Method, device and computer program for use in routing - Patents.com

Abstract

A method is disclosed for use in routing traffic between at least one user equipment (UE) and a point of presence (PoP) of a virtual network function (VNF). The PoP is disposed in a network comprising a multi-access edge computing infrastructure comprising a plurality of routers, at least some of the plurality of routers being linked, with at least one link linking the plurality of routers with the PoP. The method includes: a) obtaining an indication of which of at least two possible paths for traffic between the at least one UE and the PoP and passing through a router in the plurality of routers corresponds to a smaller estimate of an impact on the centrality of the PoP; and b) thereby enabling selection of the respective path for routing traffic between the at least one UE and the VNF.

Description

[Technical field]

[0001] This specification relates to routing traffic between User Equipment (UE) and Virtual Network Function (VNF) Points Of Presence (PoPs). [Background technology]

[0002] Various technical specifications define how network functions and services are deployed and managed globally by network operators and service providers. Virtual Network Functions (VNFs) are virtualized network services that run on open computing platforms, some of which could previously be performed by proprietary, dedicated hardware technologies.

[0003] Examples of VNFs include virtualized routers, firewalls, wide area network (WAN) optimization, and network address translation (NAT) services. Most VNFs run in virtual machines (VMs) on common virtualization infrastructure software such as VMWare or KVM.

[0004] VNFs can be linked together like building blocks in a process known as service chaining to construct composite services. When it uses VNF technology, service chaining, and the application provisioning process is simplified and shortened.

[0005] Respective technical specifications include, for example, the European Telecommunications Standards Institute standard for Network Functions Virtualization (ETSI NFV).

[0006] The placement of Virtual Network Functions (VNFs) is important for optimizing the performance of a network. The VNF placement problem is to find a point of presence (PoP) of a network function in an open computing platform of a network to meet quality of service (QoS) requirements and user requests, e.g., to optimize, e.g., minimize, the latency for a UE to access the VNF and / or optimize the throughput of communication between the VNF and the UE.

[0007] Figure 1 shows an example of an overall network (home network, enterprise network and several outdoor base station networks) formed for several access hierarchies, similar to Non-Patent Document 1, the details of which are shown at the end of this specification. This 5G network may be an access network, or an access network and a private core network forming a non-public network (NPN). The management part of the VNF includes two logical entities. These logical entities can be deployed in the core network as middleware between the core network and the access network, or they can be deployed in the access network.

[0008] The two management entities in FIG. 1, although they may also be used in other cases, are: A network statistics collector that collects various measurements from the network, i.e. latency measurements, topology information, UE location, traffic parameters, or any data required to determine the VNF PoP. A dynamic VNF orchestrator that takes network data, including data from a network statistics collector, and solves a network optimization problem to determine locations or points of presence (PoPs) for a virtual network function(s) in the network that optimize the QoS requirements of the network and the requests of users.

[0009] The network optimization problem solved by the VNF orchestrator can be, for example, an integer linear programming problem (as shown in Non-Patent Document 1) that minimizes the respective objective function when the latency of access to the VNF functions is optimized.

[0010] For example, the following equations and constraints can be used to describe the VNF placement problem for multiple VNF functions:

number

[0011] That is, by solving the above optimization problem, we can find a path P from UEk to host j that minimizes the objective function of the following equation: k The binary variable X of network function i deployed on host j on i,j,k can be found.

number

[0012] Variable l i,j,k is the path P from UEk to host j. k is the latency measured for network function i located on host j above.

[0013] Variable R i is the required throughput for network function i, and the variable W j is the hardware capability of host j.

[0014] Variable θ i is the maximum latency budget required for network function i.

[0015] Variable b i,j,k is a user and j Hosted at path P k The required throughput between VNFi and the network is the variable C. m is the capacity of the link.

[0016] Several techniques can be used by the orchestrator to solve the VNF placement problem, among which is the branch-and-bound technique (as shown in Non-Patent Document 2) which divides the network into zones, tests different positions of the virtual network functions for each zone, and after testing, finds the position that minimizes the objective function.

[0017] These techniques often do not scale with an increasing number of deployed virtual functions and network growth or as the topology of the network changes (i.e., the time required to solve the optimization problem grows exponentially with the number of functions and the size of the network).

[0018] One way to reduce complexity in the orchestrator is to use a betweenness centrality metric to determine the PoPs of the virtual network functions. In this case, 1. The orchestrator finds a path in the network from the location of UEk towards a different hostj by taking into account the link capacity and hardware capabilities of the host; 2. The orchestrator places the VNF on the host that is most central in the topology, i.e., the node with maximum betweenness centrality.

[0019] This technique has lower complexity than solving an integer linear programming problem when the centrality metric is computed by Brandes' recursive algorithm (presented in [3]). [Prior art documents] [Non-patent literature]

[0020] [Non-Patent Document 1] Cziva, Richard, Christos Anagnostopoulos, and Dimitrios P. Pezaros, “Dynamic, latency-optimal vNF placement at the network edge,” Ieee infocom 2018-ieee conference on computer communications. IEEE, 2018. [Non-Patent Document 2] BARTOK, David and MANN, Zoltan Adam, “A branch-and-bound approach to virtual machine placement” (Proceedings of the 3rd HPI cloud symposium “operating the cloud. 2015. p. 49-63)” [Non-Patent Document 3] BRANDES, Ulrik, “A faster algorithm for betweenness centrality” (Journal of mathematical sociology, 2001, vol. 25, no 2, p. 163-177) Summary of the Invention [Problem to be solved by the invention]

[0021] Once a VNF's PoP is deployed, it may become necessary to deploy the VNF's PoP to a different node in the network, i.e., to relocate the VNF, depending on changes in UE distribution, traffic demands, and function invocations. However, such a move entails a certain cost.

[0022] For example, in case of centrality-based VNF deployment, the centrality of a VNF's PoPs may change with the change in UE distribution in the deployment when the network is reconfigured such that certain access points and routers are activated to serve an updated distribution, and when previously active access points and routers become inactive for traffic related to the VNF. Thus, a large change in the centrality metric may be found and thus the VNF may be reallocated, resulting in an increase in the deployment cost of the VNF (this additional cost may be, for example, in the form of an additional delay for the UE to access a new PoP of the VNF due to application context transfer between the old and new PoPs). In other words, the sensitivity of the centrality metric to changes in traffic distribution may cause a relocation of the VNF during UE movement. This may cause an increase in deployment complexity since the centrality should be recalculated, new PoPs should be determined, and the VNF should be relocated to ensure low latency at the new PoP.

[0023] The present specification aims to improve the above situation. [Means for solving the problem]

[0024] To this end, a method for use in routing traffic between at least one User Equipment UE and a Point of Presence PoP of a Virtual Network Function VNF located in a network comprising a multi-access edge computing infrastructure comprising a plurality of routers, at least some of the plurality of routers being linked, at least one link linking the plurality of routers with the PoP, the method comprising: a) obtaining an indication of which of at least two possible paths for traffic between at least one UE and a PoP, passing through a router in a plurality of routers, corresponds to a smaller estimate of an impact on the centrality of the PoP; b) thereby enabling selection of a respective path for routing traffic between at least one UE and a VNF; The present invention relates to a method, comprising:

[0025] The method may be performed, for example, by a router or a MEC server, or may be performed by a user equipment (UE) when such UE acts as a router in the network.

[0026] Therefore, the above situation is addressed by introducing traffic steering techniques among the nodes of the network that limit the impact of traffic changes on the centrality of the VNF PoP and thus reduce the probability of VNF relocation. In other words, the traffic steering techniques are used to mitigate the impact of dynamic traffic and UE distribution on the virtual network function placement metrics. The impact of traffic changes on the centrality of the VNF PoP may be limited by a given node. A negative impact on the centrality of a given node may be limited by this node, and a positive impact on the centrality of other nodes may also be limited by this node.

[0027] These techniques are to reduce the probability of relocation of a VNF's PoP over time and improve the robustness of centrality-based VNF placement to dynamic traffic and UE distribution changes.

[0028] Furthermore, the present specification defines centrality as: Betweenness centrality, Closeness centrality and Graph centrality, Stress centrality and The present invention is directed to the above-identified method, which can be based on at least one of the following:

[0029] Alternatively or additionally, in the above specified method, the smaller estimate of the effect on centrality is The past topology of the network, The current topology of the network, The expected topology of the network, and The topology is configured to change due to at least one of a handover of a UE from one access point to another access point, a switching off of a router in the plurality of routers, and a switching on of a router in the plurality of routers.

[0030] Also, in any of the above-identified variations of the method, the network may further comprise a dynamic VNF orchestrator capable of dynamically changing the placement of PoPs in the network, the method further comprising: Obtaining a signal to initiate a next change in the placement of the PoPs with a delay that is based on selecting a path having a smaller estimate of the impact on the centrality of the PoPs; Thereby, making available the transmission information required to operate the VNF from the previous PoP to the new PoP; may include.

[0031] In the above method, when the network further comprises a network statistics collector adapted to prepare network statistics, the dynamic VNF orchestrator is configured to retrieve the network statistics; The delay may be further based on network statistics that take into account traffic between the at least one UE and the VNF that is routed along the selected path.

[0032] In any of the above-identified variations of the method, obtaining the indication may include generating the indication by comparing estimates of influence on centrality of the PoPs.

[0033] The method includes, in order to obtain the indication in a), obtaining an indication of centrality of a router of the plurality of routers that is located on at least one of at least two possible paths for traffic between the at least one UE and the PoP; receiving as an input an indication of a cumulative dependency of routers of a plurality of routers located on at least one of at least two possible paths for traffic between at least one UE and a PoP; The present invention can further include at least one of the following:

[0034] In any of the above identified variants of the method, obtaining the indication in a) may be performed in a node of the network different from the router, preferably in a Mobile Edge Computing MEC server.

[0035] The above method may further comprise, after b), sending a signal to at least one router in the selected path to update its routing scheme according to the selection.

[0036] In any of the above identified variants of the method, said obtaining in a) may comprise receiving an indication at an input of the router, preferably in the form of a signal which updates a routing scheme in the router to use the selected path.

[0037] In any of the above identified variations of the method, the centrality may be based on betweenness centrality, and each estimate of the impact on a PoP's centrality due to selecting one of the at least two possible paths is calculated as a weighted sum of the change in cumulative dependency of routers on the path.

[0038] In the above, the change may include at least one of (a) a change that has occurred since a previous estimation, and (b) an expected change.

[0039] In any of the above identified variants of the method, each estimate of the effect on the centrality of a PoP due to selecting one of the at least two possible paths may be calculated by an iterative method using the Brandes algorithm.

[0040] In any of the above-identified variations of the method, the plurality of routers may include a set of intervening routers including routers estimated to contribute more than a threshold to the VNF PoP centrality, and the smaller estimate of the impact on the PoP centrality may be based on an estimate of a change in cumulative dependency of at least some of the intervening routers.

[0041] Additionally, this specification is directed to methods, UEs, routers, dynamic VNF orchestrators, network statistics collectors, and signals as described and illustrated in the figures and configured to carry out the subject matter herein.

[0042] For example, the invention also covers a computer program comprising instructions which, when the program is executed by a processing circuit, cause the processing circuit to carry out the method defined above. The invention also covers a non-transitory computer storage medium storing instructions for such a computer program.

[0043] 6 and 7 are general flow charts that may illustrate the general algorithm of such a program that may be executed by the MEC server and / or router.

[0044] For another example, the present invention also covers a device comprising a processing circuit (PROC, MEM as shown in the example of Fig. 8) configured to implement the method defined above. As further shown in Fig. 8, this device may also comprise a communication interface (COM) configured to be connected to a network.

[0045] The device may be, for example, a router or a MEC server, as described herein below, or may be a user equipment (UE) when such a UE acts as a router in the network.

[0046] Detailed descriptions and advantages of possible embodiments of the invention are given below with reference to the accompanying drawings. [Brief description of the drawings]

[0047] [Figure 1] FIG. 1 illustrates an example network comprising an example of a multi-access edge computing infrastructure. [Diagram 2] FIG. 1 illustrates a schematic diagram of an example of a network in which a VNF PoP must be located on the path between an Internet Gateway and any UE, and the placement of the VNF PoP and the Internet Gateway is subject to selection. [Diagram 3] FIG. 1 illustrates an example of a network in which traffic steering between multiple routers can be used to maintain placement of PoPs for VNFs. [Figure 4] FIG. 1 illustrates a schematic diagram of an example of a network in which the methods disclosed herein can be used to maintain the placement of a VNF's PoP in the same location for longer. [Diagram 5] 5 illustrates an example of routing enabled by a method according to an embodiment in the example network of FIG. 4. [Figure 6] 1 is a flow chart of an example of a method having steps according to a more centralized embodiment. [Figure 7] 1 is a flow chart of an example of a method having steps according to a more distributed embodiment. [Figure 8] FIG. 1 illustrates a schematic diagram of a device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0048] Virtual Network Function (VNF) placement is a key component of modern orchestration systems, and one relatively simple technique is to use a centrality metric to find VNF placement, i.e., points of presence (PoPs) in the network. This centrality-based placement is based on the contribution of routers to the path, i.e., “data path”, between the VNF’s PoP and the UE that the packets of an active user terminal follow. When traffic and / or UE distributions change dynamically, the centrality of the nodes that comprise the VNF’s PoP can change significantly. Thus, VNFs are sometimes relocated, which is an expensive process that increases the overall latency of communication between the user terminal and the VNF’s PoP. VNF relocation may also cause a degradation in the quality of service provided to the UE, as the UE is first disassociated from the previous VNF PoP before communication is re-established with the new VNF PoP.

[0049] In this disclosure, centrality may be based on at least one of: (a) betweenness centrality, (b) closeness centrality, (c) graph centrality, and (d) stress centrality.

[0050] The centralities can be calculated, for example, using the respective Brandes iterative algorithm (presented in Non-Patent Document 3).

[0051] For example, in the example of FIG. 2 (where the triangles can be sub-networks of a software defined WAN (SD-WAN)), the Brandes iteration to calculate the betweenness centrality of a node v considered as a candidate for the placement of a VNF PoP is given by the following relationship (shown in Non-Patent Document 3):

number

[0052] Variable σ s,vis the number of shortest paths between node s and node v in the network, where each node s is a candidate for placement of an Internet gateway in this example. Node w corresponds to a UE. The variable σ s,w is the number of paths between nodes s and w that are the shortest paths among the paths that pass through node v (in other words, node v is the number of paths between nodes s and w that are the shortest paths P s (w) is the predecessor of node w on (w).

[0053] Variable δ s (w) is known as the cumulative dependency of node s on node w. The cumulative dependency indicates the contribution of the path from node w to the path that reaches node s.

[0054] The placement of the VNF PoPs is selected to maximize the centrality function.

[0055] Similarly, for a VNF that performs functions related to a connection with VNF traffic between two UEs (e.g., a proxy connection) rather than a connection between a UE and an Internet gateway, the preferred placement of the VNF's PoP may be, where possible, to be located on a large portion of the shortest path between any two UEs in the deployment.

[0056] When the UE distribution and / or UE traffic is changing in the deployment area, the centrality of a VNF evolves approximately as follows:

number

[0057] Betweenness centrality ∂(C B In the above formula for the change in (v), ∂σ s,v is the change in the number of paths between node s and the VNF's PoP, and ∂σ s,w is the change in the number of paths between node s and UE w caused by the change in UE traffic distribution, and the variable ∂δ s(w) is the change in cumulative dependency of node s on node w (UE).

[0058] The method according to the invention aims to limit the change in betweenness centrality, or the different centrality metrics used to place VNF PoPs, so as to reduce the probability of PoP movement.

[0059] To this end, a router, an orchestrator, a UE and / or another node is used in routing traffic between at least one user equipment UE and a point of presence PoP of a virtual network function VNF arranged in a network comprising a multi-access edge computing infrastructure comprising a plurality of routers, at least some of the routers being linked and at least one link links the plurality of routers with the PoP, the method comprising: Obtaining an indication of which of at least two possible paths for traffic between at least one UE and a PoP and passing through a router in the plurality of routers corresponds to a smaller estimate of an impact on the centrality of the VNF to the PoP (e.g., a smaller estimate of a reduction in centrality of the VNF or a smaller increase in centrality of nodes other than the PoP to the VNF); - thereby enabling selection of a respective path for routing traffic between at least one UE and a VNF; A method can be used, including:

[0060] A "small effect" or "small estimate of effect" on centrality is here defined as -A small reduction in PoP centrality, or - A small increase in centrality of other nodes (outside PoPs), where PoP is the node with maximum betweenness centrality.

[0061] The multi-access edge computing infrastructure may conform in part or in whole to the specifications in ETSI GS MEC 003 V2.1.1(2019-01), “Multi-access Edge Computing (MEC); Framework and Reference Architecture”.

[0062] When such a method is used, the impact of changes in traffic and UE location on the topology of the network between the VNF PoP and node s (which may be a router capable of hosting an Internet Gateway function and, if required, an intermediate router for traffic passing from the Gateway through the VNF PoP and then through the UE) and, as a result, the associated centrality changes, make a small contribution to the overall change in centrality of the VNF's PoPs. The impact of changes in UE location / traffic distribution on the VNF centrality is essentially determined by the variation in the cumulative dependency of node s on node w (the UE).

[0063] That is, in the case of betweenness centrality, the change is approximately given by the following formula:

number

[0064] The ability to choose between two possible paths is only possible in very dense networks, e.g. 100m 2For example, in the case of 5G, the network can be ultra-dense in the sense defined in CHEN, Shanzhi, QIN, Fei, HU, Bo et al., "User-centric ultra-dense networks for 5G: Challenges, methodologies, and directions," IEEE Wireless Communications, 2016, vol. 23, no 2, p. 78-85.

[0065] The traffic may be, for example, Layer 3 or higher traffic, or it may be application level traffic, or IP traffic, or Ethernet traffic.

[0066] In other words, this specification discloses techniques to limit the contribution of cumulative dependency changes to the centrality of a PoP in a network where one path from at least two paths can be selected to be established from an intermediate router to a PoP. This limitation can be implemented by steering traffic from an intermediate router with a larger cumulative dependency contribution towards an intermediate router with a smaller cumulative dependency contribution. In other words, this specification discloses techniques to steer traffic between routers towards a VNF PoP such that the cumulative dependency is more balanced along the data path.

[0067] The expression "more balanced" may mean that the difference between the change in two values ​​of the cumulative dependency over time is below a threshold value in absolute value.

[0068] example: The use of traffic steering to maintain a PoP for a VNF can be illustrated with respect to a router in a line deployment with one UE present in the deployment, as shown in FIG.

[0069] In Figure 3, there are six routers R1-R6 in line. First, assume that there is a single active path from the UE towards router R2 where the VNF PoP is located. At this location, the cumulative dependency at router R2 is 1. This cumulative dependency is higher than the cumulative dependency at routers R3 and R1 because none of the UE's application packets are forwarded to routers R1 or R3, so the cumulative dependency at routers R3 and R1 is equal to 0.

[0070] When a UE moves from location (1) where it accesses the network through router R2 to location (2) where it accesses the network through router R3, in the absence of traffic steering at router R3, the PoP of the VNF must be reassigned from router R2 to router R3. The cumulative dependency at R3 increases to 1 and the cumulative dependency at R2 decreases to 0.

[0071] However, to maintain the centrality of the VNF PoP, the relocation of the VNF PoP can be prevented by activating the path forwarding packets between routers R2 and R3.

[0072] In addition, traffic can be steered between routers R2 and R3 such that a new active path between them can be added while maintaining sufficient QoS. In particular, UE-VNF traffic packets can be proactively forwarded to router R3 such that there is the same cumulative dependency on routers R2 and R3 at a given time. That is, a new path between R2 and R3 is added for the UE's packets.

[0073] A path (2,3) is determined, such as the path that balances the cumulative dependency of routers R2 and R3. In particular, in the example of FIG. 4, by activating this path for packets of traffic between the VNF and the UE, the cumulative dependency of R3 becomes equal to the cumulative dependency of R2. The packets are forwarded to the new router R3, and when the UE moves to R3, this router forwards the packets to the VNF PoP and the UE without relocating the VNF PoP.

[0074] As in the previous examples, this specification proposes embodiments based on a routing scheme that can balance (or make more balanced) the cumulative dependency of intermediate routers of a UE, for example, when these intermediate routers may be on the active data path from the UE to the VNF PoP. However, in these embodiments, the routing scheme is performed in a network with a topology such as in Figures 4 and 5, i.e. a network that does not have a linear router deployment as in Figure 3. In this network topology, there is a choice of which of at least two possible paths to activate for traffic passing through a router in the plurality of routers between at least one UE and a PoP. According to this specification, the path selected is the one that corresponds to a smaller estimate of the impact on the centrality of the PoP.

[0075] For example, in accordance with this specification, a routing scheme may be used in the network shown in Figure 4 as shown in Figure 5. When a UE moves from access point AP#1 to access point AP#2, the selection of a path from router R2 to a PoP of a VNF is made based on a comparison of estimates of the impact on the centrality of the VNF. For example, in Figure 5, the path that includes a segment directly connecting routers R2 and R1 results in the smallest estimate of impact and is shown as selected.

[0076] The routers along which the path that limits the estimate of the impact on the centrality of a VNF passes can be logical routers equipped with Packet Discovery Rules (PDRs), i.e., tables that allow the identification of other nodes (including routers) that send packets to a given router, and Forwarding Action Rules (FARs) that identify to which neighboring nodes (including routers) a given router is sending packets.

[0077] This logical router may be located, for example, in a user terminal, or a base station, or an intermediate router in the network.

[0078] The example of Figure 5 is also an example in which the network includes a dynamic VNF orchestrator that dynamically changes the placement of PoPs in the network, the orchestrator being located in a Mobile Edge Computing (MEC) server.

[0079] In this example, the orchestrator obtains, by generation, an indication of which of at least two possible paths for traffic between at least one UE and a PoP passing through a router in the plurality of routers corresponds to a smaller estimate of the impact on the centrality of the PoP.

[0080] At the same time, router R2 also receives such an indication from the orchestrator over the network.

[0081] Any other router may similarly receive such an indication with respect to router R2 or any other router used to communicate UE-VNF traffic in the event of a UE handover.

[0082] Furthermore, it may be necessary to relocate VNFs after some changes in UE distribution, and according to the present specification, such relocation can be performed with a delay that is based on the selection of paths with a smaller estimated impact on the centrality of the PoPs.

[0083] Thus, the orchestrator can generate and obtain a signal to initiate the next change in the placement of PoPs.

[0084] Each of the routers can also get by receiving a signal from the orchestrator to initiate the next change in the placement of the PoPs.

[0085] The method performed by using the orchestrator in the MEC may include the following steps. (1) The cumulative dependency δ of a multi-access edge computing (MEC) control node towards every other node s s receiving sufficient information from every router v to determine (v) and to compute the betweenness centrality of the node; (2) The orchestrator at the MEC node takes this information and sets up the VNF PoP on the node with maximum betweenness centrality. (3) The orchestrator at the MEC node determines the intermediate router as the router closest to the UE. "Close" here means having high closeness centrality to the UE node, i.e. the closeness centrality is above a threshold, or - "Nearby" means routers receiving packets from UEs located in the same geographical area, i.e. where the distance between the UE and the base station is less than a threshold, e.g. less than about 250 meters in case of very dense environments (the threshold depends on the particular deployment). (4) From the determined intermediate routers, the orchestrator at the MEC node determines the node that has the highest change in cumulative dependency as the UE moves within the area. (5) The orchestrator at the MEC node updates the packet detection rules (PDRs) and forwarding action rules (FARs) placed (i.e., recorded) in the intervening routers as follows: -Identifying neighbors for which each router has high (time) change in cumulative dependency (e.g., PDR / PTDR changes can occur when the set of neighbors changes over time), where "high change" means exceeding a threshold; - Each router forwards packets to identified neighboring routers with high variance so that the cumulative dependency variance can be reduced. (6) Proceed to Step 2 (but with the updated PDR and FAR).

[0086] The forwarding action rules (FARs) may satisfy an existing technical standard, TS 23.501, that otherwise specifies rules for forwarding packets between routers.

[0087] The packet detection rules (PDR) may satisfy existing TS 23.501 which otherwise specifies rules for detection of packets from a given router.

[0088] Furthermore, in more centralized embodiments, i.e., using an orchestrator for more steps, the techniques described above may be performed with one or more steps shown in Figure 6. For example, steps (2)-(5) above may be performed with the method of Figure 6, which shows a periodic loop that consciously updates the PDR and FAR to maintain the centrality of the VNF's PoP (this is not shown in Figure 6, but in a typical method, the next cycle in the loop (2->3->4->5->2) is not performed, e.g., if the number of cycles reaches a maximum number or if the network decides to relocate the VNF).

[0089] However, in general, an orchestrator is not required for all of the steps to perform steering, and thus this specification encompasses more decentralized embodiments as well.

[0090] For example, the method may include the following steps. (1) The step in which the centrality of the VNF is calculated through the propagation of the cumulative dependency in the network; that is, the router transmits, in the direction towards the VNF, its cumulative dependency and the cumulative dependencies of the preceding routers on the path from the UE to the VNF; the step in which the centrality is calculated at the VNF. (a) The router at each node w transmits its cumulative dependency towards all the nodes in the network that are considered for activation when the UE moves; also, the router at each node w stores a local copy of the transmitted and received cumulative dependencies. (2) The step in which the nodes w calculate their betweenness centrality based on Brandes' iterative algorithm. (3) The step in which the nodes cooperate to enable the determination of intermediate routers such as routers close to the UE. (a) Here, "close" means having a high proximity centrality with respect to the UE node, that is, the proximity centrality exceeds a threshold, and / or (b) "close" means a router that receives packets from a UE located in the same geographical area, that is, a location where the distance is less than a threshold, and / or (c) "close" can be defined by the condition that the cumulative dependency acc_d of the router exceeds a threshold A, that is, acc_d > A, which is used to determine whether the router is located "close" to the VNF PoP (the threshold "A" can be set to 1, for example; the MEC server or the network can set this threshold). (d) The intermediate router can be defined using an additional condition that the cumulative dependency acc_d of the router is less than a threshold B, that is, acc_d < B (here, the threshold B can be set to the acc_d of the VNF PoP); and / or (e) The nodes can cooperate by exchanging their cumulative dependencies in a way that is useful for their calculations and transmitting, respectively, the list of their neighbors determined to be intermediate routers. (4) From the determined intervening routers, each node w determines the node with the highest change in cumulative dependency when the UE is moving in the area. (5) Each intervening router w updates its packet detection rules (PDR) and forwarding action rules (FAR) as follows: -Each router identifies neighboring routers with high cumulative dependency change; - Each router forwards packets to the identified neighbor routers in such a way that the change in cumulative dependency is reduced. (6) Proceed to Step 2 (with the updated PDR and FAR).

[0091] Furthermore, in a more decentralized embodiment (semi-centralized embodiment), i.e., an embodiment using an orchestrator in fewer steps than the more centralized embodiment, the technique described above can be performed with one or more steps shown in FIG. 7. Overall, in the method of FIG. 7, routers on the possible paths calculate or estimate their own cumulative dependencies on the centrality of the VNF PoP and exchange signals with information about these cumulative dependencies to communicate the information about the cumulative dependencies to the MEC. The MEC sends a signal to update the PDR filter in the node to trigger communication along the selected path from some routers if, based on the received estimate of the cumulative dependencies, it is estimated that the impact on the centrality of the VNF PoP exceeds a threshold. For example, it can be estimated that the reduction in the cumulative dependency at this router is greater than a given threshold. For example, this threshold can be set for a relative cumulative dependency reduction. Not all routers on the possible path between the UE and the VNF PoP need to calculate their cumulative dependencies and / or exchange information about the cumulative dependencies with their neighbor routers (also on the possible path). These actions can be performed only on intermediate routers, which can be defined as routers whose cumulative dependency is higher than a threshold.

[0092] Thus, the methods disclosed herein that enable traffic steering to limit the probability of VNF relocation can be performed particularly efficiently by optionally determining a set of intermediate routers that contribute more than a threshold to the centrality of the VNF PoP. To that end, the set of intermediate routers can be determined as routers that have low closeness or betweenness centrality in the network (i.e., lower than a threshold) and that are used to forward traffic to the VNF PoP. Alternatively, the set of intermediate routers can be determined as routers that have low betweenness centrality and are close to the UE's current location (where "close" means having a distance less than a threshold).

[0093] A cumulative dependency can then be calculated for this set of intervening routers between the UE and the PoP of the VNF. For example, this calculation can be periodic, and a change in the cumulative dependency can be calculated for the intermediate routers.

[0094] Traffic steering is performed such that packets are routed to balance the changes in cumulative dependency between routers, such that UE(w)'s traffic is forwarded from routers with high increases in cumulative dependency towards routers with low cumulative dependency.

[0095] Instead of recalculating the centrality of a VNF, the techniques disclosed herein allow focusing on monitoring the cumulative dependencies of a subset of nodes to estimate the possible impact on the centrality of the VNF and limit the probability of VNF relocation by selecting paths with more limited impact. As a result, the complexity can also be reduced relative to a full recalculation of the centrality. Thus, the disclosed techniques can scale better than previously used techniques due to the decentralized traffic steering and the possibility of localized traffic steering based on the estimation or prediction of the cumulative dependency.

[0096] Moreover, the techniques disclosed herein can be used in conjunction with artificial intelligence designed to provide predictions of cumulative dependencies that can be used to advance update traffic steering rules at selected intervening routers when one or more UEs are expected to move while being served by a VNF.

[0097] For example, as shown in Fig. 8, a device DEV (such as a router or MEC server) may comprise a processing circuit CIR including a processor PROC and a memory MEM storing at least instruction data of a computer program according to the invention. The processor PROC can access the memory MEM in order to read and execute the instructions stored in the memory. Moreover, the device DEV further comprises at least a communication interface COM with a network NTW.

Claims

1. 1. A method for use in routing traffic between at least one User Equipment (UE) and a Point of Presence (PoP) of a Virtual Network Function (VNF) located in a network comprising a multi-access edge computing infrastructure comprising a plurality of routers, at least some of which are linked, at least one link linking the plurality of routers with the PoP, the method comprising: a) obtaining an indication of which of at least two possible paths for the traffic between the at least one UE and the PoP, passing through a router in the plurality of routers, corresponds to a smaller estimate of an impact on the centrality of the PoP; b) thereby enabling selection of the respective paths for routing the traffic between the at least one UE and the VNF; A method comprising:

2. The centrality is Betweenness centrality, Closeness centrality and Graph centrality, Stress centrality and The method of claim 1 , wherein the method is based on at least one of:

3. The smaller estimate of the influence on the centrality is A past topology of the network; and A current topology of the network; and an expected topology of the network; and Based on at least one of 3. The method of claim 1 or 2, wherein the topology is configured to change due to at least one of: a handover of a UE from one access point to another access point; a switching off of a router in the plurality of routers; and a switching on of a router in the plurality of routers.

4. When the network further comprises a dynamic VNF orchestrator capable of dynamically changing the placement of the PoPs in the network, the method further comprises: obtaining a signal to initiate a next change in the placement of the PoP with a delay that is based on the selection of the path having the smaller estimate of the influence on the centrality of the PoP; Thereby, making available the transmission information required to operate the VNF from the previous PoP to the new PoP; The method according to any one of claims 1 to 3, comprising:

5. When the network further comprises a network statistics collector adapted to prepare network statistics, the dynamic VNF orchestrator is configured to retrieve the network statistics; The method of claim 4 , wherein the delay is further based on the network statistics taking into account the traffic between the at least one UE and the VNF that is routed along the selected path.

6. The method of any one of claims 1 to 5, wherein said obtaining said indication comprises generating said indication by comparing said estimates of said influence to said centralities of said PoPs.

7. To obtain the indication in a), obtaining an indication of centrality of a router of the plurality of routers that is located on at least one of at least two possible paths for the traffic between the at least one UE and the PoP; receiving as input an indication of a cumulative dependency of routers of the plurality of routers located on at least one of at least two possible paths for the traffic between the at least one UE and the PoP; The method of claim 6 , comprising at least one of:

8. The method according to any one of claims 1 to 7, wherein the obtaining of the indication in a) is performed in a node of the network other than the router, preferably in a Mobile Edge Computing (MEC) server.

9. The method according to any one of claims 1 to 8, further comprising, after b), sending a signal to the at least one router of the selected path to update its routing strategy according to the selection.

10. 10. A method according to any preceding claim, wherein said obtaining in a) comprises receiving said indication at an input of said router, preferably in the form of a signal which updates a routing scheme in said router to use said selected path.

11. 3. The method of claim 2, wherein the centrality is based on the betweenness centrality, and each of the estimates of the impact on the centrality of the PoP due to selecting one of the at least two possible paths is calculated as a weighted sum of changes in cumulative dependency of routers on the path.

12. 12. The method according to any one of claims 1 to 11, wherein each of the estimates of the influence on the centrality of the PoP due to selecting one of the at least two possible paths is calculated by an iterative method using the Brandes algorithm.

13. 12. The method of claim 11, wherein the plurality of routers comprises a set of intervening routers including routers estimated to contribute more than a threshold to the centrality of the VNF PoP, and the smaller estimate of the influence on the centrality of the PoP is based on an estimate of the change in cumulative dependency of at least some of the intervening routers.

14. A device comprising processing circuitry for carrying out the method of any one of claims 1 to 13.

15. A computer program comprising instructions which, when executed by a processing circuit, cause the processing circuit to carry out a method according to any one of claims 1 to 13.

Images