GB2462120A

GB2462120A - Selecting a unique time point for initiation of an activity in a communications network

Info

Publication number: GB2462120A
Application number: GB0813648A
Authority: GB
Inventors: Matthew John Robinson
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 2008-07-25
Filing date: 2008-07-25
Publication date: 2010-01-27
Also published as: GB0813648D0

Abstract

A communications network comprises a plurality of nodes, each adapted to select a different time point within a time interval for initiation of an activity. Each node selects a time point for itself and monitors for notifications of time points selected by other nodes. Where no notifications are detected by a first node, the node selects a time point for itself. Where at least one notification is detected, the first node selects a time point for itself so as to avoid coincidence of the time point with any of the detected time points. Each node notifies the other nodes of its selected time point. In an alternative embodiment, the nodes may use this arrangement to select a different time points for each of a plurality of activities. The activity is typically related to traffic engineering (TE) such as may occur in an multi protocol label switched MPLS-TE network and the invention may be used to prevent simultaneous re-calculation of label switched paths (LSPs) using link-state protocols such as OSPF or IS-IS, particularly when calculated using "least fill" tie-break rules, for which some of the nodes might otherwise attempt to make calculations based upon outdated network information. RSVP-TE may be used to set up and maintain the LSPs across the network and a Desync Parameter LSA (link state advertisement), including an opaque LSA, may be used to carry the information necessary to select the unique time point.

Description

Node Management The invention is directed to managing nodes in a network in general and, in particular, to avoiding certain synchronous node activity in a network.

Traffic engineering is a well-known technique that is applied in communications networks with the aim of avoiding congestion and meeting commitments made to users for Quality of Service (QoS). Traffic engineering involves planning the distribution of data traffic across the network, so as to avoid faulty equipment whilst not overloading any particular node or link. In packet-switched networks, this is achieved by changing routing patterns implemented by router nodes in the network e.g. using Multi Protocol Label Switching (MPLS) Traffic Engineering (TE) features.

In link-state routing protocols such as OSPF and IS-IS, each node (or router) holds a local copy of a database which effectively contains a map of the network showing which nodes are connected to which other nodes. Each node then independently calculates the best routes for traffic to different destinations based on the map created by that node. To enable each node to have an up-to-date view of the overall network connectivity, link-state nodes share information about themselves and their local links with other routers. Link-state advertisement (LSA) is a mechanism used in OSPF for communicating this information to all other routers in the same OSPF area. Hence OSPF is often used in MPLS traffic engineering to manage routing patterns in router-based networks. As both equipment performance and traffic loads can vary over time, the map and, as a result, the traffic routing pattern may need to be adjusted from time to time.

One activity related to traffic engineering that each node in a network will undertake periodically is, therefore, the re-optimisation of paths which originate from that node. In MPLS TE, these are known as Label Switched Paths (LSPs). Although failure of links in a network will cause any LSPs which traverse those links to reroute, it will not directly cause other LSPs to be recalculated. Likewise the addition of extra links or nodes to the network does not have any instant effect on routing. Periodic re-optimisation allows the node to recalculate the "best" path to use for each LSP based on the current state of the network.

("Best" can be determined by a number of different algorithms -the network operator will select the approach which most closely matches their goals. One such approach is "least fill" -this selects the smallest number of hops across the network, but also aims to spread the traffic as evenly as possible across parallel links in the network) A problem can occur if LSP path re-optimisation is attempted by a number of nodes at the same time. This is especially apparent when "least fill" tie-break rules are used to select paths. The paths that a particular node will select are based on the currently reported reservations of other LSPs. The "least fill" tie-break algorithm works by comparing parallel paths of broadly equivalent preference and then choosing to place the LSP (and hence the traffic) on the least-used path. The aim of this is to balance traffic across the different links in the network and so achieve a fairly even usage. If the node performs its path selection at a unique time, then it should be able accurately to identify the least used links to place the LSPs on. However, if a number of nodes undertake this calculation simultaneously, the nodes will choose new paths based on the old state of the network and, as a result, may each choose the same link as part of their LSPs. This could cause that link to become more heavily loaded that was intended. Hence, to achieve stability, it is desirable for each router to undertake its re-optimisation activity at a different time to the other routers.

This situation could be addressed by specifying centrally a different activation time for each node of a network, however, this task would soon become onerous for any network above a limited size and would also involve a constant process of updating to reflect changes in network equipment or topology. An alternative approach might be for each node to introduce a random amount of jitter in the timing of their re-optimisation, however, due to the random nature of the timing, coincidences may still occur with possible undesired results.

In order to achieve a system which is simple to implement and scalable to large networks, it is desirable for node activity timing to be controlled in a distributed manner.

The above issues with conventional networks are addressed by the invention, as set out in the claims. In particular, the invention provides a communications network comprising a plurality of nodes, in which each of the nodes is adapted to select a different time point within a time interval for initiation of an activity; in which a first node of the plurality of nodes comprises a time-point selector for selecting a time point for the first node and a monitor for detecting notifications of time points selected by other ones of the plurality of nodes; in which the first node is adapted, where no notification of a time point selected by other nodes of the plurality of nodes is detected, to select a time point for the first node, in which the first node is adapted, where at least one notification of a time point selected by other nodes of the plurality of nodes is detected, to select a time point for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected; and in which the first node is adapted to notify the other nodes of the plurality of nodes of the time point selected by the first node.

The invention also provides a communications network comprising a plurality of nodes, in which each of the nodes is adapted to select a different time point within a time interval for initiation each of a plurality of activities; in which a first node of the plurality of nodes comprises a time-point selector for selecting, for each of the activities, a time point for the first node and a monitor for detecting notifications of time points selected by other ones of the plurality of nodes; in which the first node is adapted, where no notification of a time point selected by other nodes of the plurality of nodes is detected for one of the activities, to select a time point for that one of the activities for the first node, in which the first node is adapted, where at least one notification of a time point selected by other nodes of the plurality of nodes is detected for one of the activities, to select a time point for that activity for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point for that activity whose notification has been detected; and in which the first node is adapted to notify the other nodes of the plurality of nodes of each time point selected by the first node.

The invention also provides a method of managing, within a time interval, initiation of activity by a plurality of nodes in a communications network, in which each of the nodes selects a different time point within the time interval for initiation of the activity; in which the method includes the steps of: with respect to a first node, monitoring for notifications of time points selected by other ones of the plurality of nodes; and where no notification of a time point selected by other nodes of the plurality of nodes is detected, the method includes selecting a time point for the first node and notifying the other nodes of the plurality of nodes of the time point selected; where at least one notification of a time point selected by other nodes of the plurality of nodes is detected, the method includes selecting a time point for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected; and notifying the other nodes of the time point selected by the first node.

The invention also provides a method as claimed in any above claim including the steps of managing, within a time interval, initiation of a plurality of activities by each of a plurality of nodes in a communications network, in which, for each of the activities, each node selects a different time point within the time interval for initiation of the activity; in which the method includes the steps of: for each of the activities, with respect to a first node monitoring for notifications of time points selected by other ones of the plurality of nodes; and where no notification of a time point selected by other nodes is detected for one of the activities, the method includes selecting a time point for the first node for that activity and notifying the other nodes of the time point selected by the first node; where at least one notification of a time point selected by other nodes for one of the activities is detected, the method includes selecting a time point for that activity for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected for that activity; and notifying the other nodes of the time point selected by the first node.

According to an aspect of the invention, the selected time point is one of: i) a time point at a midpoint between the detected time point and one extreme of the time interval; where no detected time point is located between the detected time point and the extreme of the time interval; ii) a midpoint between a first and a second detected time point; where no detected time point is located between the first and the second detected time points; and iii) a time point at a midpoint of a time period consisting of the combination of a first time period extending between a first detected time point and one extreme of the time interval and a second time period extending between a second extreme of the time interval and a second detected time point; where no detected time point is located in the time period so defined.

According to further aspects of the invention, a time point is selected by identifying the largest time period between detected time points and selecting the mid point of the identified period; or by detecting more than one period of equal largest size between detected time points and selecting one of the periods based on a predetermined preference.

According to further aspects of the invention, a preset length for the interval is communicated to the plurality of nodes from a subset of the nodes and, further, each node monitors for notification of the interval length and, where notification of more than one length is detected by a node, the node selects the largest length.

According to a further aspect of the invention, each node is provided with a node identifier and, where a conflict between the time points selected by at least two of the nodes is detected, the node identifiers of the conflicting nodes are compared and the selection of one of the conflicting nodes is confirmed based on the values of the node identifiers and the rest of the conflicting nodes are required to select a different time point.

According to a further aspect of the invention, where a conflict in which at least two of the nodes notify selection of the same time point is detected; each of the conflicting nodes waits for a delay period before notifying the time point again, in which the lengths of the delay periods are determined by one of: setting each delay period to a different value and selecting the delay periods at random.

According to further aspects of the invention, each time point is defined in terms of an offset from one extreme of the interval, in which the offset into the time interval of the selected time point is subject to rounding such that it is an exact multiple of 2" of the time interval, where n is an integer; and further, the smallest value for n is selected such that the selected time point is distinct from any other detected time point. According to a further aspect of the invention, the monitoring is carried out by the first node.

In order to aid understanding, embodiments of the invention will now be described by way of example with reference to the drawings in which: Figure 1 shows a network of nodes in which the invention may be implemented; Figure 2 shows a network of nodes of Figure 1 indicating poor path placement; Figure 3 shows a network of nodes of Figure 1 indicating improved path placement resulting from application of the present invention; Figures 4a and 4b show schematic representations of node timing; Figure 5 illustrates the structure of a link state advertisement according to the invention.

MPLS TE features normally involve the use of a link state routing protocol such as OSPF or ISIS. The following embodiments are described using OSPF (as defined in IETF RFC 2328), but the same techniques could easily be used with IS-IS.

The following embodiments use a Desync Parameter LSA which is a form of type 10 Opaque LSA devised by the present inventors for implementation of the current invention. An OSPF type 10 Opaque LSA is arranged with a flooding scope so that it is area-local.

As shown in Figure 1, a network 10 comprises a plurality of routers 14 (A to D and W to Z), each arranged via links 18 to receive data in the form of packets and to forward received data in the form of packets to another one of the routers to which it is directly connected in dependence on routing information contained in a header section in the data packets according to the MPLS standard (as specified in IETF RFC 3032 and related RFCs).

Figure 2 shows two routes 20, 22 set up across network 10. As shown in Figure 2, route 20 connects routers A and C and route 22 connects routers B and D. The routing of Figure 2 is not optimal as both routes 20 and 22 pass through nodes W and Y and along link 18 between the two. Figure 3 shows routes 20, 22 set up across network 10 in an optimum arrangement. As with the routing of Figure 2, route 20 connects routers A and C; route 22 connects routers B and D. Unlike the arrangement of Figure 2, in Figure 3, while route 20 again passes through nodes W and Y, route 22 now passes through nodes X and Z so that the two routes do not share any common equipment in network 10.

Each router 14 is arranged to exchange network control information for traffic engineering with the other routers 14 of network 10. Information about routers 14 and links 18 in network is exchanged using the OSPF routing protocol. Each router 14 can then use this information to calculate which paths to use across network 10. For traffic engineering operation, the RSVP protocol with traffic engineering extensions is used to set up and maintain LSPs across the network. RSVP -TE is defined in IETF RFC 3209.

The OSPF routing protocol exchanges information between all directly connected (neighbouring) routers. This information is exchanged in the form of Link State Advertisements (LSAs). Each LSA describes a small portion of the network and may be restricted to parameters referring to a single router. The full set of LSAs for a network make up the OSPF link state database for that network. The operation of OSPF includes mechanisms to ensure that all routers in the network have an up-to-date copy of this database (i.e. that all routers have the same "map" of the network).

Network nodes such as routers normally have an onboard real time clock. This is typically synchronised to a known reference using network time protocol (NIP) as specified in IETF RFC 1305, or other network synchronisation standards. This means that each router in the network has a consistent view of "the current time of day". This shared view of the time of day is exploited by the present invention in order to allow routers to select different times at which to schedule their traffic engineering activities so that the traffic engineering activities of different routers do not coincide, i.e.: to not occur within a certain time window of each other.

The time window defining "coincidence" may, at one extreme, be determined by the resolution of the onboard real time clock (typically one millisecond), although it could be set to a value of ten milliseconds or more in order to more effectively avert the problems associated with synchronous activity. References to "selection of the same time point" will similarly be understood to imply selection of time points within the same time window.

A first embodiment of the present invention uses a Desync Parameter LSA carrying the following parameters: OSPF Router ID RID Interval length being used by Router RIL Offset from start of interval 0 Interval length for network NIL According to a preferred embodiment, one or more router 14 is configured to advertise a preset value for network interval length (NIL). In this case, the router or router so configured originates a Desync Parameter LSA containing the preset value for parameter NIL.

Desync Parameter LSA originated by routers that have not been configured with a preset value for network interval length (NIL), have the parameter NIL set to zero. Typically, five percent of routers might be configured with the preset value for NIL or another fraction appropriate to ensuring efficient distribution of preset NIL in the subject network topology.

Each router determines the router interval length (RIL) it should use based on advertised values of NIL the router has detected. In order to select a value for RIL, the set of all values of network interval length NIL advertised by other routers are examined and the largest value is chosen as the value of AlL to use and advertise. According to an alternative embodiment, the router interval length AlL is specified by local configuration.

The router then calculates the offset 0 from the start of the chosen interval (RIL). In the unlikely event that no Desync Parameter LSA is detected, the router sets offset 0 to a predetermined value. Otherwise, the values of offset 0 advertised by the other routers in detected Desync Parameter LSA are processed, as described next. Any values of offset 0 which exceed the interval length RIL being used by the router are excluded from further consideration (advertisement of such values should only occur as a transient event occurring before all the routers in the network have determined a common value for RIL).

Processing of the parameters received in the Desync Parameter LSA will now be described with reference to Figures 4a and 4b. Figure 4a shows an interval 40 of length RIL, determined as described above, and illustrates the situation where three Desync Parameter LSAs have been received by a router (referred to here as the "local router"). Each Desync Parameter LSA received has a different value of offset 0, shown as offsets 41, 42 and 43.

The offsets are shown relative to the start point of interval 40, which start point we label as time zero, for simplicity. Each offset 0 is identified with and defines a time point, i.e. the point in time located at offset 0 from the start of the interval. The local router calculates the size of the gaps between each received value of offset 0 and the next highest value of offset 0.

This process gives values for gap A, 45 (between offsets 41 and 42), gap B, 46 (between offsets 42 and 43) and gap C, 47. Gap C is special in that it "wraps round" the end of interval 40, i.e. from offset 43 to offset 41 in the forward direction. Figure 4b illustrates the time relationships of Figure 4a at a smaller scale to include two successive instances of interval 40, i.e.: one extending from time zero to time RIL and a second extending from time RIL to time 2RIL. Interval 40 repeats seamlessly, as illustrated in Figure 4b, so that gap C can be thought of as extending from the last offset of one instance of interval 40 to the first offset in the following instance of interval 40.

For most gaps, e.g.: gaps A and B the calculation performed is as follows: gap, O -0 In the case of the highest value of offset O, the calculation performed is as follows: gape RIL + O 0n The calculated gaps are compared in size and the largest of these gaps is chosen. If there are several gaps of equal, largest size, then any one of these may be chosen (for example, randomly or based on a method such as always picking the gap located closest to the start of interval 40). The value of offset 0 for the local router is then calculated as: °Ioca = °chosen + ( gap00 / 2) where °Ioca is the value of the offset selected for the local router; gap0s is the value of the gap calculated as being the largest (or selected from those calculated to be the largest, if more than one); is the value of the offset at the start of the chosen gap (i.e.: defining the limit of the gap closest to the start of interval 40).

The local router continues to monitor the OSPF database for changes in the values of Desync Parameter LSA parameters received from other routers in the network, If changes are detected, they are processed as described next. If a change is detected in an advertised value of NIL, the value of interval length RIL being used by this router is recalculated (as described, above), If the value of AlL changes as a result of this re-calculation, then the offset 0 for the local router from start of interval 40 may need to be recalculated as well (as described, above).

If the value of AlL decreases as a result of the recalculation, then the value of offset 0 selected by the local router may lie outside the new interval, as defined by the new value for AlL. If this is the case then a new value for offset 0 will need to be chosen. If RIL increases then there is more "space" available and no immediate need to choose a new value for 0, however, current values for offset 0 for all routers will be effectively shifted towards the lower end of the new interval (at least until a reboot or similar event occurs). Given that the value of RIL is unlikely to change very often, a simple and acceptable solution would be for each router to recalculate 0 whenever AlL changes, however, it may prove beneficial for each router to wait a short, random time before recalculating 0, so as to avoid too many simultaneous recalculations across the network.

Distribution of the various Desync Parameter LSAs will require a short period of time. This leads to the possibility that a router remote from the local router could select a value of offset 0 before it receives notification of the offset previously selected by the local router. (There is also the possibility that one or more routers may have selected values for 0 while isolated from the main network, resulting in a similar effect when they are connected to the main network.) If the local router detects advertisement of a new value of offset 0 from another router, it checks if this new offset matches the value of offset 0 in use by the local router. If two routers were to use the same value of offset, both routers might try to schedule activity simultaneously. To avoid this, upon detection of the advertisement of a matching value of 0, one of the two routers concerned chooses a new value of offset 0. For example, the router with the lowest value of RID will change its offset 0. Selection of a new value of 0 is achieved by repeating the process, set out above. According to an alternative embodiment, routers wait a random time before undertaking the process of choosing a new value of offset o so as to reduce the risk of several routers simultaneously changing to the same value of offset 0.

If any of the parameters (RIL, NIL, 0, RID) relating to the local router change, the local router issues a revised Desync Parameter LSA. Normal OSPF procedures will then advertise this information to all of the other routers.

According to a preferred embodiment, the calculation of time points is based upon integer arithmetic for simplicity and consistency across all routers in a network. In particular, when halving an odd number, the result is rounded down i.e. 69/2 is taken as 34 (not 34.5).

According to a further embodiment, once the integer time point values have been calculated, they are translated into floating point values for the scheduling calculations: including calculation of gap, local offset and scheduled time.

According to a further preferred embodiment, when a router initialises (i.e. after a power up or reset), it first downloads its OSPF database before advertising its own Desync Parameter LSA. This is a refinement to reduce the risk of the router having to change its selected time point based on information received from the active routers in the network immediately following initialisation.

An example of how parameters (e.g.: RIL, NIL, 0 and RID) could be advertised to other routers using OSPF by means of the Desync Parameter LSA will now be described with reference to Figure 5. The Desync Parameter LSA builds on the OSPF Opaque LSAs described in IETF RFC 5250, although other methods for packaging the parameters could also be used (e.g. using "Type, Length, Value" formats and/or extending the OSPF Traffic Engineering LSAs of IETF RFC 3630).

As shown in Figure 5, Desync Parameter LSA 50 is formatted as eight sections 51-58 each of 32 bits (as indicated by ordinals 59 added along the top of the drawing -these numbers do not form part of the Desync Parameter LSA). A type-b LSA is proposed for this task. A type-LSA is an opaque LSA with area flooding scope (as defined in IETF RFC 5250) and is a good match for the current traffic engineering implementations of IETF RFC 3630.

The 32-bit link-state ID 52 of the Opaque LSA is divided into an Opaque type field (the first 8 bits) and a type-specific ID (the remaining 24 bits). For this example, the type field is set to 128 (in the "experimental" number space). The type-specific ID is normally set to 0, although different values may be used to support multiple timing operations, if required (i.e. if the router needs to undertake multiple activities and wishes these to be scheduled independently).

The value of OSPF Router ID RID can be automatically determined from the 32-bit "Advertising Router" field 53 which is included in all OSPF LSAs.

Interval lengths RIL and NIL and offset 0 are each represented as 32-bit positive integers 56, 58 and 57 respectively, and, according to a preferred embodiment, express time-periods in milliseconds. This allows for periods of over 4 million seconds (i.e.: more than one month) to be expressed, which is ample for traffic engineering applications. For other applications, the data format could be modified to express different time scales, as required.

Once the values of router interval length RIL and offset 0 have been calculated for the local router, they are used to schedule the next occurrence of the activity by the router. In principle, the task is scheduled to be run at each: scheduled time = (N * RIL) + 0; for all N where N is an integer.

For this calculation to yield a comparable result across all routers, a common datum must be used for this scheduling, i.e.: the start time of the first interval of length RIL. For this embodiment we consider the datum to be Oh on 1 January 1900 (this matches the definition of an NTP timestamp as specified in IETF RFC 1305). Alternative embodiments may choose to use a different datum.

In practice, in order to initiate the scheduled task, the router only needs to determine the next occurrence of the scheduled time for that task and to configure an internal timer to fire at that time and to repeat at intervals equal to AlL. Whenever the local values of RIL and/or 0 are recalculated, the timer will need to be reconfigured accordingly.

According to a further enhancement of the invention, the activity is scheduled such that it only takes place over a subset of the day, for example restricting the activity to a "quiet time" such as 2am to 4am each day.

According to a further enhancement of the invention, calculation of the local value of offset 0 includes rounding such that the result is an exact multiple of 2 of time interval RIL, where n is an integer and the lowest value of n is selected such that the resultant value of offset 0 is different from any other 0 notified in the network. This enhancement may improve the long-term results in networks where there is considerable churn (addition and deletion) of routers from the OSPF network.

We have described example implementations of the de-synchronisation scheme, including extensions to the open shortest path first (OSPF) protocol version 2 to support advertisement of the required parameters, using Opaque Link State Advertisements. We have also described a method for nodes to determine for themselves the time at which they should carry out the required activity so as to avoid synchronising activities with other nodes. The result of the application of the scheme is a set of nodes for which the timing, for each node, of an activity is optimised by a distribution of the activities over a predetermined time interval so as to ensure the separation between activity times at different nodes.

The above embodiments are to be understood as illustrative examples of the invention.

Further embodiments of the invention are envisaged and will be evident to the skilled reader.

It is to be understood that any feature described in relation to any one embodiment may be used alone, or in combination with other features described, and may also be used in combination with one or more features of another of the embodiments, or any combination of the embodiments. Furthermore, equivalents and modifications not described above may also be employed without departing from the scope of the invention, which is defined in the accompanying claims.

In particular, as will be understood by those skilled in the art, the de-synchronisation scheme, described above, is not restricted to the selection of paths in the context of traffic engineering but extends to other network events where synchronous activity by a plurality of nodes is to be avoided. The scheme may equally apply to a plurality of activities carried out by each node, with either the same interval or different intervals being used for each activity. The invention is not restricted to OSPF but may also be implemented in networks operating according other routing protocols such as IS-IS or any other means of advertising the parameters to other nodes.

As will be understood by those skilled in the art, the invention may be implemented in computer program product software, any or all of which may be contained on various storage media so that the program can be loaded onto one or more general purpose computers or could be downloaded over a computer network using a suitable transmission medium. The computer program product used to implement the invention may be embodied on any suitable carrier, readable by a suitable computer input device, such computer program product comprising optically readable marks, magnetic media, punched card or tape, or on an electromagnetic or optical signal.

The text of the abstract is incorporated hereby: a communications network comprising a plurality of nodes, each adapted to select a different time point within a time interval for initiation of an activity. Each node comprises a time-point selector for selecting a time point for itself and a monitor for detecting notifications of time points selected by other nodes.

Where no notification of a time point selected by other nodes of the plurality of nodes is detected by a first, exemplary node, it selects a time point for itself. Where at least one notification of a time point selected by another node is detected by the first node, it selects a time point for itself so as to avoid coincidence of the time point it selects and the, or each, time point whose notification has been detected. Each node is adapted to notify the other nodes of the time point it has selected. In an alternative, each of the nodes is adapted to select a different time point within a time interval for initiation each of a plurality of activities so as to avoid coincidence of the time point it selects for one of the activities and the, or each, time point for the same activity whose notification has been detected. The activity is typically related to traffic engineering in the network.

Claims

CLAIMS1. A method of managing initiation of an activity by a plurality of nodes in a communications network, in which each of the nodes selects a different time point within a time interval for initiation of the activity; in which the method includes the steps of: with respect to a first node monitoring for notifications of time points selected by other nodes of the plurality of nodes; and where no notification of a time point selected by other nodes of the plurality of nodes is detected, the method includes selecting a time point for the first node and notifying the other nodes of the plurality of nodes of the time point selected; where at least one notification of a time point selected by other nodes of the plurality of nodes is detected, the method includes selecting a time point for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected; and notifying the other nodes of the time point selected by the first node.
2. A method of managing initiation of a plurality of activities by each of a plurality of nodes in a communications network, in which, for each of the activities, each node selects a different time point within a time interval for initiation of the activity; in which the method includes the steps of: for each of the activities, with respect to a first node monitoring for notifications of time points selected by other nodes of the plurality of nodes; and where no notification of a time point selected by other nodes is detected for one of the activities, the method includes selecting a time point for the first node for that activity and notifying the other nodes of the time point selected by the first node where at least one notification of a time point selected by other nodes for one of the activities is detected, the method includes selecting a time point for that activity for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected for that activity; and notifying the other nodes of the time point selected by the first node.
3. A method as claimed in any of claims 1 and 2 including the step of detecting the notification of at least one time point and selecting one of: i) a time point at a midpoint between the detected time point and one extreme of the time interval; where no detected time point is located between the detected time point and the extreme of the time interval; ii) a midpoint between a first and a second detected time point; where no detected time point is located between the first and the second detected time points; and iii) a time point at a midpoint of a time period consisting of the combination of a first time period extending between a first detected time point and one extreme of the time interval and a second time period extending between a second extreme of the time interval and a second detected time point; where no detected time point is located in the time period so defined.
4. A method as claimed in any above claim including selecting a time point by identifying the largest time period between detected time points and selecting the mid point of the identified period.
5. A method as claimed in any of the above claim including selecting a time point by detecting more than one period of equal largest size between detected time points and selecting the mid point of one of the periods based on a predetermined preference.
6. A method as claimed in any above claim, in which a preset length for the interval is communicated to the plurality of nodes from a subset of the nodes.
7. A method as claimed in any above claim, in which each node monitors for notification of the interval length and, where notification of more than one length is detected by a node, the node selects the largest length.
8. A method as claimed in any above claim, in which each node is provided with a node identifier, the method including the steps of detecting a conflict between the time points selected by at least two of the nodes; comparing the node identifiers of the conflicting nodes; confirming the selection of one of the conflicting nodes based on the values of the node identifiers and requiring the rest of the conflicting nodes to select a different time point.
9. A method as claimed in any of claims 1 to 7, including the steps of detecting a conflict in which at least two of the nodes notify selection of the same time point; each of the conflicting nodes waiting for a delay period before notifying the time point again, in which the lengths of the delay periods are determined by one of: setting each delay period to a different value and selecting the delay periods at random.
10. A method as claimed in any above claim, including defining periods of inactivity during which execution of the method of claim 1 is inhibited.
11. A method as claimed in any above claim, in which each time point is defined in terms of an offset from one extreme of the interval, in which the offset into the time interval of the selected time point is subject to rounding such that it is an exact multiple of 2 of the time interval, where n is an integer.
12. A method as claimed in claim 11, including selecting the smallest value for n such that the selected time point is distinct from any other detected time point.
13. A method as claimed in any above claim in which the monitoring is carried out by the first node.
14. A computer program product carrying executable code adapted to perform the method of any above claim.
15. A communications network comprising a plurality of nodes, in which each of the nodes is adapted to select a different time point within a time interval for initiation of an activity; in which a first node of the plurality of nodes comprises a time-point selector for selecting a time point for the first node and a monitor for detecting notifications of time points selected by other ones of the plurality of nodes; in which the first node is adapted, where no notification of a time point selected by other nodes of the plurality of nodes is detected, to select a time point for the first node, in which the first node is adapted, where at least one notification of a time point selected by other nodes of the plurality of nodes is detected, to select a time point for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point whose notification has been detected; and in which the first node is adapted to notify the other nodes of the plurality of nodes of the time point selected by the first node.
16. A communications network comprising a plurality of nodes, in which each of the nodes is adapted to select a different time point within a time interval for initiation of each of a plurality of activities; in which a first node of the plurality of nodes comprises a time-point selector for selecting, for each of the activities, a time point for the first node and a monitor for detecting notifications of time points selected by other ones of the plurality of nodes; in which the first node is adapted, where no notification of a time point selected by other nodes of the plurality of nodes is detected for one of the activities, to select a time point for that one of the activities for the first node, in which the first node is adapted, where at least one notification of a time point selected by other nodes of the plurality of nodes is detected for one of the activities, to select a time point for that activity for the first node so as to avoid coincidence of the time point selected by the first node and the or each time point for that activity whose notification has been detected; and in which the first node is adapted to notify the other nodes of the plurality of nodes of each time point selected by the first node.
17. A communications network as claimed in any of claims 15 and 16, in which the time point selected by the first node is one of: i) a time point at a midpoint between a detected time point and one extreme of the time interval; where no detected time point is located between the detected time point and the extreme of the time interval; ii) a midpoint between a first and a second detected time point; where no detected time point is located between the first and the second detected time points; and iii) a time point at a midpoint of a time period consisting of the combination of a first time period extending between a first detected time point and one extreme of the time interval and a second time period extending between a second extreme of the time interval and a second detected time point; where no detected time point is located in the time period so defined.
18. A communications network as claimed in any of claims 15 to 17, in which the time point selected by the first node is the mid point of the largest time period between detected time points.
19. A communications network as claimed in any of claims 15 to 18, in which a subset of the nodes is adapted to communicate a preset length for the time interval to the plurality of nodes.
20. A communications network as claimed in any of claims 15 to 19, in which each node is adapted to monitor for notification of the interval length and, where notification of more than one length is detected by a node, to select the largest length.
21. A communications network as claimed in any of claims 15 to 20, in which each node comprises a node identifier and in which the network is adapted to detect a conflict between the time points notified by at least two of the nodes; compare the node identifiers of the conflicting nodes; confirm the selection of one of the conflicting nodes based on the values of the node identifiers and require the rest of the conflicting nodes to select a different time point.
22. A communications network as claimed in any of claims 15 to 21 adapted to detect a conflict between the time points notified by at least two of the nodes; in which each of the conflicting nodes is adapted to wait for a delay period before notifying the time point again, in which the length of each delay period is set to one of a different value for each node and at random value.
23. A communications network as claimed in any of claims 15 to 22, in which each time point corresponds to an offset from one extreme of the interval, in which the offset into the time interval of each time point selected by a node is subject to rounding such that it is an exact multiple of 2 of the time interval, where n is an integer.
24. A communications network as claimed in claim 23, in which the value for n is the smallest value such that the selected time point is distinct from any other detected time point.