
both in terms of simplifying the management of the network and improving operating efficiency
and user experience.
Summary of the Invention:
Exemplary embodiments of the invention aim to at least partially address some of the
problems with the prior art described above.
According to one aspect of the invention, there is provided a communications apparatus
for controlling call admission at a node in a network, the communications apparatus comprising:
means for obtaining an outage rate for one or more current calls at the node; means for obtaining
a blocking rate of admission requests at the node; means for adjusting a load threshold in
dependence on the outage rate and the blocking rate; means for obtaining an estimate of a current
network load at the node; and means for controlling admission of a call at the node in
dependence on the estimated current load and the load threshold. Thus, the load threshold value
can be adjusted based on parameters that directly affect the experience of a user of the network,
that is blocking rate and outage rate, to automatically reach a desired trade-off between the
blocking and outage when making admission control decisions.
The communications apparatus may be arranged to obtain one or more of the blocking
rate, the outage rate and the estimate of current network load from one or more other elements in
the network, or may obtain one or more of these parameters by determining them itself, for
example from measurements of network operation.
The outage rate may be defined as the fraction of connected users not currently meeting
minimum quality of service requirements, and in particular may be defined as the fraction of
connected users not currently receiving at least a minimum required data rate.
Adjustment of the load threshold value may include: reducing the load threshold value by
a first value if the obtained outage rate is more than a predetermined maximum outage value;
increasing the load threshold value by a second value if the obtained outage rate is less than a
predetermined minimum outage value; increasing the load threshold by the second value if
> Ϊ ' {^outage ~ ^outage )
where Pb{k) is the obtained blocking rate, is the obtained outage rate (or probability),
s e minimum desired outage rate and is the blockage-to-outage ratio, and to decrease
the load threshold if,
) < - g -
and/or only increasing the load threshold value when it is determined that blocking of new call
requests is occurring at the node.
The first and second values may be manually predetermined, and may be equal or
different values. Alternatively, the first and second values may be calculated, for example based
on the obtained outage rate such that the load threshold may be reduced more quickly if a very
large outage rate is experienced and/or the load threshold may be increased more quickly if low
outage is experienced.
The blocking rate may be determined by measuring a number of blocked admission
requests, N (k), and a total number of transmission requests, Nr (k), during a measurement
interval, and using the equation:
Nr {k) =
(*)=
otherwise
Nr (k)
The outage rate may be determined by block-averaging a measured fraction of users,
P (n) , not currently meeting minimum quality of service requirements during an outage
measurement period, AT , over a threshold adjustment period, ATl , and using the equation:
where N
The outage may alternatively be determined by performing a continuous exponential
averaging of a measured fraction of users, P0 (n) , not currently meeting minimum quality of
service requirements during an outage measurement period, , using the equation:
P0 {k)= \ - 0 ) -P0 {k -\)+P 0P0 {n) ,
where 0 < 0 <1 is an averaging weight.
In a further example, particularly applicable to LTE networks, a scheduler may supply
measurement data that can be used in obtaining the outage rate using the equation:
where 7 (t ) is a throughput estimate for each bearer which is updated every subframe, R a
required bit rate for each bearer, B(t) is the number of bearers present in sub-frame t, and the
indicator function l(x) is defined
The node may be a base station, and in particular, the node may be a base station in a
mobile communications network. According to some implementations, the mobile
communications network is compliant with the LTE standards.
According to another aspect of the invention, there is provided a method of adjusting a
load threshold for use in call admission control at a node in a network, the method comprising:
obtaining an outage rate for one or more current calls at the node; obtaining a blocking rate of
admission requests at the node; and adjusting the load threshold in dependence on the outage rate
and the blocking rate.
According to another aspect of the invention there is provide a computer implementable
product comprising computer implementable instructions for causing one or more computer
devices to become configured as any of the above communications apparatus.
The invention provides, for all methods disclosed, corresponding computer programs or
computer program products for execution on corresponding equipment, the equipment itself
(user equipment, nodes or components thereof) and methods of updating the equipment.
Brief Description of the Drawings:
An exemplary embodiment of the invention will now be described, by way of example,
with reference to the accompanying drawings in which:
Figure 1 schematically illustrates a mobile telecommunication system of a type to which
the invention is applicable;
Figure 2 shows a graph illustrating blocking and outage regions;
Figure 3 illustrates a flow chart of a threshold update method according to an
embodiment of the invention;
Figure 4 schematically illustrates a base station forming part of the system shown in
Figure ; and
Figure 5 shows a graph illustrating an alternative definition of blocking and outage
regions.
Description of the Exemplary Embodiments:
Overview
Figure 1 schematically illustrates a mobile (cellular) telecommunication system 1 in
which users of mobile telephones (MT) 3-0, 3-1, and 3-2 can communicate with other users (not
shown) via one of the base stations 5-1 or 5-2 and a telephone network 7. A number of uplink
and downlink communications resources (sub-carriers, time slots etc) are available for the
wireless link between the mobile telephones 3 and the base stations 5. In this embodiment, the
base stations 5 allocate downlink resources to each mobile telephone 3 depending on the amount
of data to be sent to the mobile telephone 3. Similarly, the base stations 5 allocate uplink
resources to each mobile telephone 3 depending on the amount and type of data the mobile
telephone 3 has to send to the base station 5.
The base station generally provides services to the mobile telephones by establishing a
radio bearer for each service that is requested. For example, one radio bearer may be defined for
carrying video data, one may be defined for carrying audio data, one may be defined for carrying
bursty web traffic etc. Various types of radio bearers are defined, depending on the quality of
service associated with the type of data to be carried by the radio bearer. For example, a radio
bearer used for video or audio traffic may be provided in a higher class than radio bearers used
for internet traffic - as users can tolerate delays with internet traffic but cannot tolerate delays in
the delivery of video data. At the time of establishing a new connection, radio bearers are
established that will provide the required service with the desired quality of service.
When a new connection is to be established, the base station 5 must make sure that it has
the resources to be able to provide the desired service and must be able to provide the service
without affecting the service of existing connections. To be able to achieve this, the base station
5 controls admission of new connections into the network based on the existing load on the base
station 5. A new call will only be admitted to the network if the existing load on the base station
5 and the expected load caused by the new connection is less than a threshold load value, plh .
In this embodiment, the threshold load value, p l , is adjusted based on a blocking rate (or
probability) for calls in the system and a measurement of Quality of Service (QoS). Blocking
rate or probability is determined by counting the number of call attempts made, and the number
of blocked calls, at the base station 5.
The QoS measure used in the present embodiment is based on a measurement of outage.
The probability of outage, or outage rate, is defined as the fraction of connected users who are
not currently meeting their minimum QoS requirement. This provides a measurement of QoS
that can be continuously measured and updated, and that tends to increase smoothly as the
system approaches overload. Thus, action may be taken before significant numbers of calls are
blocked at the base station, reducing the impact on users of the network. This has advantages
over QoS measurement methods which rely on using the number of dropped calls as an
indication of QoS, because dropped calls usually only occur when the system has already
become heavily overloaded. Accordingly, it is difficult to detect the onset of overload, based on
the number of dropped calls, and thus avoid the system becoming heavily overloaded (when a
number of dropped calls are experienced).
Furthermore, blocking probability (or blocking rate) and QoS are both directly related to
user experience, so it would therefore be advantageous to use measurements of these quantities
to control the setting of p , such that adjustment of the threshold value depends directly on the
user experience of the network.
Base Station
Figure 4 is a block diagram illustrating the main components of each of the base stations
5 shown in Figure 1. As shown, each base station 5 includes transceiver circuitry 31 which is
operable to transmit signals to and to receive signals from the mobile telephones 3 via one or
more antennae 33 and which is operable to transmit signals to and to receive signals from the
telephone network 7 via a network interface 35. A controller 37 controls the operation of the
transceiver circuitry 31 in accordance with software stored in memory 39. The software
includes, among other things, an operating system 4 1 and a threshold calculation module 43, a
blocking rate calculation module 44, an outage rate calculation module 45, a load estimator
module 46 and an admission control module 47. The operating system 4 1 is operable to control
operation of the base station. The threshold calculation module 43 is operable to calculate a load
threshold value for use in admission control decisions in accordance with the invention.
Blocking rate calculation module 44 and outage rate calculation module 45 respectively calculate
a current blocking rate and a current outage rate, which are supplied to the threshold calculation
module 43 for use in calculating the load threshold. The load estimator module 46 is responsible
for determining load estimates for existing radio bearers and newly requested radio bearers and
for reporting these load estimates to the admission control module 47. The admission control
module 47 is responsible for using the load estimates determined by the load estimator module
46, along with the load threshold value calculated by the threshold calculation module 43 to
determine whether or not to admit a requested new radio bearer; for congestion control; or for
load balancing mobile telephones 3 between cells.
In the above description, the base station 5 is described, for ease of understanding, as
having a number of discrete modules (such as the threshold calculation module 43, the blocking
rate calculation module 44, the outage rate calculation module 45, the load estimator module 46,
the admission control module 47 etc). Whilst these modules may be provided in this way for
certain applications, for example where an existing system has been modified to implement the
invention, in other applications, for example in systems designed with the inventive features in
mind from the outset, these modules may be built into the overall operating system or code and
so these modules may not be discernible as discrete entities.
Threshold Adjustment Method
The basic mechanism of the threshold adjustment method is illustrated in Figure 2. This
mechanism is implemented in the threshold calculation module 43 to determine the threshold
load value to be used in admission control decisions.
Over a configurable measurement interval T (for example around 60 seconds), the
current blocking rate and outage rate are estimated by the blocking rate calculation module 44
and outage rate calculation module 45. If the measured blocking and outage rates lie in the
region labelled 'Blocking Region' in Figure 2, then the load threshold, p , is increased by a
small amount to try to reduce the blocking rate at the base station 5. If the measured blocking
and outage rates lie in the Outage Region' then the load threshold, p h , is decreased by a small
amount to try to reduce the outage. Over time, the system will tend to converge to a point
somewhere on the boundary between the Blocking Region and Outage Region. The exact point
to which it converges depends on the level of offered traffic, the offered traffic being the total
throughput that the network would have to provide in order to meet the requirements of all users
who want to access the system.
The boundary between the Blocking Region and Outage Region is controlled by three
parameters which can be set by the network operator.
The first parameter is This represents the maximum outage rate that can be
tolerated under any circumstances. If the measured outage rate is higher than P then the
load threshold, p t , is always decreased. The purpose of this parameter is to prevent the outage
rate from becoming too high if the offered traffic is very high. In very high offered traffic
conditions, the algorithm maintains the outage rate at P and allows the blocking to rise as
the offered traffic increases. P is a value between 0 and 1.
The second parameter is • f the measured outage rate is below this level, it is
assumed that the outage rate is low enough to be ignored, and the threshold, pl , is always
increased if blocking occurs. This is based on the assumption that when the offered traffic is low
and the system is lightly loaded, there is little or no reason to block any calls, so any blocking
event should result in the threshold being increased. P , is a value between 0 and 1 and is
smaller than
The last parameter is the blocking-to-outage ratio, . This allows the operator to control
the required balance between the blocking rate and outage rate. At each update, the algorithm
effectively compares Pblo i g with (Poulage - ¾ )and adjusts the load threshold, p,h,
according to which one is larger. Thus, reducing will reduce the blocking rate and increase
the outage rate, whereas increasing will increase the blocking rate and reduce the outage rate.
can be any non-negative value. This parameter reflects the fact that as offered traffic
increases, a higher outage rate may be tolerated, in order to reduce the blocking rate. If required,
can be set to zero, in which case it has no effect, and the algorithm will simply attempt to
minimise the blocking rate while keeping the outage rate below P max
outage
In the Outage Region the load threshold, pt , is only decreased if blocking is occurring,
unless P (m,k) ™ . This helps to avoid the threshold drifting towards zero when the
offered traffic load is very low.
The threshold adjustment method is illustrated in the flow chart of Figure 3. The method
shown in the flow chart is executed at the end of each measurement interval. P (k) and P (k)
are respectively the estimated outage and blocking rate at the end of the k measurement
interval. pl (k) is the load threshold at the start of the k measurement interval.
and p down are step-size parameters which control the rate of adjustment. These
parameters are both greater than zero.
and ™" are parameters which set respectively the maximum and minimum
allowed values of pt . These parameters may be used to restrict the operating range of the
threshold adjustment algorithm.
As illustrated in Figure 3, execution of the threshold adjustment method starts with a
determination of whether a measured outage P {k) rate determined by outage rate calculation
module 45 is greater than the maximum allowable outage rate P , at decision block 22. If it
is determined that P0 >P , then execution proceeds to block 34 where the load threshold is
decremented by p , else execution proceeds to block 24 to determine if the blocking rate
determined by blocking rate calculation module 44, Pb (k), is greater than zero (i.e. if any
blocking is occurring). If no blocking is occurring execution proceeds to block 30 and no change
is made to the current load threshold value. If blocking is occurring execution proceeds to block
26. A determination is then made as to whether P (k) P , if so execution proceeds to block
32 and the threshold is incremented by Sp tip , else a final comparison is made of, Pb (k) >
o l g - , ) • greater than • (P ul g - P ), then execution proceeds to block
32 to increment the threshold, and if P (k) is not greater than • Poutage - P ™ ) e execution
proceeds to block 34 to decrement the threshold by Spjow„. Finally, a check is made, in block 36,
to ensure that the adjusted load threshold remains within the desired range, and if not the load
parameter is adjusted to or as appropriate.
Blocking Rate Calculation
Measurement of the blocking rate Pb(k) can be achieved by simply measuring the
number of blocked admission requests made during a measurement interval and comparing the
number of blocked requests with the total number of admission requests. More formally, if
N (k) is the total number of AC admission requests made during the k measurement interval,
and Nb (k) is the number of these attempts that were blocked by the admission control module
47. Then Pb(k) is calculated in the blockage rate calculation module 44 as follows:
N (k) =
P (k)=
b k )
otherwise
Note that Pb(k)= if Nr ( )=0 (i.e. if there are no admission control requests).
Outage Rate Calculation
The outage rate is defined above as the fraction of bearers that are currently not achieving
their Quality of Service (QoS) requirement. One way of expressing (or approximating) the QoS
requirement is as a required data rate, Rreq , such that a bearer is considered to be in outage if the
throughput it is currently achieving is lower than Rr q .
One potential problem with measuring outage in this way is that it depends on the
measurement interval. For example, suppose a call lasts for 60 seconds. During the first 30
seconds, the user achieves 95% of the required rate. During the final 30 seconds, the user
achieves 10% of the required rate. If the outage is measured, say, every 10 seconds, then this
user would be in outage for 50% of the time. But if the outage is measured every 60 seconds
then this user will not be in outage at all, because the average achieved rate over the call is
greater than the required rate. In general, the measured outage is likely to decrease as the
measurement period is increased, because there is more chance that the throughput will average
out to a value greater than the required rate. There is no 'correct' period for measuring outage.
However, one way to choose the measurement period is to consider how long it would take for
the outage to become noticeable to the user. In the case of a streaming service, for example, this
might be of the order of a few seconds. It will be appreciated by the skilled practitioner that
different measurement periods would be appropriate depending on the type of service offered.
Two possible methods of measuring outage and calculating an outage rate in the outage
rate calculation module 45 are now described which are particularly suitable for use in wireless
communications standards implemented in accordance with the LTE standard.
Direct Throughput Measurement
Suppose that during each outage measurement period of duration AT0 seconds, the
following quantities are collected for each bearer:
S, The total number of bits delivered for bearer i.
The number of sub-frames in which bearer i has data to send (i.e. for which the
queue is not empty).
A bearer is considered to be in outage if , < , where re is the required bit rate in
bits/sub-frame.
Let P0 («) be the fraction of bearers which are in outage during the n outage
measurement period. P0 («) may be block-averaged over the threshold adjustment period ATlh to
obtain P (k) as follows.
(3)
=o
where N =
Alternatively, instead of block averaging, a continuous exponential averaging can be
applied, i.e. P (k) = (1 - ) •P (k - 1) + P0 (n) , where 0 < 1 is an averaging weight.
(Note that the averaging over AT smoothes the estimate, but it is not equivalent to
extending the outage measurement period, because decisions about whether a bearer is in outage
are still made every AT seconds).
A disadvantage of this scheme is that it requires memory to store S, , a J
each bearer.
From Scheduler Throughput Measurement
In an LTE system, the allocation of radio resources to individual bearers is controlled by
the MAC scheduler. Typically the MAC scheduler will employ a Proportional Fair (PF)
scheduling algorithm, or a derivative thereof.
A PF-based scheduler maintains a throughput estimate (t) for each bearer, which is
updated every sub-frame, typically by exponential averaging.
In any sub-frame, we may consider a bearer to be in outage if T t ) 0 and 0 < C < 1 are configurable parameters (see Figure 5).
Another possible enhancement concerns the step-size parameters and Sp m . In
order to recover quickly from overload situations, it may be advantageous to use a larger value of
Pdown when outage is high. Similarly, at low offered load the threshold may increase slowly
because blocking events are rare, so it may be advantageous to use a larger value of when
outage is low. Some possible solutions are as follows:
• Use a larger value of p
down when outage is high, e.g. when P (k)>P .
• Use a larger value of 8p
p
when outage is low, e.g. when P (k)

oulag
WO 2012/043869 PCT/JP2011/072910
where Pb(k) is the determined blocking rate, P u ge is the obtained outage probability, P ™ is
the minimum desired outage rate and is the blockage-to-outage ratio, and to decrease the load
threshold if,
8. The communications apparatus of any of claims 4 to 7 wherein at least one of said first
and second values are calculated in dependence on the obtained outage rate.
9. The communication apparatus of any preceding claim, wherein the means for obtaining
the blocking rate, P (k), is configured to measure a number of blocked admission
requests, Nb k ) , and a total number of transmission requests, Nr (k), during a measurement
interval, wherein the blocking rate is determined using:
N (k) =
otherwise
Nr (k)
10. The communication apparatus of any preceding claim, wherein the means for obtaining
the outage rate, P0(k) , is configured to block-average a measured fraction of users, P0 n) , not
currently meeting minimum quality of service requirements during an outage measurement
period, AT , over a threshold adjustment period, ATh , using:
where N =
AT
1. The communication apparatus of any preceding claim, wherein the means for obtaining
the outage rate, P (k) , is configured to perform continuous exponential averaging of a measured
WO 2012/043869 PCT/JP2011/072910
fraction of users, P0 (n) , not currently meeting minimum quality of service requirements during
an outage measurement period, AT , using
P0 {k) = - 0 ) -P0 {k -\)+P P0 n)
where 0 < 1 is an averaging weight.
12. The communication apparatus of any preceding claim, wherein the means for obtaining
the outage rate, P0 (k) , is configured to determine the outage rate using
where Tf (t) is a throughput estimate for each bearer which is updated every subframe, R "q is a
required bit rate for each bearer, B(t) is the number of bearers present in sub-frame t, and the
indicator function l(x) is defined
13. The communication apparatus of any preceding claim, wherein the node comprises a base
station.
1 . The communication apparatus of any preceding claim, wherein the network comprises a
mobile communication network.
15. The communication apparatus of claim 14, wherein the mobile communication network
comprises a network compliant with the Long Term Evolution (LTE) standard.
16. A method of adjusting a load threshold for use in call admission control at a node in a
network, comprising:
obtaining an outage rate for one or more current calls at the node;
WO 2012/043869 PCT/JP2011/072910
obtaining a blocking rate of admission requests at the node; and
adjusting the load threshold in dependence on the outage rate and the blocking rate.
17. A computer implementable product comprising computer implementable instructions for
causing one or more programmable computer devices to become configured as the
communications apparatus of any of claims 1 to 15.