(19)
(11)EP 1 561 311 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
24.12.2014 Bulletin 2014/52

(21)Application number: 03758161.8

(22)Date of filing:  22.10.2003
(51)Int. Cl.: 
H04L 12/707  (2013.01)
H04L 12/751  (2013.01)
(86)International application number:
PCT/FI2003/000788
(87)International publication number:
WO 2004/040858 (13.05.2004 Gazette  2004/20)

(54)

DYNAMIC LOAD DISTRIBUTION USING LOCAL STATE INFORMATION

DYNAMISCHE LASTVERTEILUNG MITTELS LOKALER ZUSTANDSINFORMATION

REPARTITION DE CHARGE DYNAMIQUE FAISANT APPEL A DES INFORMATIONS D'ETAT LOCALES


(84)Designated Contracting States:
AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LI LU MC NL PT RO SE SI SK TR

(30)Priority: 01.11.2002 US 286477

(43)Date of publication of application:
10.08.2005 Bulletin 2005/32

(73)Proprietor: Nokia Corporation
02610 Espoo (FI)

(72)Inventors:
  • HEINER, Andreas
    FIN-02280 Espoo (FI)
  • BALANDIN, Sergey
    FIN-00920 Helsinki (FI)

(74)Representative: Papula Oy 
P.O. Box 981
00101 Helsinki
00101 Helsinki (FI)


(56)References cited: : 
EP-A2- 1 087 576
US-A- 6 098 107
US-A1- 2002 103 631
EP-A2- 1 146 768
US-A1- 2002 067 693
US-A1- 2002 147 842
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description

    FIELD OF THE INVENTION



    [0001] The present invention relates to communications. In particular, the present invention relates to a novel and improved dynamic load distributing method and system using local state information in a communication network.

    BACKGROUND OF THE INVENTION



    [0002] A communication network comprises nodes connected by links. A node refers to a network element implementing a protocol used for communicating with other nodes. A link refers to a communication facility or medium over which nodes can communicate. A communication network may either be connectionless such as an IP-based (Internet Protocol) network, or connection-oriented such as a fixed telephone network. When a traffic flow comprising data packets is transmitted from a given source node to a given destination node, there are often multiple different routes available via various nodes and links between the source and destination node. These routes are referred to as paths. In the following the term path is used to refer specifically to a loop-free path.

    [0003] Different paths have different costs associated with them. For example, a given path may have higher delay than another path. Thus it is advantageous to determine these costs beforehand and select a path based on this information. Since a traffic flow comprising packets may be distributed via multiple paths simultaneously, more than one paths may also be selected.

    [0004] The process of determining costs and selecting paths is often referred to as load distribution. Load in this context refers to the amount of packets or bytes being transferred over nodes and links thus loading them. Distributing traffic flows over multiple paths in a communication network has several advantages, such as fast failure recovery due to preexistence of multiple available paths between source and destination, higher throughput for a given infrastructure and reduced packet delay. Load distribution may be static, in which case assignment of jobs to hosts is done probabilistically or deterministically, without consideration of runtime events. Alternatively load distribution may be dynamic, in which case it is designed to overcome problems of unknown or uncharacteriseable workloads, non-pervasive scheduling and runtime variation, i.e. any situation where availability of hosts, composition of workload or interaction of human beings can alter resource requirements or availability. Dynamic load distribution systems typically monitor workload and hosts for any factors that may affect the choice of the most appropriate assignment and distribute jobs accordingly.

    [0005] A prior art method of static load distributing based on cost optimization is based on an estimate of static traffic demand between all node pairs and a weight function that is a decreasing function of the load over a link. Optimizing is performed using this heuristic cost function so that the cost of the total network is minimal, i.e. throughput through the network is optimized. The usefulness of the method is completely dependent on the accuracy of the traffic demand estimates.

    [0006] A prior art method of dynamic load distributing based on cost adjustment is generally referred to as Optimized Multi Path. Costs associated with links are slowly changed based on changes of network state parameters. If OSPF-protocol (Open Shortest Path First) is used in connection with Optimized Multi Path, information about the changes is disseminated in opaque LSAs (Link State Advertisement). Adjustments in link costs are made gradually enough to insure stability, yet to provide reasonably fast reaction to traffic demands. Paths between two nodes are based on the adjusted costs, so the paths may change over time.

    [0007] Prior art further includes document US 2002/0147842 A1 which discloses systems and method for selecting open shortest path first (OSPF) aggregates and aggregate weights for a particular area. Prior art further includes document EP 1087576 A2 which discloses a method and apparatus for constraint-based route selection using biased cost for a flow from a plurality of network paths connecting a source to a destination.

    [0008] Prior art further includes EP 1146768 A2 which discloses a load distribution device that allows a failure recovery process to be executed at a high performance rate and in a short time. EP 1146768 A2 merely teaches path selection wherein the entire path is calculated and selected at once when a connection setup request message is received.

    [0009] There are problems with prior art load distributing. Mechanisms that efficiently distribute traffic generally require network-wide state information to optimize or adjust path selection. These mechanisms use state information collected over varying time scales. Therefore they are reactive and operate on an hours to days time scale. Load distribution mechanisms that do not affect the paths between source and destination, on the other hand, distribute the load in a node over the available paths optionally including the costs of available paths. These mechanisms do not consider state information other than that which they may have themselves. Another problem associated with dynamic load distribution is maintaining flow integrity at all times. If flow integrity is not maintained, packet re-ordering generally occurs. In the case of TCP-based (Transmission Control Protocol) traffic this will reduce throughput considerably, whereas in the case of UDP-based (User Datagram Protocol) traffic applications may not be able to recover from packet reversal.

    [0010] Thus there is obvious need for a solution providing efficient dynamic load distribution using local state information in a communication network as well as using information of the neighbors to improve efficiency by taking into account the local environment.

    SUMMARY OF THE INVENTION



    [0011] The present invention concerns a dynamic load distributing method and system using local state information in a communication network comprising nodes connected by links. One or more nodes transmit their state information to one or more neighboring nodes. A neighboring node to a present node refers to a node which is directly, i.e. via one link, connected to the present node. If a neighboring node cannot process the received state information, it generates a corresponding signaling packet and transmits it to its respective neighboring nodes and so on until the point where the state information can be processed. This allows limiting signaling to a very small area.

    [0012] A traffic flow comprising packets is originated at a source node to be transmitted to a destination node via available paths. Weights are determined for the available paths. Each weight is determined as a decreasing function of cost of the corresponding path, state information of the present node i.e. the node determining the weight, and received state information of one or more nodes at the corresponding path. One of the paths is selected based on the determined weights. The traffic flow is dispatched to the next node of the selected path. In response to the present node not being the destination node, said determining weights, selecting paths and dispatching the traffic flow are repeated at said next node via which the traffic flow is forwarded.

    [0013] In an embodiment of the invention said transmitting of state information is performed by using OSPF-protocol (Open Shortest Path First), e.g. opaque LSA (Link State Advertisement) type 9.

    [0014] In an embodiment of the invention at one or more nodes the traffic flow is dispatched to the next nodes in fractions proportional to the weights determined for the paths which said nodes are a part of.

    [0015] In an embodiment of the invention packet classification, based on e.g. a prior art algorithm called the longest prefix match (LPM), is performed on arriving packets at one or more nodes, one or more fields of the arriving packets is hashed to obtain a hash value, and the obtained hash value is associated with a next node of the path with the highest determined weight.

    [0016] In an embodiment of the invention the cost of a path comprises delay associated with said path. In an embodiment of the invention the state information of a node comprises load on links from said node to next nodes.

    [0017] In an embodiment of the invention the state information of a node comprises CPU load of said node.

    [0018] In an embodiment of the invention the state information of a node comprises buffer load of said node.

    [0019] In an embodiment of the invention the communication network is a connectionless network.

    [0020] In an embodiment of the invention the communication network is a connection-oriented network.

    [0021] In an embodiment of the invention the communication network is an IP network.

    [0022] In an embodiment of the invention the communication network is an IP-RAN network (Internet Protocol Radio Access Network).

    [0023] The invention improves load distribution for short temporal traffic fluctuations. Load distribution according to the present invention minimizes the probability of congestion in downstream nodes due to sending traffic too aggressively to its optimal path. Alternative paths may be used as hot-standby paths for fast traffic protection. For example, in case of failure of the most favored link to next node traffic may be diverted quickly, a new set of multiple paths may be calculated and traffic will be distributed gracefully over the new set of feasible paths. Additionally, the invention is simple to implement as it only requires additional hash tables, means to determine state information of a present node and received state information of one or more nodes at the corresponding path, as well as means to relay said state information of one or more nodes at the corresponding path. Further, the present invention can be used in a multi-vendor environment.

    BRIEF DESCRIPTION OF THE DRAWINGS



    [0024] The accompanying drawings, which are included to provide a further understanding of the invention and constitute a part of this specification, illustrate embodiments of the invention and together with the description help to explain the principles of the invention. In the drawings:

    Fig 1 is a diagram illustrating a dynamic load distributing method using local state information according to one embodiment of the present invention, and

    Fig 2 is a block diagram illustrating a dynamic load distributing system using local state information according to one embodiment of the present invention.


    DETAILED DESCRIPTION OF THE INVENTION



    [0025] Reference will now be made in detail to the embodiments of the present invention, examples of which are illustrated in the accompanying drawings.

    [0026] Figure 1 illustrates a dynamic load distributing method using local state information in a communication network comprising nodes connected by links according to one embodiment of the present invention.

    [0027] One or more nodes transmit their state information to one or more neighboring nodes, phase 10. In the embodiment of the invention disclosed in Figure 1 this is performed by using OSPF (Open Shortest Path First) opaque LSA (Link State Advertisement) type 9. The state information may be transmitted either periodically, or in special cases such as when a given node begins to become congested. A traffic flow comprising packets is originated at a source node to be transmitted to a destination node via available paths, phase 11. In phase 12 weights are determined for the available paths. Each weight is a decreasing function of cost of the corresponding path, state information of the present node i.e. the node determining the weight, and received state information of one or more nodes at the corresponding path. In the embodiment of the invention disclosed in Figure 1 the cost of a path comprises delay associated with said path, the state information of a node comprises load on links from said node to next nodes, CPU load of said node, and buffer load of said node.

    [0028] Packet classification, based e.g. on the longest prefix match (LPM), is performed on the packets, one or more fields of the packets is hashed to obtain a hash value, and the obtained hash value is associated with a next node of the path with the highest determined weight, phase 13. One path is selected based on the determined weights, phase 14. The traffic flow is dispatched to the next node of the selected path, phase 15. Said determining weights, selecting paths and dispatching the traffic flow are repeated at one or more nodes via which the traffic flow is forwarded until the traffic flow reaches the destination node, phases 16 and 17. At one or more nodes the traffic flow may be dispatched to the next nodes in fractions proportional to the weights determined for the paths which said nodes are a part of (not illustrated in Figure 1).

    [0029] Figure 2 illustrates a dynamic load distributing system using local state information in a communication network according to one embodiment of the present invention. In the embodiment of the invention disclosed in Figure 2 the communication network is an IP-RAN network (Internet Protocol Radio Access Network), i.e. a connectionless network. The network comprises multiple nodes N connected by links L. The node originating a traffic flow comprising packets to be transmitted is referred to as a source node SN. The node being destination for the traffic flow transmitted via available paths is referred to as destination node DN.

    [0030] In the embodiment of the invention disclosed in Figure 2 one or more nodes comprise a weight determiner WD for determining weights for the available paths. Each weight is determined as a decreasing function of cost of the corresponding path, state information of the present node, and received state information of one or more nodes at the corresponding path. In the embodiment of the invention disclosed in Figure 2 the cost of a path comprises delay associated with said path, whereas the state information of a node comprises load on links from said node to next nodes, CPU load of said node and/or buffer load of said node. One or more nodes further comprise a path selector PS for selecting a path based on the determined weights. One or more nodes further comprise a dispatcher DP for dispatching the traffic flow to the next node of the selected path. One or more nodes further comprise a fractionalizer FR for dispatching the traffic flow to the next nodes in fractions proportional to their respective weights.

    [0031] In the embodiment of the invention disclosed in Figure 2 one or more nodes further comprise a transmitter TR for transmitting state information of the node to one or more neighboring nodes. The transmitter may use e.g. OSPF opaque LSA type 9 for transmitting state information. The transmitter may transmit state information either periodically, or in special cases such as when the corresponding node begins to become congested. One or more nodes further comprise a classifier CL for performing packet classification on arriving packets, and a hasher HS hashing one or more fields of the classified packets to obtain a hash value, and for associating the obtained hash value with a next node of the path with the highest determined weight.

    [0032] In the embodiment of the invention disclosed in Figure 2 the weight determiner, path selector, dispatcher, fractionalizer, transmitter, classifier and hasher may be implemented in software and/or hardware.

    [0033] It is obvious to a person skilled in the art that with the advancement of technology, the basic idea of the invention may be implemented in various ways. The invention and its embodiments are thus not limited to the examples described above, instead they may vary within the scope of the claims.


    Claims

    1. A dynamic load distributing method using local state information in a communication network comprising nodes connected by links, wherein the method comprises the steps of:

    transmitting (10) from one or more nodes state information of said node to one or more neighboring nodes,

    originating (11) a traffic flow comprising packets at a source node to be transmitted to a destination node via available paths, and characterized by:

    determining (12) weights for the available paths, each weight determined as a decreasing function of cost of the corresponding path, state information of the present node, and received state information of one or more nodes at the corresponding path,

    selecting (14) a path based on the determined weights,

    dispatching (15) the traffic flow to the next node of the selected path, and

    in response to the present node not being the destination node, repeating (12, 14, 15, 16) said determining weights, selecting paths and dispatching the traffic flow at said next node via which the traffic flow is forwarded.


     
    2. The method according to claim 1, wherein said transmitting of state information is performed by using OSPF-protocol.
     
    3. The method according to claim 1, wherein at one or more nodes the traffic flow is dispatched to the next nodes in fractions proportional to their respective weights.
     
    4. The method according to claim 1, wherein the method further comprises the steps of:

    performing (13) packet classification on arriving packets at one or more nodes,

    hashing (13) one or more fields of the classified packets to obtain a hash value, and

    associating (13) the obtained hash value with a next node of the path with the highest determined weight.


     
    5. The method according to claim 1, wherein the cost of a path comprises delay associated with said path.
     
    6. The method according to claim 1, wherein the state information of a node comprises load on links from said node to next nodes.
     
    7. The method according to claim 1, wherein the state information of a node comprises CPU load of said node.
     
    8. The method according to claim 1, wherein the state information of a node comprises buffer load of said node.
     
    9. The method according to claim 1, wherein the communication network is a connectionless network.
     
    10. The method according to claim 1, wherein the communication network is a connection-oriented network.
     
    11. The method according to claim 1, wherein the communication network is an IP network.
     
    12. The method according to claim 1, wherein the communication network is an IP-RAN network.
     
    13. A dynamic load distributing system using local state information in a communication network comprising nodes (N) connected by links (L), one of the nodes being a source node (SN) for originating a traffic flow comprising packets to be transmitted via available paths, and one of the nodes being a destination node (DN) for receiving the transmitted packets, the nodes comprising:

    a transmitter (TR) for transmitting state information of said node to one or more neighboring nodes, and characterized by :

    a weight determiner (WD) for determining weights for the available paths, each weight determined as a decreasing function of cost of the corresponding path, state information of the present node, and received state information of one or more nodes at the corresponding path,

    a path selector (PS) for selecting a path based on the determined weights, and

    a dispatcher (DP) for dispatching the traffic flow to the next node of the selected path.


     
    14. The system according to claim 13, wherein said transmitter (TR) uses OSPF-protocol for transmitting state information.
     
    15. The system according to claim 13, wherein one or more nodes further comprise:

    a fractionalizer (FR) for dispatching the traffic flow to the next nodes in fractions proportional to their respective weights.


     
    16. The system according to claim 13, wherein one or more nodes further comprise:

    a classifier (CL) for performing packet classification on arriving packets, and

    a hasher (HS) for hashing one or more fields of the classified packets to obtain a hash value, and for associating the obtained hash value with a next node of the path with the highest determined weight.


     
    17. The system according to claim 13, wherein the cost of a path comprises delay associated with said path.
     
    18. The system according to claim 13, wherein the state information of a node comprises load on links from said node to next nodes.
     
    19. The system according to claim 13, wherein the state information of a node comprises CPU load of said node.
     
    20. The system according to claim 13, wherein the state information of a node comprises buffer load of said node.
     
    21. The system according to claim 13, wherein the communication network is a connectionless network.
     
    22. The system according to claim 13, wherein the communication network is a connection-oriented network.
     
    23. The system according to claim 13, wherein the communication network is an IP network.
     
    24. The system according to claim 13, wherein the communication network is an IP-RAN network.
     


    Ansprüche

    1. Dynamisches Lastverteilungsverfahren, das Informationen über den lokalen Zustand in einem Kommunikationsnetz, das Knoten umfasst, die durch Übertragungsstrecken verbunden sind, verwendet, wobei das Verfahren die folgenden Schritte umfasst:

    Senden (10) von Zustandsinformationen des Knotens von einem oder von mehreren Knoten an einen oder an mehrere Nachbarknoten,

    Erzeugen (11) eines Verkehrsflusses, der Pakete umfasst, bei einem Quellknoten, damit sie über verfügbare Wege zu einem Zielknoten übertragen werden, und gekennzeichnet durch:

    Bestimmen (12) von Gewichtungen für die verfügbaren Wege, wobei jede Gewichtung als eine fallende Funktion der Kosten des entsprechenden Wegs, von Zustandsinformationen des vorliegenden Knotens und von empfangenen Zustandsinformationen eines oder mehrerer Knoten auf dem entsprechenden Weg bestimmt wird,

    Auswählen (14) eines Wegs auf der Grundlage der bestimmten Gewichtungen,

    Absenden (15) des Verkehrsflusses an den nächsten Knoten des ausgewählten Wegs, und

    Wiederholen (12, 14, 15, 16) des Bestimmens von Gewichtungen, des Auswählens von Wegen und des Absendens des Verkehrsflusses bei dem nächsten Knoten, über den der Verkehrsfluss weitergeleitet wird, in Ansprechen darauf, dass der vorliegende Knoten nicht der Zielknoten ist.


     
    2. Verfahren nach Anspruch 1, wobei das Senden von Zustandsinformationen unter Verwendung des OSPF-Protokolls ausgeführt wird.
     
    3. Verfahren nach Anspruch 1, wobei der Verkehrsfluss bei einem oder bei mehreren Knoten an die nächsten Knoten in Anteilen abgesendet wird, die proportional zu ihren entsprechenden Gewichtungen sind.
     
    4. Verfahren nach Anspruch 1, wobei das Verfahren ferner die folgenden Schritte umfasst:

    Ausführen (13) einer Paketklassifizierung an ankommenden Paketen bei einem oder bei mehreren Knoten,

    Hash-Codieren (13) eines oder mehrerer Felder der klassifizierten Pakete, um einen Hash-Wert zu erhalten, und

    Zuordnen (13) des erhaltenen Hash-Werts zu einem nächsten Knoten des Wegs mit der höchsten bestimmten Gewichtung.


     
    5. Verfahren nach Anspruch 1, wobei die Kosten eines Wegs eine dem Weg zugeordnete Verzögerung umfassen.
     
    6. Verfahren nach Anspruch 1, wobei die Zustandsinformationen eines Knotens die Last auf Übertragungsstrecken von dem Knoten zu nächsten Knoten umfassen.
     
    7. Verfahren nach Anspruch 1, wobei die Zustandsinformationen eines Knotens die CPU-Last des Knotens umfassen.
     
    8. Verfahren nach Anspruch 1, wobei die Zustandsinformationen eines Knotens die Pufferlast des Knotens umfassen.
     
    9. Verfahren nach Anspruch 1, wobei das Kommunikationsnetz ein verbindungsloses Netz ist.
     
    10. Verfahren nach Anspruch 1, wobei das Kommunikationsnetz ein verbindungsorientiertes Netz ist.
     
    11. Verfahren nach Anspruch 1, wobei das Kommunikationsnetz ein IP-Netz ist.
     
    12. Verfahren nach Anspruch 1, wobei das Kommunikationsnetz ein IP-RAN-Netz ist.
     
    13. Dynamisches Lastverteilungssystem, das Informationen über den lokalen Zustand in einem Kommunikationsnetz, das Knoten (N) umfasst, die durch Übertragungsstrecken (L) verbunden sind, verwendet, wobei einer der Knoten ein Quellknoten (SN) zum Erzeugen eines Verkehrsflusses, der Pakete umfasst, damit sie über verfügbare Wege zu einem Zielknoten übertragen werden, ist und wobei einer der Knoten ein Zielknoten (DN) zum Empfangen der übertragenen Pakete ist, wobei die Knoten Folgendes umfassen:

    einen Sender (TR) zum Senden von Zustandsinformationen des Knotens an einen oder an mehrere Nachbarknoten, und gekennzeichnet durch:

    eine Gewichtungsbestimmungseinrichtung (WD) zum Bestimmen von Gewichtungen für die verfügbaren Wege, wobei jede Gewichtung als eine fallende Funktion der Kosten des entsprechenden Wegs, von Zustandsinformationen des vorliegenden Knotens und von empfangenen Zustandsinformationen eines oder mehrerer Knoten auf dem entsprechenden Weg bestimmt wird,

    eine Wegauswahleinrichtung (PS) zum Auswählen eines Wegs auf der Grundlage der bestimmten Gewichtungen, und

    eine Absendeeinrichtung (DP) zum Absenden des Verkehrsflusses an den nächsten Knoten des ausgewählten Wegs.


     
    14. System nach Anspruch 13, wobei der Sender (TR) das OSPF-Protokoll zum Senden von Zustandsinformationen verwendet.
     
    15. System nach Anspruch 13, wobei einer oder mehrere Knoten ferner Folgendes umfassen:

    eine Anteilbildungseinrichtung (FR) zum Absenden des Verkehrsflusses an die nächsten Knoten in Anteilen, die proportional zu ihren jeweiligen Gewichtungen sind.


     
    16. System nach Anspruch 13, wobei einer oder mehrere Knoten ferner Folgendes umfassen:

    einen Klassierer (CL) zum Ausführen einer Paketklassifizierung an ankommenden Paketen, und

    eine Hash-Codierungseinrichtung (HS) zum Hash-Codieren eines oder mehrerer Felder der klassifizierten Pakete, um einen Hash-Wert zu erhalten, und zum Zuordnen des erhaltenen Hash-Werts zu einem nächsten Knoten des Wegs mit der höchsten bestimmten Gewichtung.


     
    17. System nach Anspruch 13, wobei die Kosten eines Wegs eine dem Weg zugeordnete Verzögerung umfassen.
     
    18. System nach Anspruch 13, wobei die Zustandsinformationen eines Knotens die Last auf Übertragungsstrecken von dem Knoten zu nächsten Knoten umfassen.
     
    19. System nach Anspruch 13, wobei die Zustandsinformationen eines Knotens die CPU-Last des Knotens umfassen.
     
    20. System nach Anspruch 13, wobei die Zustandsinformationen eines Knotens die Pufferlast des Knotens umfassen.
     
    21. System nach Anspruch 13, wobei das Kommunikationsnetz ein verbindungsloses Netz ist.
     
    22. System nach Anspruch 13, wobei das Kommunikationsnetz ein verbindungsorientiertes Netz ist.
     
    23. System nach Anspruch 13, wobei das Kommunikationsnetz ein IP-Netz ist.
     
    24. System nach Anspruch 13, wobei das Kommunikationsnetz ein IP-RAN-Netz ist.
     


    Revendications

    1. Procédé de répartition de charge dynamique faisant appel à des informations d'état locales dans un réseau de communication comprenant des noeuds connectés par des liaisons, dans lequel le procédé comprend les étapes consistant à :

    transmettre (10) depuis un ou plusieurs noeuds des informations d'état dudit noeud à un ou plusieurs noeuds voisins,

    émettre (11) un flux de trafic comprenant des paquets au niveau d'un noeud source afin qu'il soit transmis à un noeud de destination par l'intermédiaire de trajets disponibles, et caractérisé par les étapes consistant à :

    déterminer (12) des poids pour les trajets disponibles, chaque poids étant déterminé sous la forme d'une fonction décroissante du coût du trajet correspondant, d'informations d'état du noeud présent, et d'informations d'état reçues d'un ou plusieurs noeuds sur le trajet correspondant,

    sélectionner (14) un trajet sur la base des poids déterminés,

    distribuer (15) le flux de trafic au noeud suivant du trajet sélectionné, et

    en réponse au fait que le noeud présent n'est pas le noeud de destination, répéter (12, 14, 15, 16) ladite détermination de poids, sélectionner des trajets et

    distribuer le flux de trafic au niveau dudit noeud suivant par l'intermédiaire duquel le flux de trafic est réacheminé.


     
    2. Procédé selon la revendication 1, dans lequel ladite transmission d'informations d'état est effectuée en utilisant un protocole OSPF.
     
    3. Procédé selon la revendication 1, dans lequel, au niveau d'un ou plusieurs noeuds, le flux de trafic est distribué aux noeuds suivants par fractions proportionnelles à leurs poids respectifs.
     
    4. Procédé selon la revendication 1, dans lequel le procédé comprend en outre les étapes consistant à :

    effectuer (13) une classification de paquets sur des paquets entrants au niveau d'un ou plusieurs noeuds,

    hacher (13) un ou plusieurs champs des paquets classés afin d'obtenir une valeur de hachage, et

    associer (13) la valeur de hachage obtenue à un noeud suivant du trajet ayant le poids déterminé le plus élevé.


     
    5. Procédé selon la revendication 1, dans lequel le coût d'un trajet comprend le retard associé audit trajet.
     
    6. Procédé selon la revendication 1, dans lequel les informations d'état d'un noeud comprennent la charge imposée à des liaisons depuis ledit noeud vers des noeuds suivants.
     
    7. Procédé selon la revendication 1, dans lequel les informations d'état d'un noeud comprennent une charge de CPU dudit noeud.
     
    8. Procédé selon la revendication 1, dans lequel les informations d'état d'un noeud comprennent la charge de tampon dudit noeud.
     
    9. Procédé selon la revendication 1, dans lequel le réseau de communication est un réseau sans connexion.
     
    10. Procédé selon la revendication 1, dans lequel le réseau de communication est un réseau orienté connexion.
     
    11. Procédé selon la revendication 1, dans lequel le réseau de communication est un réseau IP.
     
    12. Procédé selon la revendication 1, dans lequel le réseau de communication est un réseau IP-RAN.
     
    13. Système de répartition de charge dynamique faisant appel à des informations d'état locales dans un réseau de communication comprenant des noeuds (N) connectés par des liaisons (L), l'un des noeuds étant un noeud source (SN) destiné à émettre un flux de trafic comprenant des paquets devant être transmis par l'intermédiaire de trajets disponibles, et l'un des noeuds étant un noeud de destination (DN) destiné à recevoir les paquets transmis, les noeuds comprenant :

    un émetteur (TR) destiné à transmettre des informations d'état dudit noeud à un ou plusieurs noeuds voisins, et

    caractérisé par :

    un dispositif de détermination de poids (WD) destiné à déterminer des poids pour les trajets disponibles,

    chaque poids étant déterminé sous la forme d'une fonction décroissante du coût du trajet correspondant,

    d'informations d'état du noeud présent, et

    d'informations d'état reçues d'un ou plusieurs noeuds sur le trajet correspondant,

    un sélecteur de trajet (PS) destiné à sélectionner un trajet sur la base des poids déterminés, et

    un distributeur (DP) destiné à distribuer le flux de trafic au noeud suivant du trajet sélectionné.


     
    14. Système selon la revendication 13, dans lequel ledit émetteur (TR) utilise un protocole OSPF pour transmettre des informations d'état.
     
    15. Système selon la revendication 13, dans lequel un ou plusieurs noeuds comprennent en outre :

    un dispositif de fractionnement (FR) destiné à distribuer le flux de trafic aux noeuds suivants par fractions proportionnelles à leurs poids respectifs.


     
    16. Système selon la revendication 13, dans lequel un ou plusieurs noeuds comprennent en outre :

    un classificateur (CL) destiné à effectuer une classification de paquets sur des paquets entrants, et

    un hacheur (HS) destiné à hacher un ou plusieurs champs des paquets classés afin d'obtenir une valeur de hachage, et à associer la valeur de hachage obtenue à un noeud suivant du trajet ayant le poids déterminé le plus élevé.


     
    17. Système selon la revendication 13, dans lequel le coût d'un trajet comprend un retard associé audit trajet.
     
    18. Système selon la revendication 13, dans lequel les informations d'état d'un noeud comprennent la charge imposée à des liaisons depuis ledit noeud vers des noeuds suivants.
     
    19. Système selon la revendication 13, dans lequel les informations d'état d'un noeud comprennent une charge de CPU dudit noeud.
     
    20. Système selon la revendication 13, dans lequel les informations d'état d'un noeud comprennent la charge de tampon dudit noeud.
     
    21. Système selon la revendication 13, dans lequel le réseau de communication est un réseau sans connexion.
     
    22. Système selon la revendication 13, dans lequel le réseau de communication est un réseau orienté connexion.
     
    23. Système selon la revendication 13, dans lequel le réseau de communication est un réseau IP.
     
    24. Système selon la revendication 13, dans lequel le réseau de communication est un réseau IP-RAN.
     




    Drawing









    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description