MANAGEMENT OF CONFIGURATION CHANGES IN CLUSTERED NETWORK NODES
Field of the Invention
The present invention relates to telecommunications, in particular to
configuring nodes in a telecommunications network.
Description of the Related Art
5 Decentralised algorithms for self-configuring of nodes in networks suffer from
the risk that they may not converge to usable solutions. This is particularly so in large
networks having many interacting nodes. For example, in many situations in network
optimisation, the configuration of a network node, for example a base station for
cellular communications or an optical switch, is dependent on the configuration of a
10 neighbouring network node, and vice versa. This means that when a network node
changes its configuration, in the sense of changing a property or characteristic of the
network node, this triggers neighbouring nodes to change theirs, which causes their
neighbouring network nodes to change theirs, and so on. This is a problem that can
cause a lot of disruption in the network, and may even cause a catastrophic failure if
15 the disruption is severe and propagates throughout the network.
The behaviour of large networks of interacting nodes that are distributed in the
sense of lacking central control, cannot be predicted precisely. This because such
systems are complex and accurate analysis is difficult or not possible. Furthermore,
the information available to individual nodes is limited.
20 Conversely, in systems having central or centralised control, a central control
has a good picture of the overall system and can therefore decide on a good network
configuration, for example as regards implementing self-configuring algorithms in
network nodes, and so implement that configuration in a well-controlled manner.
However, centralised approaches suffer scalability issues in that they are difficult to
25 apply to a large scale network, for example to a rapidly changing network with a large
number of nodes, such as femtocell deployments.
In this text, femtocell base stations are sometimes referred to as femtos.
Summarv
The reader is referred to the appended independent claims. Some preferred
features are laid out in the dependent claims.
An example of the present invention is a method of configuring nodes of a
telecommunications network, in which nodes react to changes in configuration of at
least one of their respective neighbour nodes. The method includes the steps of:
identifying a cluster of neighbouring nodes,
identifying which nodes in the cluster are in a frontier region adjacent to
another cluster,
adapting the configuration of nodes in the frontier region in response to
changes in the configuration of other nodes in the frontier region,
adapting the configuration of nodes in the cluster in response to changes in the
configuration of other nodes in the cluster whilst considering the configuration of the
nodes in the frontier region as set.
Preferred embodiments of the present invention partition the network into
clusters and implement configuration of the nodes in the boundary region. The nodes
in the boundary region limit, to within a cluster, the propagation of changes due to
each node adapting its setting in response to a change in a corresponding setting of a
neighbouring node. In other words, by dividing the network up into clusters with
boundary regions, oscillations and failures may propagate through a cluster but be
prevented from continuing into a further cluster. In some embodiments, as the size of
clusters is known, hardware can be selected appropriately to perform the
computational tasks involved.
Brief Description of the Drawings
Embodiments of the present invention will now be described by way of
example and with reference to the drawings, in which:
Figure 1 is a diagram illustrating known propagation of cell coverage changes
through a distributed cellular network (PRIOR ART),
Figure 2 is a diagram illustrating known oscillations in cell coverage
optimisation in the distributed cellular network shown in Figure 1 (PRIOR ART),
Figure 3 is a diagram illustrating a network of clusters of cells according to a
first embodiment of the invention,
Figure 4 is a diagram illustrating in-cluster optimisation of cell coverage in the
network shown in Figure 3,
Figure 5 is a diagram illustrating cell coverage in the network at a later stage in
which frontiers between clusters start to be identified,
Figure 6 is a diagram illustrating cell coverage in the network at a later stage at
which frontiers between clusters become outlined,
Figure 7 is a diagram illustrating cell coverage in the network at a later stage at
which frontiers become better defined by boundary optimisation,
Figure 8 is a diagram illustrating the resultant cell coverage in the network,
Figure 9 is a diagram illustrating the cell coverage optimisation process from
the perspective of an individual cell in the network shown in Figure 3,
Figure 10 is a diagram illustrating a network according to a second
embodiment of the invention in which clusters overlap.
Figure 11 is a diagram illustrating a network according to a third embodiment
of the invention, in which some cells of a cluster, those cells forming an inner and
boundary region, lie within another cluster, and
Figure 12 is a diagram illustrating a network according to a fourth embodiment
of the invention, in which a few cells of a cluster, those cells forming a boundary
region only, lie within another cluster.
Detailed Description
As an example of network node optimisation, the inventors considered cell
coverage optimisation specifically in femtocell deployments, in other words where the
netwrk nodes are femtocell base stations. The inventors considered cell coverage
optimisation as an example because it is easy to visualise. Of course, other properties
or attributes of nodes may be optimised in addition or instead.
The inventors considered the known approaches to cell coverage optimisation
in femtocell deployments. Here the coverage of a femtocell is adjusted to achieve
objectives such as load balancing, minimising interference, and preventing coverage
holes. This is done by changing transmit power of pilot channels and changing the
base station antenna configuration. The inventors realised that in known systems
where distributed algorithms are used, a change in one part of the network may
propagate throughout the network, as shown in Figure 1 (PRIOR ART). Figure 1
shows three neighbouring femtocells (A,B,and C) at four sequential instances in time
(steps i, ii, iii, and iv). For example, as shown in Figure 1, FemtocellA decreasing its
coverage (step ii) causes its neighbour FemtocellB to increase its coverage (step iii) in
consequence so as to fill a coverage gap. This in turn causes FemtocellC to decrease
(step iv) its coverage area. Such changes in coverage can lead to temporary conditions
where quality of service is reduced and control signalling (as opposed to user traffic)
is increased.
There is also a risk that unstable oscillatory behaviour may occur, where the
network fails to converge to a stable configuration. For example, as shown in Figure 2
(PRIOR ART) two neighbouring femtocells (here denoted D,E,) can get in a loop
where each alternately expands and contracts its coverage area to adapt to the
coverage area of the other, but a stable coverage area configuration is not achieved.
Such disruptions are undesirable.
Turning now to an embodiment of the invention, we again consider coverage
optimisation as an example because it is easily visualised. We consider how femtocell
having backhaul connections to a common Digital Subscriber Line Access Multiplier
(DSLAM) can be consider as a cluster, how each cluster optimises cell coverage
within the cluster ("inner optimisation"), how frontiers between clusters are identified,
and how femtocells at the frontiers are optimised in their coverage ("boundary
optimisation"). By the use of boundary optimisation, cell coverage areas of femtocells
at the frontiers become fixed such that disruptions and perturbations in cell coverage
areas of femtocells within a cluster are contained within that cluster.
Clustering of femtocells
As shown in Figure 3, femtocell base stations (a few of which are denoted 2
for ease of understanding) are each connected via respective Digital Subscriber Line
(xDSL) connections, in other words, Internet backhaul connections, to a Digital
Subsciber Line Access Multiplier (DSLAM) shared with other femtocells. All the
femtocells connected to the same DSLAM form a cluster. In Figure 3, four such
clusters of femtos, denoted ClusterA, ClusterB, ClusterC and ClusterD are shown, a
femto in clusterA is shown as a black triangle, a femto in clusterB is shown as a
white-centred circle, a femto in clusterC is shown as a black square, and a femto in
clusterD is shown by a black circle. The DSLAMs are connected together in a
Metropolitan Access Network (MAN) that includes a central office 4 having backbone
connection 6 to the rest of the telecommunications world (not shown). Information is
exchanged between DSLAMs via the MAN.
Optimisation within clusters ("Inner optimisation")
As shown in Figure 4, initially each cluster independently starts its own femto
coverage area optimisation process. In this example, the optimisation process uses
genetic programming, as is known from, for example, the paper by Ho L T W, Ashraf
I, and Claussen H entitled "Evolving Femtocell Coverage Optimisation Algorithms
Using Genetic Programming" in Proc. IEEE PIMRC 09, September 2009, and more
generally the book by John Koza "Genetic Programming: On the Programming of
Computers by Means of Natural Selection, Press, 1992.
In some other, otherwise similar, embodiments (not shown) alternative
methodologies to genetic programming are used, such as Reinforcement Learning and
Neuro-fuzzy logic. In some embodiments (not shown) two or more methodologies are
used in combination.
As shown in Figure 4, initially each cluster has no knowledge of its respective
neighbours so a frontier between clusters is not defined, and hence can be considered
as merely virtual. The optimisation within each cluster is aimed at maximising overall
coverage within each cluster area and identifying the frontiers, as explained in relation
to Figures 5 and 6 below.
Defining frontiers between clusters
As shown in Figure 5, each femto 2 has an associated femtocell coverage area,
a few of which are indicated by the reference numeral 10 for ease of understanding.
In maximising the coverage within each cluster, adjacent clusters will overlap in their
coverage area so as to define a respective frontier 12. This can be seen by comparison
of Figures 5 and 6, where optimisation of coverage areas in clusterA and ClusterB has
the effect of filling the area 8 in Figure 5 where there is no coverage, so as to give the
coverage shown in Figure 6 including defining the frontier between clusterA and
clusterB.
The frontier definition process is by feedback information from mobile user
terminals. When a mobile user terminal senses it is in the overlapping coverage area
of two overlapping femtocells but those two femtocells are connected to different
Digital Subscriber Line Access Multipliers (DSLAMs), in other words the two femtos
are in different clusters, then the mobile user terminal informs the two femtos of this
situation. The two femtos, in turn, each forwards this information to its respective
DSLAM, where the information is used to update a database table identifying femtos
at the frontier.
In some situations some of the femtos at the frontier are known in advance.
Femtos identified at a frontier are held steady in their coverage areas
The DSLAM of each cluster removes the femtos that are identified as being at
along frontier from the within cluster cell coverage optimisation process. In this
within-cluster process, their coverage areas are then considered to be steady rather
than variable. Accordingly when a change or perturbation in cell sizes propagates
through a cluster, these femtos at the frontier have steady coverage areas so act to
inhibit or prevent the change or perturbation moving into a neighbouring cluster.
Coverage areas of femtos identified at a frontier are optimised ("Boundary
Optimisation")
As shown in Figure 7, the cells along the frontier with a neighbouring cluster,
are optimised as to their cell coverage without further considering the rest of the
femtos in their respective clusters. This is to provide maximum coverage in the
frontier region only. For the purposes of illustration dashed lines is shown merely to
indicate the frontier region edges, such femtos between a frontier (solid line) and
frontier region edge (dashed line) are considered at the respective frontier for cell
coverage optimisation purposes.
In this example, the optimisation process uses genetic programming, as is
known from, for example, the paper by Ho L T W, Ashraf I, and Claussen H entitled
"Evolving Femtocell Coverage Optimisation Algorithms Using Genetic
Programming" in Proc. IEEE PIMRC 09, September 2009, and more generally the
book by John Koza "Genetic Programming: On the Programming of Computers by
Means of Natural Selection, MIT Press, 1992. In some other, otherwise similar,
embodiments (not shown) alternative methodologies to genetic programming are
used, such as Reinforcement Learning and Neuro-fuzzy logic. In some embodiments
(not shown) two or more methodologies are used in combination.
This optimisation process is performed by the Digital Subscriber Line Access
Multiplier (DSLAM) elected to do that for each frontier. In an alternative embodiment
(not shown), this process can be performed in a distributed manner by the relevant
femtos. In another alternative embodiment (not shown) this optimisation process is
performed by an external entity, for example a computational element.
The result of boundary optimisation is shown in Figure 8 for comparison with
Figure 7. It will be seen for example that coverage gaps along the frontier have been
closed.
Relationship between optimisation within cluster and optimisation along frontier
The above described processes of optimisation within cluster and optimisation
along frontier are basically independent so that propagations of change a through a
cluster is stopped by the femtos along its frontiers from continuing into other clusters.
This means any disruptions are limited to within one cluster so limiting its effect.
In this example, changes in frontier cell coverage affect within-cluster
coverage, but not vice versa.
Converging to an overall cell coverage solution
Both processes of optimisation within cluster and optimisation along frontier
are independent in the sense that both are continuously seeking to optimise the
coverage areas of femtos to provide an overall best solution. Of course, this approach
is able to react to topology changes, for example as new femtos are introduced or are
switched on or off.
As each cluster is defined by the Digital Subscriber Line Access Multiplier
(DSLAM) to which its femtos are connected, the maximum number of femtos that
may be in the cluster is known in advance. Accordingly, the maximum number of
femtos in the frontier region is limited as is the computational complexity in reaching
a convergent solution. This means the computational hardware for the inner and
boundary optimisations may be optimised in terms of speed, power consumption, size
etc for these processes at the scale of the numbers of femtos involved. In this example,
the computational hardware is located in the DSLAMs. In an alternative distributed
approach (not shown), the hardware is distributed in the femtos. In a further
alternative embodiment, the hardware is in an external entity (e.g. a computational
element).
The processes from the perspective of an individual femto
To further explain the above mentioned approach, let us consider an individual
femto within a cluster. As shown in Figure 9, upon booting up (step a) , the femto is
automatically included in the inner optimisation process (step b) which deals with
power hence coverage area of femtos within the cluster. The femto is given a status of
"normal" for this purpose.
A query is then made as to whether (step c) the femtocell is in a frontier
region. This query is made upon every algorithm iteration (In an alternative
embodiment, the query could be made every time frame). Upon the femto being
discovered (step d) as being in a frontier region, because of a notification message
from a mobile user terminal to that effect, or a procedure of neighbour femto
discovery, then the status of the femto is changed (step e) to "frontier".
At this point, although the "frontier"-status femto is still a part of the inner
optimisation process (step h) its power level and hence coverage area is set (step g)
for that purpose as being steady. On the other hand, the "frontier"-status femto is
included in the boundary optimisation process (step f), which does not consider the
non-frontier region femtos.
The effect of introducing frontiers that prevent all nodes in a network from
being reconfigured when a change occurs in one cluster may mean that a theoretical
optimum global coverage configuration may not be achievable in consequence. In
some embodiments (not shown) the deviation from this ideal may be measured or
evaluated and may be used as a parameter in clustering and optimisation methods.
Variants
In the above examples, Digital Subscriber Line Access Multipliers (DSLAMs)
are used to coordinate the identification of boundary region femtos. An alternative is
to instead do this in a distributed manner. Another alternative is for an external entity,
for example, a computational element, to do this.
In the above example, femtos were considered as clustered by being
connected to the same DSLAM. In other embodiments, other groupings are possible,
such as grouping femtos according to their paging area codes.
The order in which boundary optimisation and within cluster optimisation are
undertaken may depend on the given scenario and constraints. For example, in some
other embodiments, for example if femtos are topology-aware such that the femtos at
frontiers are identified without within cluster optimisation, then boundary
optimisation is performed before within cluster optimisation. In some other
embodiments (not shown) for example in critical applications where femtos need
setup times that are minimised, the boundary optimisation and within cluster
optimisation are performed in parallel.
As shown in Figure 10, in a slightly different scenario to that described with
reference to Figures 3 to 8A, clusters can slightly overlap. This may occur in, for
example, in real xDSL deployments with regions served by different Digital
Subscriber Line Access Multipliers (DSLAMs) somewhat overlapping. The method as
described in relation to Figures 3 to 8 applies except that a thicker frontier region
results.
As shown in Figures 11 and 12, some other scenarios are where some
femtocell base stations belong to a first cluster, lie within a second cluster, detached
from the main first cluster.
As shown in Figure 11, a substantial number of femtos from clusterA lying
within clusterB can be considered as an "island" having both cells in an inner region
14 and in a frontier region. This is where the inner region is identified for example
from information std in a topology record table in the network. For femtos in the inner
region 14 a within-region optimisation (within-cluster type optimisation) is
performed. For femtos in the frontier region 16, a boundary optimisation is
performed.
As shown in Figure 12, a smaller number of femtos from clusterA lying within
clusterB can be considered as an "atoll" having a frontier region 18 only . For femtos
For femtos in the frontier region 18, only boundary optimisation is performed.
The present invention may be embodied in other specific forms without
departing from its essential characteristics. The described embodiments are to be
considered in all respects only as illustrative and not restrictive. The scope of the
invention is, therefore, indicated by the appended claims rather than by the foregoing
description. All changes that come within the meaning and range of equivalency of
the claims are to be embraced within their scope.
A person skilled in the art would readily recognize that steps of various abovedescribed
methods can be performed by programmed computers. Some embodiments
relate to program storage devices, e.g., digital data storage media, which are machine
or computer readable and encode machine-executable or computer-executable
programs of instructions, wherein said instructions perform some or all of the steps of
said above-described methods. The program storage devices may be, e.g., digital
memories, magnetic storage media such as a magnetic disks and magnetic tapes, hard
drives, or optically readable digital data storage media. Some embodiments involve
computers programmed to perform said steps of the above-described methods.
Claims:
1. A method of configuring nodes of a telecommunications network, in which nodes
react to changes in configuration of at least one of their respective neighbour nodes,
the method including the steps of:
identifying a cluster of neighbouring nodes,
identifying which nodes in the cluster are in a ontier region adjacent to another
cluster,
adapting the configuration of nodes in the frontier region in response to changes in the
configuration of other nodes in the frontier region,
adapting the configuration of nodes in the cluster in response to changes in the
configuration of other nodes in the cluster whilst considering the configuration of the
nodes in the frontier region as set.
2. A method according to claim 1, in which said adapting the configuration of nodes
in the frontier region and said adapting the configuration of nodes in the cluster occur
in parallel.
3. A method according to claim 1 or claim 2, in which the nodes are cellular base
stations.
4. A method according to claim 3, in which the characteristic of the base station
configuration being adapted is cell size.
5. A method according to claim 3 or 4, in which the identifying which nodes in a
cluster are in a frontier region is from information of locations of mobile user
terminals that are in a region of coverage overlap of base stations in different clusters.
6. A method according to any of claims 3 to 5, in which each base station of a
cluster is connected to a shared backhaul node, in particular a Digital Subscriber Line
Access Multiplier, associated with the respective cluster.
7. A method according to any preceding claim, in which said method of configuring
nodes is undertaken in said another cluster also.
8. A method according to any preceding claim in which said adapting the
configuration of nodes in the frontier region occurs continuously and/or iteratively.
9. A method according to any preceding claim in which said adapting the
configuration of nodes in the cluster to the adapted configuration of other nodes in the
cluster occurs continuously and/or iteratively.
10. A telecommunications network comprising nodes, in which, the nodes are
configured to react to changes in a characteristic of their respective neighbour nodes,
the network being configured to:
identify a cluster of neighbouring nodes,
identify which nodes in the cluster are in a frontier region adjacent to another cluster,
adapt the characteristics of nodes in the frontier region in response to changes in the
characteristics of other nodes in the frontier region,
adapt the characteristics of nodes in the cluster in response to changes in the
characteristics of other nodes in the cluster whilst considering the characteristics of
the nodes in the frontier region as set.
1 . A telecommunications network according to claim 10, configured to conduct in
parallel said adapting the characteristics of nodes in the frontier region and said
adapting the characteristics of nodes in the cluster.
12. A telecommunications network according to claim 10 or claim 1lin which the
nodes are cellular base stations and the characteristic of the nodes that is being
adapted is cell size.
3. A telecommunications network according to claim 12, in which the nodes are
femtocell base stations.
1 . A telecommunications network according to claim 10, in which the network is
configured to adapt the characteristic of nodes in response to their neighbour cells, in
said another cluster also.
15. A telecommunications network according to any of claims 10 to 14, in which
said network is configured to adapt, in a continuous and/or iterative manner, the
characteristics of nodes in the frontier region in response to the adjustment of the
characteristics of other nodes in the frontier region and the characteristics of nodes in
the cluster in response to the adjustment of the characteristics of other nodes in the
cluster.