Redundancy and load balancing in a telecommunication unit and system

ES 2 289 302 T3

DESCRIPTION

Redundancy and load balancing in a telecommunications unit and system.

Background of the invention

The present invention relates to network element redundancy and load balancing in a telecommunications system, and especially to the use of parallel gateway nodes, such as GGSN (GPRS Gateway Support Node) nodes in a system. mobile packet switching. To provide a concrete example, the invention will be described in the context of a packet-switched mobile communication system.

The continuous evolution of applications carried over mobile systems places ever-increasing demands on mobile networks. To allow heavy traffic, efficient use of the radio network is important, determining the capacity of the system. In many applications, packet-switched connections are more efficient than circuit-switched connections. They are particularly well suited for bursty data transmission, for example for the use of the Internet. In this case, a high bit rate is required to load a new page, although, on the other hand, data traffic is practically non-existent when the page is viewed. However, in circuit-switched connections, the capacity of the connection is reserved all the time for a certain user, thereby wasting resources and the user must also pay for it. In a packet-switched system, resource allocation is based on the amount of data transmitted and not on the duration of the connection.

GPRS (General Packet Radio Service) is a technique that enables packet-switched data transmission to be used, for example, in the third generation UMTS (Universal Mobile Telecommunications System) mobile network. GPRS requires the introduction of new network elements, such as the GGSN, in the mobile system. The GGSN is the network element of the GPRS and UMTS mobile networks and controls the routing of data packets in the GPRS network and deals with the connection of the GPRS network with other networks, such as the Internet and other GPRS networks.

US 6,392,989 B1 discloses a method of providing a recovery plan for a connection when a connection is interrupted due to a link or switch failure. Mapping or forwarding tables are obtained from the plans and they are distributed to the switches. In the event of a failure, each switch switches to the recovery plan corresponding to the failure. When the fault is repaired or eliminated, the switches switch back to their main plans.

In the GPRS system, the logical connection between a mobile station and a GGSN supporting the mobile station is called a PDP (Packet Data Protocol) context. A redundant GGSN node comprises several processing units GTP-U (GPRS Tunneling Protocol - User Plane) and GTP-C (GPRS Tunneling Protocol Control Plane) and apply packet transmission based on PDP contexts. Redundancy is used in the GTP-U and GTP-C processing units to continue packet transmission even in error situations. Redundancy is based on getting a second processing unit to take over, in the event that the main unit cannot continue to transmit packets. Redundancy of network nodes, such as the GGSN, is typically implemented using spare units with a redundancy ratio of 1: 1, whereby one spare unit is available for each active unit. The problem with 1: 1 redundancy is that it makes the network node structure complicated and expensive, in the event that each processing unit must be backed up. Brief description of the invention

Therefore, an object of the invention is to solve the mentioned problem. One of the objectives of the invention is to reduce the hardware resources allocated to obtain redundancy. The objective is achieved by developing a method and a system that implements the method and a network element that are characterized by the aspects disclosed in the independent claims. Preferred embodiments of the invention are disclosed in the dependent claims.

The invention is based on the use of groups, comprising parallel network element units, called group nodes, to support a network element, such as a GGSN. A group node is an example of a GTP-U or GTP-C processing unit capable of serving PDP context activation requests. The processing units act as standby units for each other. The group nodes include logical nodes that represent the pairs formed by the group nodes in such a way that each pair is associated with a pair of logical nodes, and one of the logical nodes resides in the first group node and the other in the second group node. In the pair of logical nodes, one of the logical nodes is active and the other is on standby. A directed pair of logical nodes, which indicates the active and standby logical node, is referred to as a load allocation alternative.

In this case, GGSN redundancy is based on the idea that when a user plane node malfunctions, PDP contexts whose active logical node resides in the failed group node will be served by the standby logical node. par, which then becomes the active logical node.

The method, system and network element of the invention provide the advantage that 1: 1 redundancy is not

ES 2 289 302 T3 needed in the system and a pair can be defined for each node of the group even in the case that there is an odd number of group nodes, for example three. In this way, when a group node malfunctions, only 33% of the PDP contexts need to be transferred to be served by another group node considering that the network element load is divided between three group nodes.

In one of the embodiments of the invention, the network load is balanced in such a way that when a PDP context is activated, a logical node can be selected as the active node from the group node presenting the least load. . This embodiment further provides the advantage that a solution is provided for both load balancing and network element unit redundancy. In this way, the system is flexible, that is, it has high availability and reliability. This option is especially advantageous in a situation in which only part of the sessions should be highly reliable. This embodiment is particularly well suited for an environment comprising multiple all-IP (Internet Protocol) GGSN units based solely on packet-switched data transmission, in which a certain flexibility and availability are required on the part. of the system. Another advantage of this embodiment is that it alleviates the problem related to load balancing based on IP-based packet transmission, that is, the fact that the transmission speed can be significantly reduced when the load increases. .

The invention further provides a combined flexibility solution for the GTP-C and GTP-U processing units.

Brief description of the figures

The invention will now be described in more detail by means of the preferred embodiments and with reference to the accompanying drawings, in which Figure 1 is a block diagram of a simplified system of the invention, Figure 2 is a schematic representation of the external IP addressing of the invention at the user level, Figure 3 is a schematic representation of the load balancing of the invention at the user level, When an internal high-speed switch is available, Figure 4 is a schematic representation of the load balancing of the invention at the user level, when an internal high-speed switch is not available, Figure 5 is a schematic representation of the external IP addressing of the invention in the user plane and the control plane, Figure 6 is a schematic representation of the load balancing of the invention in the user plane and the control plane, when an internal high speed switch is available, Figure 7 is a schematic representation of the load balancing of the invention at the user plane and control plane, when an internal high speed switch is not available.

Detailed description of the invention

The present invention can be applied in any telecommunications system in which a redundancy of network elements is implemented using, together with active units, standby units that can be activated when the active unit malfunctions. These systems include third generation mobile systems, such as UMTS (Universal Mobile Telecommunications System), and systems based thereon, and systems, such as. GSM 1800 and PCS (Personal Communications System), corresponding to the GSM system (Global System for Mobile Communications). The invention can also be applied to other wireless systems and fixed systems.

The invention is described below using an illustrative system that is based on, but not limited to, a 3GPP-all-IP system. The 3GPP-all-IP is a system based on IP technology that makes use of the GPRS defined in the 3GPP Third Generation Association Project, implementing a redundancy of network elements using parallel reserve units in said system.

Figure 1 shows a simplified GPRS architecture showing only the parts that are essential to understanding the invention. It is clear to a person skilled in the art that a mobile system also comprises other functions and structures that need not be described in detail here.

The main parts of a mobile system are a core network CN, a radio access network RAN ​​and a mobile station, also referred to as user equipment UE. The GPRS system uses a 3G radio access network (such as the UMTS radio access network) or a radio communication network.

ES 2 289 302 T3 2G radio access (such as GSM radio access network).

The GPRS CN core network can be connected to external networks, such as the Internet. The main parts of the core network CN are a serving gateway support node SGSN and a GPRS gateway support node GGSN. The core network described in this document is based on the UMTS core network. Other types of core networks, for example IS-41, may comprise other network elements.

The main functions of the SGSN are the detection of GPRS UE mobile stations in its service area, the processing of the registrations of the new mobile stations UE, the transmission of data packets to and from a GPRS UE mobile station and the maintenance of a record on the locations of mobile stations in the service area.

The main functions of the GGSN include interaction with an external data network. The GGSN connects the operator's GPRS network to external systems, such as the GPRS systems of other operators, and to data networks, such as the Internet. The GGSN contains the PDP addresses of the GPRS subscribers and routing information, in other words the SGSN addresses. The interface on the SGSN side of the GGSN is called the Gn interface and the interface on the IP network side is called the Gi interface. The interface between the SGSN and a network managed by another network operator PLMN2 (Public Land Mobile Network) is called the Gp interface. The operation of the GGSN of the invention will be described later with reference to Figures 2, 3, 4, 5 and 6.

The mobile station UE can be a single voice terminal or it can be a multi-service terminal that provides service to a service platform and supports the loading and execution of different functions related to services. The mobile station UE comprises the mobile equipment in question and typically a removably connected identification card, which is also referred to as a subscriber identity module, SIM. The mobile station can be any device or a combination of several different devices capable of communicating in a mobile system. The subscriber identity module typically includes the subscriber identifier, executes authentication algorithms, stores authentication and encryption keys, and required subscriber information at the mobile station.

To transmit and receive GPRS data, the mobile station UE must activate at least one PDP context that it wishes to use. The term PDP refers to a protocol that transmits data in packets. This activation makes the mobile station known to the corresponding GGSN, and interworking with the network can begin. A PDP context defines different data transmission parameters, such as PDP address, QoS quality of service, and NSAPI (Network Service Access Point Identifier).

A mobile station connected to a GPRS system can initiate a PDP context activation at any time by transmitting a PDP context activation request to the SGSN. After receiving the message, the SGSN transmits a PDP context creation request towards the GGSN which creates the PDP context and transmits it towards the SGSN. The SGSN transmits an acknowledgment of the PDP context activation towards the mobile station UE in response to a successful PDP context activation, after which a virtual connection is established between the mobile station UE and the GGSN. As a consequence, the SGSN transmits data packets from the mobile station UE to the GGSN and the GGSN transmits data packets received from an external network through the SGSN to the mobile station UE. The PDP context is stored in the mobile station UE, the SGSN and the GGSN. Without limiting it to the GPRS system; PDP context is any logical context that is established for packet-switched data transmission between a terminal and the network element controlling the connection. One or more PDP contexts represent each PDP address in the mobile station UE, the SGSN and the GGSN.

The term GTP (GPRS Tunneling Protocol) refers to a protocol that is used to transmit user data between GPRS nodes in the GPRS core network. Two PDP contexts on different interfaces and connected to each other form a GTP tunnel.

A system that implements the functionality of the present invention and its network element comprises not only means of the prior art but means to implement the functions described in more detail in relation to Figures 2, 3, 4, 5, 6 or 7. More specifically, they comprise maintenance means for maintaining virtual group nodes on the group nodes, means for forming load allocation alternatives of said virtual group nodes and / or means for changing the load allocation. Additionally, the network nodes may comprise means for distributing the load on active group nodes, means for distributing the load of the network element among the group nodes, means for distributing the load of the network element based on an allocation plan of load, means to define an external IP address for load assignment alternatives, means for maintaining information about a major and minor group node associated with the load allocation alternative, switching means for transmitting data using the IP address defined for the load allocation alternative and / or means for changing the load allocation internally in the network element. It is also possible that the system and its network nodes understand all of the aforementioned media.

The present network nodes comprise processors and memory that can be used in the functions of the invention. All modifications required to implement the invention can be made in the form of added and updated software routines and / or using hardware-based solutions, such as ASIC (Application Specific Integrated Circuit) circuits, or a separate logic module.

ES 2 289 302 T3

In the following description the terms "control plane" (CP) and "user plane" (UP) are used. All information transmitted and received by a user from a mobile station, such as encoded voice data in voice calls or packets from an Internet connection, is transmitted on the user plane. The control plane is used for all mobile system related control signaling that is not typically visible to the user. Exceptions to this may occur, eg short messages may be transmitted in the control plane. At the radio interface, control plane and user plane data can be multiplexed on the same physical channel.

Implementation of the first embodiment

The following discussion describes how the invention is implemented at the user level.

Figure 2 shows a network element NE, which is for example a GGSN node. The network element comprises two or more physical GTP-U processing units A, B, C, which in this case are also called group nodes. Group nodes A, B, C are capable of serving PDP contexts. The group nodes are arranged in pairs in such a way that each group node can form a pair with any other group node. For example, if the number of available cluster nodes N = 3, the number of possible pairs P = 3. If N = 4, then P = 6. In the situation shown in Figure 2, the number of group nodes N = 3, and the possible pairs of group nodes are in this case AB, BC and CA. In each of the pairs, the first group node is a spare unit for the second group node.

According to the present invention, the group nodes contain logical nodes al, a2, b1, b2, c1, c2, which are also called virtual group nodes or virtual nodes. The number of virtual nodes is preferably twice the number of the pairs formed by the group nodes. The virtual group nodes al, a2, b1, b2, c1, c2 are logical GTP-U processing units. They are arranged in pairs in such a way that the first virtual node of the pair is active and the second is on standby. The same cluster node can comprise both active and standby virtual nodes.

A directed pair of virtual nodes has a visible characteristic outside the network element, called the load allocation alternative LBX1, LBX2, LBY1, LBY2, LBZ1, LBZ2, which is a Gn, Gp or Gi logical interface. Table 1 shows the load allocation alternatives for three processing units A, B, C, and illustrates which of the virtual nodes in the load allocation alternative is active and which is on standby.

TABLE 1

<td rowspan="2">Pair of processing units</td><td rowspan="2">Alternative load allocation</td><td colspan="2">Virtual node</td>

<td>Active</td><td>On hold</td>

<td rowspan="2">AB</td><td>LBX1</td><td>a1</td><td>b2</td>

<td>LBX2</td><td>b2</td><td>a1</td>

<td rowspan="2">BC</td><td>LBY1</td><td>b1</td><td>c2</td>

<td>LBY2</td><td>c2</td><td>b1</td>

<td rowspan="2">AC</td><td>LBZ1</td><td>c1</td><td>a2</td>

<td>LBZ2</td><td>a2</td><td>c1</td>

Two or more pairs of virtual nodes formed in the same way may have different load allocation alternatives which may present other differences apart from the formulation of the pair of virtual nodes.

Each load allocation alternative has an external IP address that is used as the user plane address of the PDP context. When a PDP context activation request is processed in the GGSN, one of the load allocation alternatives is selected. A serving virtual node is selected, that is, an active node, and a standby virtual node. The selection is based for example on the load information of the group nodes or on a specific alternative scheme.

If a group node A, B, C malfunctions, the assignment of the PDP contexts whose active unit is that group node is changed. The active virtual node serving the PDP context is placed on standby and the corresponding standby virtual node changes to active, unless the situation occurs that it is also down. In the present description, the change of active and standby unit is also referred to as "switch transition". When the corresponding standby units become active, they begin to serve the PDP contexts.

Each load allocation alternative has an individual external user plane IP address on the Gn or Gp interface to receive the data packets arriving at the GGSN. This individual address of the load allocation alternative is used as the PDP context address of the active virtual node of the load allocation alternative. The characteristic of the load allocation alternative is that it is visible on the external interfaces of the network element. The IP address is used to indicate the route through the physical interface of group node A, B, C.

ES 2 289 302 T3

The traffic in the network element NE can be distributed among the group nodes comprising active virtual nodes based on a specific load allocation plan. The traffic on the network element NE can be distributed among the group nodes of comprising standby virtual nodes, whereby the standby virtual nodes become active.

Implementation of the first embodiment using an internal switch

Figure 3 shows an implementation of the invention at the user level, when the GGSN has an internal high-speed switch K or a corresponding connection between group nodes A, B, C. By means of the internal switch K, in the interfaces of the network element can hide all the modifications necessary to recover from failures, whereby the change from the active group node to the standby node and vice versa is not visible outside the network element NE. The packets arriving at the physical interface Gi Gif or the Gn Gnf interface of the group node A, B, C are transmitted through the internal switch K to the active group node. For the sake of clarity, Figure 3 shows a switch K, although there may actually be several switches. Gnf could also illustrate the Gp interface.

According to one of the preferred embodiments of the invention, the allocation of loads is based on routing protocols (flexibility of the links based on routing), in other words, information about a primary and secondary route to the allocation alternative of loads is kept inside the GGSN. The primary path is the physical interface of the group node that comprises the active logical node of the load allocation alternative. The secondary path is the physical interface of the standby drive. When a PDP context is activated, the load allocation alternative is offered from a group node A, B, C that has available capacity. The Gnf physical Gn interface or Gif physical Gi interface, which the packets arrive at at the network node, need not reside in the group node from which the load allocation alternative is offered. In other words, packets can reach any interface and are transmitted through the network node's internal switch K to the active group node. The forwarding of information is maintained in the GGSN to allow the indication of the main and secondary route through the physical Gi or Gn interface of the group node to the active unit. The load allocation change or a switch transition between the active unit and the standby unit is not visible outside the network element, as the external IP address of the load allocation alternative is the only visible address outside the element network. When the main connection malfunctions, the data packets are guided to use the secondary path of the load allocation alternative. A bug in the first interface (eg Gn) of the GGSN is not visible to the second interface (eg Gi).

According to another preferred embodiment of the invention, the load allocation is based on a tie-layer solution (tie-layer flexibility). Also in this case it is not necessary that the physical Gn interface Gnf or the physical Gi interface Gif reside in the group node A, B, C, in which the active unit processes packets. In the link layer solution, a load allocation alternative has a physical interface dedicated to it, and the standby unit monitors the physical connection or Gif, Gnf interface of the active unit. That work can also be done by another GGSN component. If the standby unit receives information about a malfunction from the interface used by the active unit, the standby unit initiates the switchover transition. The standby drive then starts using the standby interface instead of the Gif, Gnf dedicated physical interface of the faulty drive, and packets addressed to the faulty drive are sent to the standby drive through the internal switch K . Also in this embodiment, a switching transition or an internal load allocation change of the load allocation alternative is not visible outside the network element as a change in the external IP address but as a change in the address on the link layer, as the IP address of the load allocation alternative is the only routing address visible outside the network element. When the main connection malfunctions, the data packets are guided to use the secondary path of the load allocation alternative. A bug in the first interface (eg Gn) of the GGSN is not visible to the second interface (eg Gi).

According to yet another of the preferred embodiments of the invention, the above-mentioned routing-based solution and the link layer solution can be applied simultaneously, in which case a rapid recovery from the malfunction of the device takes place first. cluster node based on link layer solution followed by routing based recovery.

Visibility of a load allocation alternative can be found on both sides of the NE network element. This means that a similar feature of the load allocation alternative can be available on the Gi and Gn side of the GGSN. The external addressing of the load allocation alternatives can be based, for example, on the range of subnet addresses used for the IP addresses of the PDP context, the endpoint address of the. IP tunneling in the case of a tunneling mechanism (such as GRE routing generic encapsulation, IP-in-IP, or the IP IPSec security protocol), or a set of label-switched paths used by the allocation alternative loading.

Subsequently, in the section "Changing the physical interface", of this description, methods are described to guide data packets so that they use a secondary route towards the alternative of load allocation in a solution based on routing or to initiate the use of an alternative physical interface in the link layer solution.

ES 2 289 302 T3

Deploying the first embodiment without any internal switch

Figure 4 shows an implementation of the invention at the user plane, when no internal high-speed switch is available in the network element. The switching transition or the change of the load allocation is not in this case an internal change of the network element, but is visible to the outside, in other words, the path changes on both interfaces Gn, Gi.

In yet another of the preferred embodiments of the invention, an integrated load allocation change or switching transition is performed. In a routing-based solution, the data packets in this case arrive at the network element NE in such a way that the external address of the load allocation alternative is marked as its IP address. The main routes for transmitting data packets are those that use the Gif physical Gi interface or the Gnf physical Gn interface of the group node A, B, C, in which the active load allocation alternative resides. When the group node malfunctions, the child routes are used. When secondary routes are used, packets arrive at the physical interface of the cluster node, where the second load allocation alternative resides and which then becomes the active unit. The change from the route to the child route is visible outside the network element. Routing protocols can be used to indicate the secondary route on the physical interface.

In an integrated load allocation change of still another preferred embodiment of the invention, it is preferable to use a link layer solution that is based on the idea that the unit awaiting the load allocation alternative monitors the interface. Gif physical Gi and Gnf physical Gn interface of the active drive. This function can also be performed by another GGSN component. If an error is detected, the standby drive begins to use the alternate physical interface of the failed drive. Later, in the section "Changing the physical interface" of this description, methods for changing the physical interface are described. The standby drive can then start using the alternate physical interface, if this is everyone's standby drive. PDP contexts whose active drive is the faulty drive. This situation can be achieved by indicating paths on the physical interface of the standby drive that replaced the failed drive. Since changes to the Gn side also need to be made to the Gi side (and vice versa), a logical Gi interface (or Gn interface) is assigned for each load allocation alternative.

The routing solution or the link layer solution can be applied to the Gi interface regardless of which solution is applied to the Gn interface, and vice versa. Subsequently, in the section "Changing the physical interface" of the present description, methods are described to guide data packets in such a way that they use a secondary route towards the alternative of load allocation in a solution based on routing or to initiate the use of alternate physical interface in link layer solution.

Implementation of the second embodiment

The following discussion describes a combined implementation of the user plane and the control plane of the invention.

Figure 5 shows a network element NE, which is, for example, a GGSN node. The network element comprises two or more physical GTP-U A, B, C processing units, which in this case are also called user plane group nodes, and two or more physical GTP-C processing units D, E, F which are also called in this case group nodes of the control plane. The user plane and control plane group nodes A, B, C, D, E, F are capable of serving PDP contexts. According to the second embodiment of the invention, the service pairs are formed with group pairs of the user plane and the control plane in such a way that a group node of the user plane and a group node of the plane controls form a pair. In a service pair, both nodes are active. Each user plane group node can form a pair with any control plane group node and vice versa. The serving pair has a standby reserve pair. Each backup pair can be the backup pair of any service pair. The spare pair further comprises a control plane group node and a user plane group node.

According to this embodiment, the group nodes of the user plane and the control plane comprise logical nodes, which are also called virtual group nodes or virtual nodes a4, b4, d4, e4. Virtual nodes are logical GTP-U or GTP-C processing units. The virtual nodes are arranged in pairs such that the first virtual node of the first pair resides in the GTP-U processing unit of the serving pair and the second virtual node of the first pair resides in the GTP-U processing unit of the pair. reserve, and in such a way that the first virtual node of the second pair resides in the GTP-C processing unit of the service pair and the second virtual node of the second pair resides in the GTP-C processing unit of the reserve pair. The pairs of virtual nodes are further arranged by secondary pairs in such a way that the first pair of virtual nodes resides in the GTP-U processing units and the second pair of virtual nodes resides in the GTP-C processing units. Table 2 shows the possible service pairs and spare pairs for three GTP-U processing units A, B, C and three GTP-C processing units D, E, F.

ES 2 289 302 T3

TABLE 2

<td>UP processing units</td><td>CP processing units</td><td>Service torque + Service torque reservation</td>

<td rowspan="4">A, B</td><td rowspan="4">OF</td><td>AD + BE</td>

<td>AE + BD</td>

<td>BE + AD</td>

<td>BD + AE</td>

<td rowspan="4">A, B</td><td rowspan="4">E, F</td><td>AE + BF</td>

<td>AF + BE</td>

<td>BF + AE</td>

<td>BE + AF</td>

<td rowspan="4">A, B</td><td rowspan="4">F, D</td><td>AF + BD</td>

<td>AD + BF</td>

<td>BD + AF</td>

<td>BF + AD</td>

<td rowspan="4">B, C</td><td rowspan="4">OF</td><td>BD + CE</td>

<td>BE + CD</td>

<td>CE + BD</td>

<td>CD + BE</td>

<td rowspan="4">B, C</td><td rowspan="4">E, F</td><td>BE + CF</td>

<td>BF + CE</td>

<td>CF + BE</td>

<td>CE + BF</td>

<td rowspan="4">B, C</td><td rowspan="4">F, D</td><td>BF + CD</td>

<td>BD + CF</td>

<td>CD + BF</td>

<td>CF + BD</td>

<td rowspan="4">C, A</td><td rowspan="4">OF</td><td>CD + AE</td>

<td>CE + AD</td>

<td>AE + CD</td>

<td>AD + CE</td>

<td rowspan="4">C, A</td><td rowspan="4">E, F</td><td>CE + AF</td>

<td>CF + AE</td>

<td>AF + CE</td>

<td>AE + CF</td>

<td rowspan="4">C, A</td><td rowspan="4">F, D</td><td>CF + AD</td>

<td>CD + AF</td>

<td>AD + CF</td>

<td>AF + CD</td>

The control plane virtual node d4, e4 and the user plane virtual node a4, b4 of the spare pair can reside on different physical units even if the virtual nodes of the service pair reside on the same physical unit and vice versa.

A directed secondary pair of virtual nodes forms a load allocation alternative LB1. A secondary directed pair of virtual nodes indicates which are the active virtual nodes and the standby virtual nodes associated with it. In the load allocation alternative LB1, the user plane virtual node a4 and the control plane virtual node d4 are active, and the user plane virtual node b4 and the control plane virtual node e4 are its nodes. virtual standby.

Two or more secondary pairs of virtual nodes formed in the same way may have different load allocation alternatives which may present other differences apart from the formulation of the secondary pair of virtual nodes.

The load allocation alternatives have external IP addresses that are used as the addresses of the PDP contexts. The direction may be different for the user plane and the control plane. When the GGSN is

ES 2 289 302 T3 processes a PDP context activation request, load allocation alternatives are selected for user plane and control plane. When a load allocation alternative is selected, the working (initial) pair and the reserve (initial) pair are selected. The user plane IP address, for example, the IP address for group nodes A and B, and the control plane IP address, for example, the IP address for group nodes D and E. Selection is based, for example, on group node load information or a specific alternative scheme.

If a group node malfunction prevents a virtual node from continuing as the active node, the backup virtual node becomes the active node. This operation can be performed separately on the user plane and the control plane. In this way, a switching transition is carried out. The active pair of virtual nodes serving the PDP context goes into standby and the corresponding pair of standby virtual nodes becomes active, unless it turns out that it is also on a faulty unit. When the corresponding standby units become active units, they begin to serve the PDP contexts. The control plane group mode load assignment should not necessarily be changed if the failing unit is a user plane group node, and vice versa.

The routing address of the load allocation alternative is used as the PDP context address of the active virtual node of the load allocation alternative. It is the characteristic of the load allocation alternative that is visible outside the network element. The IP address is used to indicate the route through the physical interface of the cluster node.

The traffic in the network element NE can be distributed among the group nodes comprising active virtual nodes based on a specific load allocation plan. The traffic in the network element NE can be distributed among the nodes. group comprising standby virtual nodes, whereupon standby virtual nodes become active.

Implementation of the second embodiment using an internal switch

Figure 6 shows the implementation of the invention at the user level, when the GGSN has an internal high-speed switch K or a corresponding connection between group nodes A, B, C, D, E, F. By means of the internal switch K, in the interfaces of the network element all the modifications required for the recovery with respect to the missing ones can be hidden, with which the change from the active group node to the standby node and vice versa is not visible outside of the NE network element. The packets arriving at the physical Gi interface Gif or the Gn Gnf interface of the group node A, B, C are transmitted through the internal switch K to the active group node. For the sake of clarity, Figure 6 shows a switch K, although in reality, several switches may be provided.

According to one of the preferred embodiments of the invention, the allocation of loads is based on routing protocols (flexibility of the links based on routings), in other words, information about a primary and secondary route to the allocation alternative load is kept inside the GGSN. The primary path is the physical interface of the group nodes that comprise the service pair. The secondary route is the physical interface of the group nodes that comprise the backup pair. When a PDP context is activated, the load allocation alternative is offered from a pair of pool nodes that have available capacity. It is not necessary that the Gnf physical Gn interface or the Gif physical Gi interface, to which the packets arrive at the network node, reside in the group nodes from which the load allocation alternative is offered. In other words, the packets can reach any interface and they are transmitted through the internal switch K of the network node to the active units. Forwarding information is maintained in the GGSN to allow indication of the primary and secondary route through the physical Gi or Gn interface of the group node to the active units. The internal load allocation switching or switching transition of a load allocation alternative is not visible outside the network element, since the IP addresses of the load allocation alternative are the only addresses visible outside the network element. When the primary connection malfunctions, the data packets are guided to use the secondary path of the load allocation alternative. A bug in the first interface (eg Gn) of the GGSN is not visible to the second interface (eg Gi).

According to yet another of the preferred embodiments of the invention, the assignment of PDP contexts is based on a link layer solution (link layer flexibility). Also in this case, it is not necessary that the physical Gn interface Gnf or the physical Gi interface Gif reside in the group node pair, in which the active units process packets. In the link layer solution, a load allocation alternative has a physical interface dedicated to it, and the standby units monitor the Gif, Gnf physical interface or connection of the active units. This function can also be performed by another GGSN component. If the standby units receive information about a malfunction from the interface used by the active units, the standby units initiate the switchover transition. The standby drives then start to use the standby interface instead of the Gif, Gnf dedicated physical interface of the faulty drive, and packets addressed to the faulty drive are sent to the standby drives through the internal switch K . In this embodiment, the switching transition or internal load allocation change of the load allocation alternative is not visible outside the network element as a change in the external IP address but as a change in the address at the layer since the IP address of the load allocation alternative is the routing address visible outside the network element. When the main connection malfunctions, the data packets are routed to use the

ES 2 289 302 T3 secondary route of the load allocation alternative. A fault in the first interface (eg Gn) of the GGSN is not visible to the second interface (eg Gi).

According to yet another of the preferred embodiments of the invention, the aforementioned routing-based solution and the link layer solution can be applied simultaneously, in which case a rapid recovery from node malfunction first takes place. group based on the link layer solution followed by a routing based recovery.

Visibility of a load allocation alternative can be found on both sides of the NE network element. This means that there may be a similar characteristic of the load allocation alternative on the Gi and Gn side of the GGSN. The external addressing of the load allocation alternatives can be based, for example, on the range of subnet addresses used for the IP addresses of the PDP context, the endpoint address of the IP tunnel in the case of a tunneling mechanism ( such as GRE, IP-in-IP, or IPSec), or a set of label switched trails used by the load allocation alternative.

Subsequently, in the section "Changing the physical interface" of the present description, methods are described to guide data packets in such a way that they use a secondary route towards the alternative of load allocation in a solution based on routing or to start use an alternate physical interface in the link layer solution.

Implementation of the second embodiment if no internal switch

Figure 7 shows an implementation of the invention at the user plane, when no internal high-speed switch is available in the network element. In that case, the change of load allocation or switching transition is not an internal change of the network element, but is visible to the outside. In other words, the path of the data packets changes on both Gn, Gi interfaces.

In still another preferred embodiment of the invention, an integrated load allocation change is performed. In this case, in a routing-based solution, the data packets arrive at the network element NE in such a way that the external address of the load allocation alternative is marked as its IP address. The main route for transmitting data packets is the one that uses the Gif physical Gi interface or the Gnf physical Gn interface of the group nodes A, B, C, D, E, F in which the active load allocation alternative resides. . When the group node malfunctions, the child path is used. When the secondary route is used, the packets arrive at the physical interface of the group nodes, in which the backup pair resides and which in this case become the active units. The change from the route to the child route is visible outside the network element. Routing protocols can be used to indicate the secondary route on the physical interface.

In an integrated load allocation change or switching transition of a preferred embodiment of the invention, it is possible to use a link layer solution that is based on the idea that the units waiting for the load monitor the Gif physical Gi interface and the Gnf physical Gn interface of the active units. This function can also be performed by another GGSN component. If an error is detected, the standby drive begins to use the alternate physical interface of the failed drive. Subsequently, in the section "Changing the physical interface" of this description, methods for changing the physical interface are described. The standby unit can then start to use the alternate physical interface, if it is the standby unit for all PDP contexts whose active unit is the faulty unit. This situation can be achieved by indicating paths on the physical interface of the standby drive that replaced the failed drive. As changes on the Gn side also need to be made on the Gi side (and vice versa), for each load allocation alternative a logical Gi interface (or Gn interface) is assigned.

The routing solution or the link layer solution can be applied to the Gi interface regardless of which solution is applied to the Gn interface, and vice versa. Next, in the section "Changing the physical interface" of the present description, methods are described to guide data packets so that they use a secondary route towards the alternative of load allocation in a solution based on routing or to start to use the alternate physical interface in the link layer solution.

Changing the physical interface

When you bind an IP address to a new link layer address in the situations described above, data packets may be lost. The forwarding, unicast and multicast modes are methods for transmitting data arriving at the network element NE towards the physical interface Gi or Gn of the active group node A, B, C, D, E, F in the situations shown in Figures 3, 4, 6 or 7 without packet loss. Another advantage of these methods is that no dedicated physical interface is required.

Forwarding mode means that one of the group nodes acts as the parent for the group IP routing address of the network element or for the external address (s) of the load allocation alternative . The main element receives the packets addressed to the group IP routing address or the external address of the load allocation alternative and forwards them to the active group node. Forwarding is based, for example, on the address of the load allocation alternative of the data packet. If the main element works

ES 2 289 302 T3 defectively, another main element is selected in the network element, towards which the packets directed to the group IP routing address are thereafter transmitted. The forwarding mode can be applied to the routing-based solutions described above in such a way that the network element may or may not have an internal switch available for it.

Unicast mode means that the data arriving at the network element NE is transmitted separately to each cluster node, even if the data has only one receiver. Each cluster node then receives the packet and either accepts it or rejects it depending on the routing address (es) of the data load allocation alternative in question. The rejection or acceptance of the packet can also be due to the content of the data packet.

Multicast mode means that the data can have multiple simultaneous receivers. In this case, the data packet arrives at the group IP routing address of the network element, from which it is forwarded in multicast to all group nodes. Depending on the routing address (es) of the data load allocation alternative, the cluster node either accepts or rejects the packet. The rejection or acceptance of the packet can also be due to the content of the data packet.

Although the invention has been described above as an example with GTP-U or GTP-C management in the GGSN, it is clear to a person skilled in the art that the invention can also be applied to other protocols. The invention can also be applied to other network elements implemented by groups. Examples of other network elements, in which the invention can be applied, are a GPRS serving node (SGSN), an IP IP-BTS base transceiver station and a radio access network gateway RAN-GW.

Although the invention has been presented above describing the redundancy of both the input and output interface and the load allocation of the network element, it is clear to a person skilled in the art that the invention can also be applied to situations in which only one of the interfaces is used.

It is apparent to a person skilled in the art that as technology advances, the basic idea of ​​the invention can be implemented in many different ways. Therefore, the invention and its embodiments are not limited to the examples described above, but may vary within the scope of the claims.