U.S. patent application number 10/550229 was filed with the patent office on 2006-11-09 for method for adapting link weights in ralation to optimized traffic distribution.
Invention is credited to Gero Schollmeier, Christian Winkler.
Application Number | 20060253607 10/550229 |
Document ID | / |
Family ID | 34745206 |
Filed Date | 2006-11-09 |
United States Patent
Application |
20060253607 |
Kind Code |
A1 |
Schollmeier; Gero ; et
al. |
November 9, 2006 |
Method for adapting link weights in ralation to optimized traffic
distribution
Abstract
The invention relates to a method which is used to adapt link
weights or, link costs, for optimized traffic distribution within a
communication network. According to said method, a loop is
circulated until an interruption criterion is met. The individual
iterations comprise the following steps: Routing of traffic within
the communication network is calculated based on link cost paths.
Parameter values used for optimization e.g. the link-related volume
of traffic, are determined for the individual links by means of the
calculated path, and an expected volume of traffic and the link are
determined for the parameter having the highest value. When the
highest determined value is equal to or less than the valve of the
previous step (interruption criterion), the link costs are
increased for the determined link. Said method enables link costs
and paths for optimized traffic distribution to be determined in an
economical manner.
Inventors: |
Schollmeier; Gero; (Gauting,
DE) ; Winkler; Christian; (Munchen, DE) |
Correspondence
Address: |
Siemens Corporation;Intellectual Property Department
170 Wood Avenue South
Iselin
NJ
08830
US
|
Family ID: |
34745206 |
Appl. No.: |
10/550229 |
Filed: |
January 14, 2005 |
PCT Filed: |
January 14, 2005 |
PCT NO: |
PCT/EP05/50149 |
371 Date: |
September 20, 2005 |
Current U.S.
Class: |
709/241 |
Current CPC
Class: |
H04L 45/12 20130101 |
Class at
Publication: |
709/241 |
International
Class: |
G06F 15/173 20060101
G06F015/173 |
Foreign Application Data
Date |
Code |
Application Number |
Jan 30, 2004 |
DE |
10 2004 004 793.6 |
Claims
1.-8. (canceled)
9. A method for adapting link weights in a communication network
formed with links for an optimized traffic distribution within the
communication network in respect of a traffic volume expected for
the communication network and in relation to a parameter relating
to the link usage, the method comprising: a) assigning start values
for the link weights to the links of the communication network; b)
calculating paths for routing of traffic within the communication
network on the basis of the link weights; c) determining the values
of the parameter for the links using the calculated paths and the
expected traffic volume; d) determining such link for which the
parameter has the highest value; and e) increasing the link weight
for the determined link, wherein f) the steps b), c), d) and e) are
repeated until the value of the parameter for the determined link
is higher than the parameter value for the determined link relative
to a preceding step d), wherein the steps a) to f) are performed
with the aid of a data processor.
10. The method in accordance with claim 9, wherein the parameter is
produced by an absolute traffic load, a relative traffic load
related to link bandwidth traffic-related costs arising for using
the link, a link availability, a delay time of the relevant link,
or the load capabilities of an end node of the relevant link.
11. The method in accordance with claim 9, wherein the start values
for the link weights are the same for all links.
12. The method in accordance with claim 10, wherein the start
values for the link weights are the same for all links.
13. The method in accordance with claim 9, wherein the paths are
calculated using the OSPF (open shortest path first) protocol or
the IS-IS (Intermediate System-Intermediate System) protocol.
14. The method in accordance with claim 10, wherein the paths are
calculated using the OSPF (open shortest path first) protocol or
the IS-IS (Intermediate System-Intermediate System) protocol.
15. The method in accordance with claim 11, wherein the paths are
calculated using the OSPF (open shortest path first) protocol or
the IS-IS (Intermediate System-Intermediate System) protocol.
16. The method in accordance with claim 9, wherein the paths
calculated if the abort criterion f) is fulfilled are used for
routing within the framework of the ECMP (Equal Cost Multi Path)
concept.
17. The method in accordance with claim 10, wherein the paths
calculated if the abort criterion f) is fulfilled are used for
routing within the framework of the ECMP (Equal Cost Multi Path)
concept.
18. The method in accordance with claim 11, wherein the paths
calculated if the abort criterion f) is fulfilled are used for
routing within the framework of the ECMP (Equal Cost Multi Path)
concept.
19. The method in accordance with claim 13, wherein the paths
calculated if the abort criterion f) is fulfilled are used for
routing within the framework of the ECMP (Equal Cost Multi Path)
concept.
20. The method in accordance with claim 9, wherein the link weights
and their increase are always expressed by whole numbers.
21. The method in accordance with claim 10, wherein the link
weights and their increase are always expressed by whole
numbers.
22. The method in accordance with claim 11, wherein the link
weights and their increase are always expressed by whole
numbers.
23. The method in accordance with claim 9, wherein the link weight
is increased for both directions of transmission of the link for
the link determined.
24. The method in accordance with claim 10, wherein the link weight
is increased for both directions of transmission of the link for
the link determined.
25. The method in accordance with claim 11, wherein the link weight
is increased for both directions of transmission of the link for
the link determined.
26. The method in accordance with claim 9, wherein the expected
traffic volume is described by the traffic matrix.
27. The method in accordance with claim 10, wherein the expected
traffic volume is described by the traffic matrix.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application is the US National Stage of International
Application No. PCT/EP2005/050149, filed Jan. 14, 2005 and claims
the benefits thereof. The International Application claims the
benefits of German application No. 10 2004 004 793.6 filed Jan. 30,
2004, both applications are incorporated by reference herein in
their entirety.
FIELD OF THE INVENTION
[0002] The invention relates to a method for adapting link weights
in a communication network formed from links for an optimized
distribution of traffic within the communication network as regards
the volume of traffic expected for the communication network and in
relation to a parameter relating to link usage.
SUMMARY OF THE INVENTION
[0003] The invention lies within the area of communication
technology and in particular deals with the transmission of data
over packet-oriented networks.
[0004] The optimization of data transmission over packet-oriented
networks, especially over IP (Internet Protocol)-based networks is
currently an important area of activity in the field of network
technologies. the transmission of data with different
quality-of-service requirements in particular being a prerequisite
for the provision of new services, such as for example the
transmission of real-time traffic over packet-oriented networks. To
optimize transmission as much as possible, different parameters,
such as delay, noise susceptibility or traffic distribution must be
taken into account.
[0005] The best possible distribution of the traffic routed over a
communication network is important, on the one hand in relation to
avoiding overload situations, on the other hand good traffic
distribution is the requirement for a good utilization of the
existing resources and especially the bandwidth available. The
traffic distribution within a communication network depends on the
routing within the network or the paths on which the traffic is
routed through the network. By far the most widespread routing
method for packet-oriented networks based on an IP protocol is what
is known as the OSPF (Open Shortest Path First) method or protocol.
In this method, for traffic to be routed between two nodes, the
shortest path in the sense of a metric between the two nodes is
defined. The path is normally determined with the aid of what are
known as link costs. The terms link weight or link weights which
are normally employed by specialists as synonymous with link costs
are also referred to below. The link costs or link weights are
values administratively assigned to the links, said values
representing a measure of how well the relevant link is being used.
In the simplest case the link costs are equal to 1 for each link,
i.e. all links are treated in the same way. For routing which
minimizes the link costs this method automatically determines the
shortest path between two nodes as regards the number of links or
the number of hops. Changing the link costs allows the
determination of the routes and thus the distribution of the
traffic within a communication network to be influenced.
Controlling the traffic distribution by adapting the link costs is
generally possible with routing methods which operate using a
metric and the allocation of link costs, e.g. in the IS-IS
(Intermediate System-Intermediate System) protocol or for example
the ECMP (Equal Cost Multiple) method, which was defined within the
framework of the OSPF protocol and makes provision for the use of a
number of paths determined by means of OSPF for routing between two
nodes.
[0006] Conventional methods for optimizing traffic distribution by
adapting the link costs are comparatively complicated. Usually the
link costs are optimized globally by optimizing a target function
which reflects the topology of the network. Because of the
complexity of this method the link costs are currently frequently
not optimized. Unavoidable adaptations of the link costs are
frequently performed manually, which entails a high risk of errors.
Thus the traffic optimization in real networks is likely to be less
than optimal.
[0007] The object of the invention is to specify a simple method
for determining the link costs in a communication network as
regards optimized traffic distribution
[0008] The object is achieved by the claims.
[0009] The thinking behind the invention is that the link costs or
link weights of heavily loaded links should be increased to reduce
the traffic load on these links. A parameter is used for the
optimization which represents the measure of the loading of the
individual links. Preferably this parameter is specified by the
traffic load of the individual links. Other options are the
relative traffic load in relation to the link bandwidth, any
traffic-dependent costs arising for link usage (possibly even in
the form of real charges arising), the availability of the links,
the run time on the individual links or the load-bearing
capabilities of the end nodes of the links concerned. The starting
point of the invention is that the volume of traffic to be carried
by the network is known or at least an estimate of said volume
exists. In traffic theory what is known as the traffic matrix is
usually used to specify the traffic volume more precisely. The
traffic matrix specifies the volume of traffic to be transported
between pairs of source and destination nodes. For traffic routed
through the communication network a source node for example
corresponds to an ingress node and a destination node to an egress
node, so that the traffic matrix features an entry in relation to
these two nodes for the traffic transported between them.
[0010] For optimization as regards the parameter the procedure is
as follows. First the link costs of the individual links are
initialized, i.e. start values are assigned to them. For the links
a distinction is preferably made between the two directions, i.e. a
connection between two nodes corresponds to two directed links,
each of which transports the traffic once from the first node to
the second and once in the other direction. The invention can then
be further developed so that, especially for symmetrical traffic
matrices, the optimization of the link costs is performed for both
directions simultaneously. After the assignment of start values for
the link costs paths for the routing of traffic within a
communication network are computed with the aid of the routing
protocol (for example OSPF or IS-IS). With the aid of the assumed
traffic volume the traffic carried on these paths and thereby the
traffic volume on the individual links can be calculated or
determined. The link is then determined for which the parameter,
for example the traffic volume, has the highest value. Subsequently
the value of the link costs for the link determined is increased,
by 1 for example. Calculating the paths, defining the values of the
parameter, determining the link for which the parameter has the
highest value and increasing the value of the link costs for the
link determined is performed as a loop which is run until such time
as the value of the parameter for the link with the highest
parameter value determined in a step is higher than the parameter
determined in the previous step or sequence for the link with the
highest parameter value. If this abort criteria is fulfilled, the
values for the link costs defined in the step before last and the
associated paths for routing within the communication network are
determined.
[0011] An improved or optimized traffic distribution is obtained
through the invention by increasing the link costs of heavily
loaded links. In comparison to conventional methods the inventive
method is simpler and can be implemented without any problems. For
determining paths within the framework of the ECMP (Equal Cost
Multiple) method it is useful to operate in the method with
whole-number link costs or whole-number increases in the link
costs, since in this way it is more likely that alternative paths
with the same accumulated costs for the connection of two nodes are
produced, so that, as provided for in the ECMP method, traffic can
be distributed to the various equivalent paths. As a start point it
is useful to initialize all link costs with the same value, for
example with 1.
[0012] The inventive method can be performed using central or
distributed computing capacity. For parts thereof, e.g. path
calculation, there can be recourse to conventional methods (which
for example are defined within the framework of the OPSF or IS-IS
protocol).
[0013] The object of the invention is explained in more detail
below within the context of an exemplary embodiment which refers to
a flowchart.
BRIEF DESCRIPTION OF THE DRAWING
[0014] The sole FIGURE shows an exemplary flowchart according to
the invention.
DETAILED DESCRIPTION OF THE INVENTION
[0015] The exemplary embodiment starts from the assumption that the
communication network is an IP network which is formed with nodes
and links and that paths are to be determined for routing within
the framework of the ECMP concept. To begin with the link costs LK
are initialized with the value 1. On the basis of these link costs
LK the ECMP paths P(LK) are then calculated. The calculation is
undertaken by determining the paths with the lowest accumulated
link costs LK for routing between two nodes, as is provided for in
the OSPF protocol.
[0016] Least-cost paths to a destination node can for example be
determined in the following way. All neighboring nodes of the
destination node are initially determined and these neighboring
nodes are assigned the link costs as node costs. This defines the
routing for this first ring of nodes around the destination node.
The procedure below is undertaken ring-by-ring, moving up from a
ring to the next higher ring in each case. A ring in this case is
specified by the fact that a specific number of links at least have
to be passed through to reach the destination node. The costs of
the individual paths are successively calculated ring-by-ring. In
this case in each step the path costs are made up of the sum of the
link costs and the node costs, with link costs being the current
costs of the outgoing link in each case and the node costs being
the actual costs of the relevant end node of this link in each
case. The costs of the paths calculated in this way are then
compared and only those paths are retained which have minimal path
costs (as is also provided for within the framework of the OSPF
concept). A number of paths with the minimum path costs are
frequently found which are retained for traffic distribution in the
sense of the ECMP concept. Other methods of determining least-cost
paths, for example the Dijkstra algorithm, can be used just as
well.
[0017] The above procedure can be used for re-determining the paths
as part of the iterations for which the inventive method makes
provision. After calculation of the paths P(LK) as a function of
the link costs, the traffic matrix VM is used to determine the
traffic volume V(L) of the individual links. The link L(Vmax,n)
with the highest or maximum traffic volume Vmax,n is identified and
the maximum traffic volume Vmax,n in this step is compared with the
maximum traffic volume Vmax,n-1 in the previous step If the maximum
traffic volume Vmax,n for the current step is less than the maximum
traffic volume Vmax,n-1 for the previous step, the link costs
LK(L(Vmax,n)) of the link L(Vmax,n) with the maximum traffic volume
Vmax,n are increased by 1 and the paths P(LK) are recalculated as a
function of the changed link costs LK. The loop is then executed
once more. Otherwise, i.e. if the maximum traffic load Vmax,n is
greater than in the previous step, the link costs LK determined in
the previous step are used as optimized link costs LKOpt and the
associated paths P(LK) are used as the paths P(LKopt) determined
for an optimized traffic distribution for routing.
* * * * *