GB2189111A - Telecommunications network and method - Google Patents

Abstract

A method of controlling a circuit-switched telecommunications network comprises calculating an implied cost to the operator of the network for each link in the network, and a surplus value generated by calls using each route through the network, equal to the revenue from the route less the implied costs of all the links in the route, to control a centralized or decentralized routing strategy, or to provide an indication of the surplus values for real time network management purposes, or to allocate extra capacity optimally to the network. The method utilizes recent information on traffic conditions, and is thus responsive to varying and uncertain traffic conditions. <IMAGE>

Description

SPECIFICATION Communications network and method The present invention relates to communications networks and methods. In particular it relates to the routing and management of traffic and the dimensioning of links in a circuit-switched network subjected to varying and uncertain traffic conditions.

With the introduction of new switching technologies in recent years, it is no longer necessary for telephone traffic (orvoice and data traffic on an integrated network) to be routed along a fixed source-destination route. Stored program controlled exchanges and common channel signalling have opened up the possibility of increased flexibility.

An important problem for a large and complex network is how the routes used affect the performance ofthe network. For example, it may be that two different routes can both be used to carry calls from a source node to a destination node, and that the proportions oftraffic offered to these routes can be varied. It is important to know how such variation affects the blocking of calls between these two nodes.

Potentially as important is the knock-on effect of such variation on the blocking of calls between other pairs of nodes of the network. Knock-on effects may greatly influence the overall performance ofthe network, and have previously made the construction of good routing strategies difficult.

In the national telecommunications network in the United Kingdom, for example, it is becoming increasingly difficu It to predicttraffic demands in the shortterm and the long term, and there is a need for a system for coping with local overloads and changes in demand arising, for example, from the introduction of new services, and from changes in pricing policy.

The present invention provides a circuit-switched communications network comprising nodes joined by links each comprising circuits, and an adaptive control system for selecting routesthroughthe network, each route using a circuit from one or more links, the system comprising: means for determining a value of thetrafficthroughthelinksandthrough all the routes through those links, and, from those traffic values, determining for each link, for each route through the link, the proportion of traffic currently through the link which is on that route. the said proportion corresponding to the ratio between the traffic value for that link and the traffic valueforthat route; means for determining for each linkthe probability that use of an additional circuitfrom that link will subsequently block a route th roug h that link due to the unavailability of a circuit in that link; means for determining, from the said traffic proportions and the said blocking probabilities a measure for each link ofthe notional costto the operator ofthe network which would be suffered by the anticipated use of a circuit ofthat link, and a measure for each route of the notional surplus value of operating that route, equivalent to a predetermined measure of benefit obtainable by operating the route less the sum of all the said measures of notional costsforthe link or all the links ofthat route, by performing an iteration to determine a currentvalue of the said notional cost from a previousvaluethereof and from the product ofthe said blocking probability and a weighted sum, forall routes through the network, ofthe said traffic proportions, the said weighted sum of the traffic proportions for each route being weighted in accordance with the said notional surplus value; and means responsive to the said notional surplus value measures and/orthe said notional cost measures to provide an output signal indicative thereof.

The term 'notional cost' as used herein is intended to encompass not only tangible costs, but also the intangible costs to the network operator which result from the likely blockage of a later call.

To assess the efficiency of any scheme which can provide a good performance to certain routes at the expense of a poor performance to other routes it is necessary to make comparisons between the value of different routes. In the method described below, the notional cost to the operator of the network which would be suffered by the anticipated use of a circuit of that link, the predetermined measure of benefit obtainable by operating the route, the notional surplus value for each route and the notional nett value for each route, are all expressed in terms of monetary cost and revenue, butthe method applies equally well where there are no revenue considerations, but instead penalties which are incurred when a call is blocked, with possible different penalties for different routes.The notional costs, benefits and values referred to above should therefore simply be viewed as quantitative assessments concerned with the relative valueto the network operator calls on different routes. For example, minimizing the num ber of attempted callsthat are blocked corresponds to maximizing the revenue generated when every successful call isvalued atone unit. Assigning to calls on different routes differing values ofthe predetermined measure of benefit then corresponds to minimizing a similarly weighted average of the total number of calls biocked.

The said means for determining traffic values and proportions is preferably decentralised. The said means for determining the measures of notional costs and notional surplusvalues is preferably also decentralised. The system thus preferably comprises arith metic processing means assigned to each node which is a potential source of a stream of signals to be transmitted along a route, the source node arithmetic processing means determining the said traffic values only in respect of those links and routes to which that source node is potentially connectable, and, from those traffic values, determining, by the iterative calculations, for each such link the said measure of notional cost and, for each such route, the said notional surplus value.

The said blocking probability is preferably a predetermined estimating function, preferably determined by addressing a memory containing a set ofvaluesfor the function for each of several values ofthetraffic through a link.

The said weighted sum is preferably weighted in accordance with the sum ofthe notional surplus value currently obtainable by using that route, and the previous value ofthe said notional costforthe link.

Preferably, the said traffic values aretime-averaged values, and are preferably determined by an iterative smoothing process, by repeatedly combining a proportion ofthe previous time-averaged traffic value with a corresponding proportion of a measured instantaneoustrafficvalue.

The said communications network is, in a preferred embodiment ofthe invention, a telephone network for transmitting telephone calls along the routes and generating revenue thereby.

Thetelecommunications network preferably comprises display means responsive to the said output signal to provide a visible indication to human operators ofthe networkofthe said notional surplus valuesforeach route and/or notional costs for each link. In a preferred embodimentofthe invention, the network comprises means responsive to the said output signal indicative of the notional surplus values to vary automatically the routing oftrafficthrough the network.

According to afurtheraspect,the invention provides a method of routing traffic in a circuit-switched communications network comprising nodes joined by links each comprising circuits, comprising determin ing values ofthe trafficthrough the links and through all the routes through those links and determining, from those trafflic values, for each link, for each route through the link, the proportion oftraffic currently through the link which is on that route the said proportion corresponding to the ratio between the traffic value for that link and the traffic value for that route:: determining for each linkthe probability that use of an additional circuitfrom that linkwill subsequently block a route through that link due to the unavailability of a circuit in that link; determining, from the said traffic proportions and the said blocking probabilities,a measureforeach linkofthenotional cost to the operator ofthe network which would be suffered by the anticipated use of a circuit ofthat link, and a measure for each route ofthe notional surplus value of operating that route, equivalentto a predetermined measure of benefit obtainable by operating the route less the sum of all the said measures of notional costs forthe link or all the links on that route, by performing an iteration to determine a currentvalue of the said notional cost from a previous value thereof and from the product ofthe said blocking probability and a weighted sum, for all routes through the network, ofthe saidtraffic proportion, the said weighted sum of the traffic proportions for each route being weighted in accordance with the said notional surplus value; and providing an output signal indicative ofthe said notional surplus value and/orthe said notional cost measures.

The said blocking probability is preferably a predetermined estimating function, preferably determined by addressing a memory containing a set of values for the function for each of several values ofthe traffic through a link.

Preferably, the weighted sum is weighted in accordance with the sum ofthe notional surplus value for that route and the previous value of the said notional costforthe link.

The time-averaged values of traffic along the links and routes are preferably determined by an iterative smoothing process by repeatedly combining a proportion of the previous time-averaged traffic value with a corresponding proportion ofthe measured instantaneoustrafficvalue.

In one form, the method is used to control a centralised or decentralised routing strategy, and-the said output signal indicative ofthe notional surplus value is used by circuit-switching means to switch the circuits to establish an optimal route dependent upon thesaidsurplusvalues. In anotherform,the method is used in real time network management, and the said output signal indicative of the notional surplus values is used by display means to provide a visible display for a human operator.In a further application ofthe invention, the said output signal is provided to means which, from values ofthe output signal accumulated over a substantial period oftime, provide an indication ofthe optimum ailocation of extra capacity, (eg. an extra circuitorcircuits in a link,oran extra link) in the network, in accordance with the said notional surplus values.

According to a still further aspect, the invention provides a method of allocating extra capacityto a network comprising nodes joined by links each comprising circuits, taking into account the entire pattern oftrafficflowthrough the network over a substantial period of operation, comprising determining time-averaged values of the traffic through the links and through all the routes th rough those links, and, from the said traffic values, determining by iterative calculations for each link a measure of the notional cost to the operator of the networkwhich would be suffered by the anticipated use of a circuit of that link, and, for each route, determining the notional surplus value which would be obtained by operating that route, equal to a predetermined measure of benefit obtainable by operating the route less the sum ofallthe said measures of notional costforthe linkor all the links ofthat route; integrating the said notional costvaluefor a linkwith respect two time overthe said substantial period of operation, to provide a measure representative of the increased benefit which could have been obtained by providing an additional circuit on that link; and allocating one or more extra circuits to that link dependent on the size ofthat increased benefit.

In orderthatthe invention may be better understood, preferred embodiments of the invention will now be described, by way of example only, with reference to the accompanying diagrammatiodraw- ings, wherein: Figure lisa representation of part of a circuitswitched telecommunications network comprising five nodes and seven interconnecting links; and Figure 2 is a graph ofthe predetermined estimating function Fk(y), as a function ofthe mean carried load yk(t) on link k having a maximum capacity Ck = 100.

A circuit-switched communications network comprises a plurality of nodes, for example those shown as A, B, C, D and E in Figure 1, intereonnected by a pluralityoflinks (forexample links 1 to 7 in Figure 1), each link having one or more circuits for carrying a telephone call. In use, a telephone call has a source node and a destination node which are to be connected by a route using one or more links. For example,asourcenodeAcould be connected to a destination node C by several alternative routes, for examplevia links(l,2), links(l,4, 7)orlinks(5, 3, 4, 7).

Limits may be imposed by a controlling system on the maximum number of links in a route.

If all the circuits on any link in a route are occupied, i.e., are ndt available, then that route is blocked. In the eventofa call being blocked on a route, the control system may decide to offer the callto an alternative route: a network controlled inthisway is known as a circuit-switched network with alternative routing.

First consider the scheme for the case where a call is offered to just one route, and the call is lost if any link along that route has all its circuits busy. Label the links of the system k = 1,2 and supposethat link kcomprises Ck circuits. Let a label r identify a route.

LetAkr = 1 or 0 according as route r passes through link kor not. Acall requesting route r is lost if on any link k with Akr = 1 there are no free cirnuits. Otherwise the call is connected and simultaneously holds one circuit on each link kwith Akr = 1 forthe holding period ofthe call. Acall carried on route rgenerates a predetermined measure of benefit, which may be a measure of the mean revenue.

Intelligence in the form of arithmetical processing ability is available for each linkkand each node. This intellegence may be located centrally, or it may be distributed over the nodes ofthe network; for example the data processing for route r might be carried out at the source node for calls on route r. The instantaneous traffic load at each link and each route is measured by conventional means; from this, the data processing means compute time-averaged traffic values, updating the time average at regular intervais,for example every five seconds or every minute oreveryten minutes.Signals representing instantaneous and/or time-averaged values of traffic, originating from local data processing means, eg. at nodes ofthe network, are then sent to any other data processing means which requires that information, preferably through the network itself using the common channel signalling capability.

Specifically, suppose that there is available for route ra measureXr(t) ofthecarried load in Erlangs on route roverthe unittime interval (t-1,t) and a measurey(t)ofthecarried load in Erlangs on linkk overthe same interval.These will bearthe relation

Time-averaged, moving estimates xr(t) and yk(t) respectively ofthe mean carried traffic on route rand link k may then be computed either by local or by central data processing means by the following two iterations:- xr(t+1) = (1 -a)xr(t) + aXr(t) (2) yk(t+1) = (l--a)yk(t)+ aYk(t) (3) The constant a is chosen to be between 0 and 1, its precise value reflecting a balance between accuracy of estimation and speed of the response. For example, ifthetime interval is five seconds then a might be 0.02, while if the time interval is one minuted then a might be 0.2.

A quantity Sr(t) may now be calculated, representing an estimated notional surplus value of a route r, by subtracting, from the mean revenue wr generated buy a call on route r,the notional cost ck of using a circuit on each link k used by the router:-

The notional surplusvaluesrofa source is the difference between: the measure of benefit wrthat would be generated by connecting a call along the route; and the expected loss of revenue, due to the increased blocking in the network, that connecting such a call would cause. The notional costokof a link is calculated from the volume oftraffic on, and the notional surplus value of, routes that pass through the link.The notional cost of a link measures the expected loss of revenue, due to increased blocking in the network, that use of that link would cause. The recursion (4) enables the notional surplus values s,to be determined using the further recursion given below, which is derivable from a statistical analysis of circuit-switched networks:

in which b is another constant chosen to be between 0 and 1, again its precise value reflecting a balance between accuracy of estimation and speed of re sponse; and in which Fk(yk(t)) is a predetermined function, based on the mean carried load yk(t), of the probability that use now of an additional circuit from link kill block a latercallthrough link k.An example of Fkfor a link of capacity Ck = 100 is given in Figure 2.

It is not essential thatthe calculation of Fk be exact: an approximation to the value of Fk may be sufficient.

Forexample, the curve shown in Figure2 may be approximated bya straight line or a series of straight lines.

The ratio xr(t)/yr(t) is the proportion of traffic through link kwhich is on routes, and is used to estimate the probability that if any later call through link k is blocked, then that call is on routes. Blocking a call on route rwill cause the loss ofthe measure of benefit wr less the notional costfor every link along router other than link k. But this difference is simply estimated by the expression (ck(t) + Sr(t)) occurring in equation (5).

It can be shown that the constant b can be chosen sothatthe recursion (5)forthe notional costs ck(t) converges, irrespective ofthe initial values assigned to ck, k = 1 to K. For example the recursion converges foranyvalue of b less that 1/K, where K isthetotal number of links in the network.

It should be noted that manyvariations of equa tions (2)- (5) can be constructed, utilizing essentially the same method. For example, the time interval for the update ofthe recursions (4) and (5) could be shorter than that used forthe moving average estimates (2) and (3). This would have the advantage of giving a faster responseto changing traffic conditions.

The computation of sr(t+ 1 ) involves only values of ck(t) on links kwhich route r passes through, and the computation of ck(t+ 1) involves only values of Sr(t) on routes rwhich pass through link L This allows the recursionsto be implemented in a decentralized manner, e.g. with local data processing means assigned specificallyto selected nodes or links and resposive only to traffic through those respective nodes or links.Output signals provided by those local data processing means, representative ofthe notional surplusvaluessr(t) and/orthe notional costs ck(t) may then be fed to data processing means at a source node, or ata common central location, for (i) determining which of several possible routes to selectfora call orforall calls from a source node; or (ii) displaying an indication of a nett expected revenue for each route, to human operators, or (iii) integrating values over a period of days, weeks or months for dimensioning studies; as described below in greater detail.

It remains to specifythefunction Fk. Let

Erlang's formula forthe loss probability of an isolated link of capacity C offered Poisson traffic at a rate v, i.e.

the probability of blocking a call of a single link of capacity C circuits offered traffic at th at rate. Then Fk(y) = v[E(v,C-1) - E(v,Ck)l (7) fixes the probabilitythat use now of an additional circuit from link kwill block a later call through link k, wherevisdefined by: y = v[1E(v,Ck)] (8) which matches the mean carried loan to a fictional offered loadvwhich wouid correspondto it (as derived, for example, in Keliy, F.P., "Reversibility and Stochastic Networks",J.Wiley,Chichester(1979).

For an account of some of the theory upon which the method is based see Kelly, F.P., "Blocking probabilities in large circuit-switched networks", Adv.Appl.

Prob 18(1986)).

Suppose then thatthe notional costs ck and the associated notional surplus values 5r have been estimated using the preceding scheme. Suppose also thatthe loss probability on routes, Lr, has also been estimated,forexample by a similar simple moving average scheme. Consider the network of Figure 1.

Then a call from nodeAto node Cwill generate a notional nettvalue, i.e. a nett expected revenue, of (1 - )Sr if offered to routes, where r is, say, the route (5,6,7). If thins quantity is negative for any route then that route should not be used: more revenue will be lost elsewhere in the network than can be generated by accepting calls on that route. With this proviso the trafficfrom nodeAto node Cshould be shared out between the different possible routes (possibly by cycling between these routes) so asto reflect the notional nett values (1Lr)sr. If one route has a higher value off (1 -Lr)srthan the others a greater share of the traffic should be directed to this route.The adjustment should be made gradually, since the effect of increasing the traffic offered to a route will be to push up the loss probability ofthe route and the implied costs ck along that route, and hence to reduce the product (1TLr)sr.

The adjustments can be made automatically, eg. by central data processing means or local data processing means situated at source nodes responsive to signals from local data processing means representing ck and 5r values, or else the notional costs ck, surplus values srand loss probabilities Lrcan be displayed in a network management centre to inform human operators and enable them to make appropriate routing decisions. For example, it may well be that grade of service as well as revenue generation is important. This can be handled efficiently by increasing the nominal parameters wrfor routes handling traffic which is suffering a poor grade of service, and allowing the system to adjust.This procedure makes explicitthetrade-offs between overall performance, as measured by revenue generation, and the grade of service on any particular route.

Afurther use of the notional costs ck is in dimensioning studies. If ck(t) is integrated, for exampie, by the central data processing means described immediately above, over a long period (of days, weeks or even months) it gives a measure ofthe increased revenue which would have been generated by an additional circuit on link k. This can be compared with the installation costs of additional capacityto help with expansion decisions.

Afurther use of the notional surplus values 5r is in the implementation oftrunk reservation schemes. A trunk reservation scheme operates as follows. If less than m circuits are free on a linkthen calls on priority routes are accepted by the link, but calls of nonpriority routes are not accepted. The value of m is typicallysmall, perhaps 5 or 10.The difficulty isto decide which routes through a linkshould be classified as priority routes. The surplus values 5r providea means for making this decision: rankthe routesthroughlinkkindecreasingorderoftheir surplus values, and classify those that appear at the top ofthe list as priority routes.

The invention can readily be applied to networks with alternative routing, i.e. those in which a call blocked on one route may attempt connection on one or more alternative routes.

Suppose the label rfixes a pair (r(1 ),r(2)) of routes between a source and destination node. Regard r(1) as the route first attempted by a call labelied r; if this route is blocked the route r(2) is attempted; if both routes are blocked then the call is lost. As before, assume a call with label rwhich is carried generates a mean revenue of wr. For simplicity assume here that the routes r(1 ) and r(2) do not have any links in common (the scheme to be described is easily extended to allow the routes to have links in common, orto allow any number of routes to be attempted).

Then replace equation (4) by

Equation (4a) is similarto the earlier equation (4) since a call attempting route r(2) which is blocked will be lost, as earlier. Equation (4b) has an additional term. This term arises since a call blocked on route r(1) is not necessarily lost: there is a probability 1 Lr(2)(t) that it will be carried on its second choice route r(2), generating there a surplus value of Sr(2)(t+ 1). As earlier, the notional costs ck and surplus values 5r estimated by this scheme can be used for various purposes. Additionally, they can now be used to decide on whetherto reroute blocked calls.For example, fSr(2) becomes negative while 5r(1) remains positive then the single route r(1) is preferred to the pairr = (r(1 ),r(2)): a call blocked on the first choice route r(1) should be discarded ratherthan offered to the second choice route r(2).

The same principles apply equally to networks where more than one attempt is madeto reroute a call: the notional surplus value sr(n) for each route other than the last choice route depends on the loss probabilities of all subsequently chosen routes (in a similarwayto that in which sr(1 ) depended on Lr(2) in equation (4b)).

Whilstthe invention has been described with reference to particular characteristics ofthe embodiments described, many modifications and variations thereof are possible within the scope of the invention.

The network, for example, need not be a telephone network, but may be a telecommunications network for packet-switch streaming data between computers, the network being capable of optimisation by considering virtual routes through the network, behaving in a similar mannerto routes in a telephone network. The methods described above for optimising the flow of telephone traffic may be applied to such virtual routes to determine the optimal useage of links orto allocate new linkswheretheywould provide the most benefit. Further, while a very smail network has been described for the purposes of illustration, the invention may be applied to networks of 10,60, several hundreds or several thousand nodes, with a complex disposition of links and of circuit capacities in those links.

Claims (31)

1. Acircuit-switched communications network comprising nodes joined by links each comprising circuits, and an adaptive control system for selecting routes through the network, each route using a circuit from one or more links, the system comprising: means for determining a value ofthe traffic through the links and through all the routes through those links, and, from those traffic values, determining for each link,foreach routethroughthelink,the proportion of traffic currently through the link which is on that route, the said proportion corresponding to the ratio between the traffic value forthat link and the traffiovalueforthat route;; means for determining for each linkthe probability that use of an additional circuit from that link will subsequently blocka route through that link due to the unavailability of a circuit in that link; means for determining,from the said traffic proportions and the said blocking probabilities, a measureforeach linkofthe notional costtothe operator ofthe network which would be suffered by the anticipated use of a circuit ofthat link, and a measure for each route ofthe notional surplus value of operating that route, equivalent to a predetermined measure of benefit obtainable by operating the route less the sum of all the said measures of notional costsforthe link or all the links ofthat route, by performing an iteration to determine a current valueofthesaid notional costfroma previousvalue thereof and from the product of the said blocking probability and a weighted sum,for all routes through the network, of the said traffic proportions, the said weighted sum of the tramic proportions for each route being weighted in accordance with the said notional surplus value; and means responsiveto the said notional surplus value measures and/or the said notional cost measures to provide an output signal indicative thereof.

2. A network in accordance with Claim 1, wherein the said traffic values aretime-averaged values.

3. A network in accordance with Claim 2, wherein the means for determining the traffic values determine the time-averaged values of traffic along the links and routes by an iterative smoothing process, by repeatedly combining a proportion ofthe previous time-averaged traffic value with a corresponding proportion of a measure instantaneous traffic value.

4. A network in accordance with any preceding claim, wherein the means for determining traffic values and proportions is de-centralised.

5. A network in accordance with any preceding claim, wherein the means for determining the said measures of notional costs and notional surplus values is de-centralise.

6. A network in accordance with Claim 5, comprising arithmetic processing means assigned to each node which is a potential source of a stream of signals to be transmitted along a route, the source node arithmetic processing means determining the said traffic values only in respect of those links and routes to which that source node is potentially connectable, and, from those traffic values, determining, by the said iteration, for each such linkthe said measure of the notional cost, and, for each such route the said measure of notional surplus value.

7. A network in accordance with any preceding claim, wherein the said blocking probability is an estimated blocking probability determined in accordancewith a predetermined estimating function ofthe proportion of unavailable circuits in the link.

8. A network in accordance with Claim 7, wherein the said predetermined estimating function is determined by the said probability determining means by addressing a memory containing a set of values for the function for each of several values ofthetraffic through a link.

9. A network in accordance with any preceding claim, comprising meansforcomputing a loss probability on each route, representative of the probability of a call offered to that route being lost, and means responsive to the said loss probabilities and the said notional surplus values forthe routes to compute a notional nettvalue for each route, equal to the productofthe notional surplusvalueandthe difference between the loss probability and unity.

10. A network according to any preceding claim, wherein the output signal is indicative ofthe measure ofthe notional surplus value of a route, and compris ing means for offering a call to a selected route, the offering means comparing the said output signal for a specified route with a predetermined value, and offering the call eitherto that route orto an alternative route depending on the outcome ofthe said com parison.

11. A network according to Claim 10 as appendant to Claim 9, wherein a call is offered to a first choice route selected in accordance with the said compari son and, ifthat is blocked, to a second choice route selected bythe said comparing means comparing the outputsignalforthat route with the predetermined value, wherein the means for determining the said measure of notional costs and notional surplus values decreases the measure of notional surplus value forthefirst choice route by a quantity depen- dant upon the loss probability ofthe second choice route and upon the notional surplus value of the second choice route.

12. A network in accordance with any preceding claim, wherein the said weighted sum ofthe said traffic proportions for each route is weighted in accordance with the sum ofthe notional surplus value of operating that route, and the previous value ofthe said notional cost ofthe link.

13. A network in accordance with any preceding claim, for routing telephone calls and generating revenuethereby.

14. A network in accordance with any preceding claim, comprising display means responsive to the said output signal to provide a visual indication to human operators ofthe network ofthe said surplus revenues for each route and/or expected revenue lossesforeach link.

15. A network according to any preceding claim, comprising means for preventing use on a specified route of a circuit on a link in accordance with the number of available circuits on the link, and the said output signal forthat route.

16. Anetworkin accordancewith Ciaim 15, wherein the said preventing means is arranged to determine whether or notthe number of circuits on the link exceeds a predetermined value.

17. A method of routing traffic in a circuit switched communications network comprising nodes joined by links each comprising circuits, comprising determining values of the traffic th roug h the links and through all the routes th rough those links and determining, from those traffiovalues, for each link, for each route through the link, the proportion oftraffic currently through the link which is on that route the said proportion corresponding to the ratio between the trafficvalueforthat link and the trafficvalue forthat route: determining for each link the probabilitythat use of an additional circuit from that link will subsequently block a route through that link due to the unavailability of a circuit in that link; determining, from the said traffic proportions and the said blocking probabilities, a measure for each linkof 'the notional cost to the operator ofthe networkwhich would be suffered by the anticipated use of a ci rcuit of that link, and a measure for each route ofthe notional surplus value of operating that route, equivalentto a predetermined measure of benefit obtainable by operating the route less the sum of all the said measures of notional costs forthe link or all the links on that route, by performing an iteration to determine a current value of the said notional cost from a previous value thereof and from the the product of the said blocking probability and a weighted sum, for all routes through the network, of the said traffic proportions, the said weighted sum of the traffic proportionsforeach route being weighted in accord ancewiththesaid notional surplus value; and providing an output signal indicative of the said notional surpius values an/o rthe said notional cost measures.

18. A method in accordance with Claim 17, wherein the said traffic values are time-averaged values.

19. A method in accordance with Claim 18, wherein the time-averaged values of traffic along the links and routes are determined by an iterative smoothing process, by repeatedly combining a proportion ofthe previous time-averaged traffic value with a corresponding proportion of a measured instantaneous traffic value.

20. A method according to any of Claims 17 to 19, further comprising computing a loss probability on each route, representative of the probability of a call offered to that route being lost, and computing from the said loss probabilities and the said notional surplus values for the routes a notional nettvaluefor each route, equal to the product ofthe notional surplus value and the difference between the loss probability and unity.

21. A method according to any of Claims 17to 20, comprising thefurtherstep of offering a call to a route selected in accordance with a comparison ofthe said outputsignalsforthat route and at least one alternative route with a predetermined value.

22. A method in accordance with Claim 21 as appendantto Claim 20, wherein a call is offered to a first choice route selected in accordance with the said comparison and, ifthat call is blocked, to a second choice route selected by comparing the output signal forthatsecond choice route with the predetermined value, and wherein the measure of notional surplus valueforthe first choice route is decreased by a quantity dependent uponthe loss probability of the second choice route and upon the notional surplus value of the said choice route.

23. Amethod accordingto anyofClaimsl7to22 wherein the measure ofthe notional costfor each link is calculated by determining, for each possible route therethrough, the proportion of traffic currently through the link which is on that possible route, calculated asthe ratio between the traffiovaluefor that link and the traffic valueforthat route; determining, in accordance with a predetermined estimating function of the proportion of unavailable circuits in the link, an estimated blocking probability that use of a additional circuitfrom that link will subsequently blockthe route through that linkdueto the unavailability of acircuitin that link; and performing an iteration to determine a currentvalue ofthe said notional costfrom the previous value thereof and from the product of the said estimated blocking probability and a weighted sum, for all the possible routes, of the said traffic proportions.

24. A method in accordance with Claim 23, wherein the said predetermined estimating function is determined by addressing a memory containing a set of values forthe function for each of several values of the traffic th rough a link.

25. A method in accordance with any of Claims 17 to 24wherein the weighted sum is weighted in accordance with the sum of the notional surplus value currently obtainable by using that route and the previous value of the said notional cost ofthe link.

26. A method in accordance with any of Claims 17 to 25, further comprising the step of using circuitswitching means, in response to the said output signal, to switch the circuits to establish an optimal route dependent upon the said notional surplus values.

27. A method in accordance with any of Claims 17 to 26, comprising the step of displaying visibly a representation ofthe said notional surplus values and/orthe said notional costs, to enable human operators to managethe network.

28. A method in accordance with any of Claims 17 to 27, comprising means responsive to the said output signal to accumulate values of the output signal over a substantial period oftime, and, from those accumulated values, to provide an indication of the optimum allocation of extra capacity in the network, in accordance with the said notional surplus values.

29. A method of allocating extra capacity to a network comprising nodes joined by links each comprising circuits, taking into accountthe entire pattern of traffic flow through the network over a substantial period of operation, comprising determining time-averaged values ofthetrafficthrough the links and through all the routes through those links, and, from the said traffic values, determining by iterative calculations for each linka measure of the notional cost to the operator of the network which would be suffered by the anticipated use of a circuit of that link, and, for each route, determining the notional surplus value which would be obtained by operating that route, equal to a predetermined measure of benefit obtainable by operating the route less the sum offall the said measures of notional cost for the linkorall the links ofthat route; integ rating the said notional costvaluefora link with respect to time over the said substantial period of operation, to provide a measure representative of the increased benefit which could have been obtained by providing an additional ci circuit on that link; and allocating one or more extra circuits to that link dependent on the size ofthat increased benefit.

30. A circuit switched communications network, substantially as described herein with reference to the accompanying drawings.

31. A method of operating a circuit switched communications network, substantially as described herein with reference to the accompanying drawings.