METHOD AND NETWORK ENTITIES FOR CONTROLLING ALLOCATION OF RADIO NETWORK RESOURCES OF NETWORK SLICES IN A WIRELESS COMMUNICATION NETWORK TECHNICAL FIELD [0001] The present disclosure relates generally to methods and one or more network entities for controlling allocation of radio network resources of network slices in a wireless communication network. The present disclosure further relates to computer programs and carriers corresponding to the above methods and network entities. BACKGROUND [0002] To meet the huge demand for higher bandwidth, higher data rates and higher network capacity, due to e.g., data centric applications, existing 4
th Generation (4G) wireless communication network technology, aka Long-Term Evolution (LTE) is being extended or enhanced into a 5
th Generation (5G) technology, also called New Radio (NR) access. The following are requirements for 5G wireless communication networks: - Data rates of several tens of megabits per second should be supported for tens of thousands of users; - 1 gigabit per second is to be offered simultaneously to tens of workers on the same office floor; - Several hundreds of thousands of simultaneous connections are to be supported for massive sensor deployment; - Spectral efficiency should be significantly enhanced compared to 4G; - Coverage should be improved; - Signaling efficiency should be enhanced; and - Latency should be reduced significantly compared to 4G. [0003] One technology evolving with 5G is network slicing. As wireless communication evolves it has become clear that different wireless applications have very different demands on the network. For some applications, reliability of the connection is most important, such as security applications. For other
applications, such as video streaming, high transmission rate is most important, for yet other applications, such as machine-type communication, other demands apply. The basic idea of network slicing is to slice the network architecture in multiple logical and independent networks that are configured to effectively meet the various demands of the different applications. For example, a first network slice has network resources dedicated for providing machine-type communication, a second network slice has network resources dedicated for providing ultra- reliable low latency communication and a third network slice has network resources dedicated for providing enhanced mobile broadband content delivery. Each network slice can here be seen as an isolated end-to-end network tailored to fulfil requirements requested by particular applications. This means that in a wireless communication network, each cell, which is managed by a network node, e.g., base station, has a certain amount of its radio network resources, aka radio access network (RAN) resources set to each network slice that the wireless communication network in which the cell is a part, provides. It is possible to dedicate certain amount of RAN resources to certain slices by using for example Radio Resource Partitioning (RRP). In a wireless communication network, for instance, providing three different network slices: slice A, slice B and slice C; slice A could have 30 % of the radio network resources, slice B 60 % and slice C 10 %. [0004] A User Equipment (UE), when being connected to the wireless communication network, can communicate different traffic streams over different network slices. In essence, the UE can be connected to multiple network slices simultaneously, for instance through the use of User Equipment Route Selection Policy (URSP). Naturally, different UEs will have different traffic mixes when it comes to these network slices. For instance, one UE might communicate more traffic on certain network slices, while other UEs has a different traffic profile. Along with this, it is possible to set up RRP resource shares differently for different cells in the radio access network. This means that, for instance, one cell could have dedicated 40% of its radio network resources to Slice A and 40 % to Slice B, while a different cell might have dedicated 60% of its radio network resources to Slice A and 20 % to Slice B.
[0005] A problem with existing technology is that neither resource control of the radio network resources of the different network slices nor traffic steering, or load balancing, of the traffic over the different network slices are controlled and optimized in a holistic manner. Today, the load balancing of the traffic over the different network slices is done in a best effort manner with the purpose of maximizing the overall throughput of the cell and spectrum efficiency. The load balancing does not consider the traffic mix of the UEs, with respect to the resources of the different network slices needed, nor the available resources of the different network slices in the cells. In a similar manner, the wireless communication network does not consider a dynamic control of the resources of the different network slices in order to optimize for both the cell throughput and the given traffic mix, or traffic profile, of the UEs in the network. [0006] This brings with it a myriad of problems and suboptimizations that will only be exaggerated when the penetration of network slicing and radio resource partitions become more prominent. The main problem arises from the fact that the UEs in the network will have different traffic profiles. In other words, the UEs will have different traffic loads on the different network slices. At the same time, when dual connectivity is applied, a UE can only be connected to a single Primary Cell (PCell). This means that the radio network resources dedicated to the different network slices will have to be controlled in a way so that the traffic needs of the UEs can be satisfied. Along with this, the different UEs must be steered and load balanced in a way that harnesses this. In existing technology, a UE could be placed in a cell which cannot satisfy its given traffic profile, or the UE may miss out on a cell placement that would have satisfied the UE´s given traffic profile. [0007] One example is when a UE is utilizing multiple network slices simultaneously. In such a case, the existing technology does not ensure that the UE is placed in a cell so that all the network slices required by the UE are given a required amount of radio network resources. The existing technology mainly looks at the overall load of the cell, and if the UE is lucky, it will be placed in a cell that caters for the needs of all the network slices required by the UE. In a more loaded
scenario, the UE could be placed in a cell where only one network slice gets a good throughput, and another network slice is starved. [0008] Another example is when the UE is mainly utilizing one network slice, called Slice A. The UE might consider joining a cell where it experiences good signal strength. But if allocation is determined with an optimization criterium of downlink throughput, and the main consideration is overall traffic load, it might be the case for that cell that it has a lot of traffic on another network slice called Slice B but very little on Slice A. In other words, there is plenty of room in Slice A, but since Slice B is overutilizing its resources, and therefore, there is a high overall utilization for the cell, i.e., high overall traffic load in the cell, the UE is not moved to this cell. This is despite the fact that the UE would have received a very good downlink throughput on this cell because Slice A had very little traffic. [0009] Consequently, there is a need of an improved method for controlling allocation of radio network resources of network slices in a wireless communication network. SUMMARY [00010] It is an object of embodiments of the invention to address at least some of the problems and issues outlined above. It is an object of embodiments of the invention to control radio network resources of network slices and traffic in a wireless communication network so that allocation of the radio network resources to UEs is optimized, or at least improved, for a diverse traffic mix on different network slices. It is possible to achieve at least of one of these objects by using methods and network entities as defined in the attached independent claims. [00011] According to one aspect, a method is provided that is performed by one or more network entities for controlling allocation of radio network resources of network slices in a wireless communication network comprising a plurality of cells. The communication network has a plurality of network slices configured. The method comprises allocating, to a first UE of a plurality of UEs residing in the wireless communication network, one or more of the plurality of cells, each of the plurality of cells having pre-allocated an amount of radio network resources for
each of the plurality of network slices, and allocating to the first UE at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the one or more cells. The allocation of one or more of the plurality of cells to the first UE and the allocation of at least part of the pre- allocated amount of radio network resources to the first UE is based on need of the first UE for radio network resources of the at least one of the plurality of network slices and load of the pre-allocated amount of radio network resources of the at least one of the plurality of network slices for each of the plurality of cells. The method further comprises re-allocating the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the plurality of cells based on trends of need for radio network resources of each of the plurality of network slices for the plurality of UEs. [00012] According to another aspect, one or more network entities is provided that is configured for controlling allocation of radio network resources of a plurality of network slices configured in a wireless communication network comprising a plurality of cells. The one or more network entities comprises a processing circuitry and a memory. Said memory contains instructions executable by said processing circuitry, whereby the one or more network entities is operative for allocating, to a first UE of a plurality of UEs residing in the wireless communication network, one or more of the plurality of cells, each of the plurality of cells having pre-allocated an amount of radio network resources for each of the plurality of network slices, and allocating to the first UE at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the one or more cells. The allocation of one or more of the plurality of cells to the first UE and the allocation of at least part of the pre-allocated amount of radio network resources to the first UE is based on need of the first UE for radio network resources of the at least one of the plurality of network slices and load of the pre- allocated amount of radio network resources of the at least one of the plurality of network slices for each of the plurality of cells. The one or more network entities is further operative for re-allocating the amount of radio network resources pre- allocated to each of the plurality of network slices of each of the plurality of cells
based on trends of need for radio network resources of each of the plurality of network slices for the plurality of UEs. [00013] According to other aspects, computer programs and carriers are also provided, the details of which will be described in the claims and the detailed description. [00014] Further possible features and benefits of this solution will become apparent from the detailed description below. BRIEF DESCRIPTION OF THE DRAWINGS [00015] The solution will now be described in more detail by means of exemplary embodiments and with reference to the accompanying drawings, in which: [00016] Fig.1 is a schematic diagram of a wireless communication network in which the present invention may be used. [00017] Fig.2 is a flow chart illustrating a method performed by one or more network entities, according to possible embodiments. [00018] Fig.3 is a diagram illustrating an example of pre-allocation of resources in different network slices of cells, different resource needs of UEs as well as allocation of cells to the different UEs and of resources of those cells to the different UEs. [00019] Fig.4 is a schematic block diagram of a control framework for performing the method according to some embodiments. [00020] Fig.5 is a part of the control framework of fig.4 in more detail. [00021] Fig.6 is a block diagram illustrating the one or more network entities in more detail, according to further possible embodiments. DETAILED DESCRIPTION [00022] Fig.1 shows a wireless communication network 100 in which the present invention may be applied. The wireless communication network 100 comprises a
radio access network (RAN) node aka network node 130 that is in, or is adapted for, wireless communication with a plurality of wireless communication devices aka wireless devices or User Equipment (UE) 140, 145. The network node 130 provides radio access in a cell 150 covering a geographical area. The network node 130, or another network node, further provides radio access in another cell 155. The two cells 150, 155 may have overlapping coverage. The wireless communication network 100 further comprises other network nodes of the RAN and of a core network (CN), which are represented with box 160 in fig.1. The wireless communication network 100 further comprises or is connected to a cloud network 170. [00023] The wireless communication network 100 may be any kind of wireless communication network that can provide radio access to wireless devices. Example of such wireless communication networks are networks based on Long Term Evolution (LTE), LTE Advanced, Wireless Local Area Networks (WLAN), Worldwide Interoperability for Microwave Access (WiMAX), WiMAX Advanced, as well as fifth generation (5G) wireless communication networks based on technology such as New Radio (NR), and any possible future sixth generation (6G) wireless communication network. [00024] The network node 130 may be any kind of network node that can provide wireless access in a cell 150, 155 to a UE 140 alone or in combination with another network node. Examples of network nodes 130 are a base station (BS), a radio BS, a base transceiver station, a BS controller, a network controller, a Node B (NB), an evolved Node B (eNB), a gNodeB (gNB), a Multi-cell/multicast Coordination Entity, a relay node, an access point (AP), a radio AP, a remote radio unit (RRU), a remote radio head (RRH) and a multi-standard BS (MSR BS). [00025] The UEs 140, 145 may be any type of device capable of wirelessly communicating with a network node 130 using radio signals. For example, the UE 140 may be a machine type UE or a UE capable of machine to machine (M2M) communication, a sensor, a tablet, a mobile terminal, a smart phone, a laptop embedded equipped (LEE), a laptop mounted equipment (LME), a USB
dongle, a Customer Premises Equipment (CPE), an Internet of Things (IoT) device, etc. [00026] Embodiments described herein suggests observing the traffic mix of the UEs in the network for the different network slices, the amount of available radio network resources for the different slices, and the cost for the different cells to provide those radio network resources to the UEs. Along with this it is suggested to collect long-term trends of at least the traffic mix of the UEs but possibly also the amount of available radio network resources for the different slices, and the cost for the different cells to provide those radio network resources to the UEs. Based on this information, allocation of UEs to cells and allocation of radio network resources of network slices to the UEs is performed. [00027] Other embodiments described herein suggests optimizing/improving the selection of PCell for a UE with a certain traffic mix over the network slices. This is performed by considering the realizable radio network resources of the different network slices of available cells in the network. Moreover, as input for this selection of PCell, a specified intent or allocation optimization criterion from the wireless communication network may be applied. The output of the optimization/improvement is the optimal PCell for the UE, or each UE in the network, where optimality is defined by the previously mentioned optimization criterion. [00028] Yet other embodiments suggest controlling the amount of resources dedicated to the various network slices of the various cells in the network. This is done by considering what the typical traffic mix of UEs in the network are. Along with this it also takes as input an overarching intent or optimization criterion for the network. These two inputs are then used to find the optimal resource allocation to the different network slices in the different cells in the network. [00029] According to yet other embodiments, an optimization criterion is applied for how to control and prioritize allocation of radio network resources of network slices to UEs and/or for selection of PCell to UEs. Such an optimization criterion may be set by the operator of the network.
[00030] According to other embodiments, the amount of traffic that a UE is sending on the different slices is observed and estimated as well as the cost for the network to provide these and the overall traffic load of the different slices in the different cells in the radio access network. Based on this information allocation of cells to UEs as well as of radio network resources of network slices of the cells is controlled. Further, a mechanism to capture the long-term trends of this information is provided. [00031] According to other embodiments, an optimization criterion, set e.g., by the operator, may be combined with the information about long-term trends of traffic load in order to optimize the resource allocation to the different network slices in the different cells in the network. [00032] Fig.2, in conjunction with fig.1, describes a method performed by one or more network entities for controlling allocation of radio network resources of network slices in a wireless communication network 100 comprising a plurality of cells 150, 155. The communication network 100 has a plurality of network slices configured. The method comprises allocating 204, to a first UE 140 of a plurality of UEs 140, 145 residing in the wireless communication network 100, one or more of the plurality of cells 150, 155, each of the plurality of cells 150, 155 having pre- allocated an amount of radio network resources for each of the plurality of network slices, and allocating to the first UE 140 at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the one or more cells. The allocation of one or more of the plurality of cells to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources to the first UE 140 is based on need of the first UE 140 for radio network resources of the at least one of the plurality of network slices and load of the pre- allocated amount of radio network resources of the at least one of the plurality of network slices for each of the plurality of cells 150, 155. The method further comprises re-allocating 206 the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the plurality of cells 150, 155 based on trends of need for radio network resources of each of the plurality of network slices for the plurality of UEs 140, 145.
[00033] The step of allocating 204 signifies allocating both a first UE to cell(s) and resources of network slices of the allocated cell(s) to the first UE in one and the same step, taking into consideration the first UE´s need of radio network resources of each of the network slices and load of the radio network resources of each of the network slices of each cell. In other words, when determining the allocation of cell(s) to the first UE and the allocation of radio network resources of network slices to the first UE, the one or more network entities takes into consideration the first UE´s need of radio network resources of the different slices and the load of the pre-allocated amount of radio network resources of the different network slices for each of the plurality of cells. Further, when the “first UE” is two or more UEs, i.e., when two or more UEs are to be allocated simultaneously, the one or more network entities may determine the allocating of cell(s) and of radio network resources of the network slices to the two or more UEs at the same time. Then the optimization problem will be to do the allocation of cell(s) and the allocation of radio network resources of network slices to each of the two or more UEs based on the need of each of the two or more UEs of radio network resources of each of the network slices and the load of the pre-allocated amount of radio network resources of each of the network slices of each of the cells. As a result, the radio network resources of the network slices are used in a very efficient way. Further, by also re-allocating the pre-allocated radio network resources of the plurality of network slices of the plurality of cells based on trends of the plurality of UEs, the pre-allocation of resources of the network slices of the cells will be updated dynamically so that it follows the way the network slice demand/need changes for users of UEs in e.g., a certain geographic area. [00034] The plurality of cells may have partly overlapping coverage areas. At least some of the plurality of cells have mutually different distribution of radio network resources between the network slices. According to an embodiment, the amount of radio network resources pre-allocated to each network slice for each cell has to be larger than zero (0). However, in other embodiments, it is possible that the amount of radio network resources pre-allocated to one or more of the network slices for one or more of the cells is zero (0) while the amount of radio network resources pre-allocated to each network slice for the other of the plurality
of cells (except the “one or more cells”) is larger than zero. "Load of the pre- allocated amount of radio network resources of a network slice” signifies the amount/percentage of the radio network resources pre-allocated to the network slice of the cell that is currently used/allocated to UEs. In other words, for one cell, in case the load is 80 % for one network slice, it means that 20 % of the pre- allocated radio network resources of that network slice is still vacant and can be used for allocation to the first UE. The trends of need for radio network resources for different network slices for the plurality of UEs may be determined on statistics of usage of radio network resources of the different network slices over a shorter or longer time period. [00035] The one or more network entities that performs the method may be realized at or in one of the network nodes, i.e., base stations, that handles one of the involved cells, i.e., network node 130. Alternatively, the one or more network entities may be arranged at or in any other network node of the wireless communication network 100 in the RAN or the core network 160. Alternatively, the one or more network entities may be realized as a group of network nodes in the cloud network 170, wherein functionality of the one or more network entities is spread out over the group of network nodes. The group of network nodes may be different physical, or virtual, nodes of the network. Still alternatively, in case the wireless communication network comprises an Open Radio Access Network (O- RAN), the one or more network entities may be implemented in a node of the O- RAN, such as an O-RAN Central Unit, an O-RAN Distributed Unit or an O-RAN Radio Unit, or as a cloud solution in O-RAN. [00036] According to an embodiment, the allocation 204 of one or more of the plurality of cells 150, 155 to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices to the first UE 140 is performed more often than the re-allocation 206 of the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the plurality of cells 150, 155. The re-allocation 206 may be performed on demand, i.e., when the trends on the need for radio network resources of each of the plurality of network slices for the plurality of UEs have
changed more than a threshold since the last re-allocation. Alternatively, the re- allocation may be performed regularly, or based on an outer trigger. [00037] According to another embodiment, the allocation 204 of the first UE 140 to one or more of the plurality of cells 150, 155 and the allocation of at least a part of the amount of pre-allocated radio network resources of at least one of the plurality of network slices of the one or more cell to the first UE 140 is further based on amount of schedulable radio resources needed for each of the plurality of cells 150, 155 to deliver the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices. [00038] The radio network resources of the network slices signify e.g., bandwidth, transmission rate, latency, scheduling requests and/or compute- and memory resources, etc. that the first UE needs on each network slice. “Amount of schedulable radio resources needed” signifies on the other hand how much time- frequency resources, e.g., radio bearers, that are needed to deliver the radio network resources of the at least of the plurality of network slices that the first UE has need for, i.e. to deliver data of a network slice at the requested transmission rate/bandwidth/latency etc. For example, when a UE experiences a first low signal quality, i.e., low Signal to Interference and Noise Ratio (SINR) in a first cell, the amount of schedulable radio resources needed to deliver a certain amount of radio network resources to the UE is higher than when the same UE experiences a second signal quality of a second cell, which second signal quality is higher than the first signal quality. So, in such a case, when determining the allocation of cell and radio network resources of network slices, from the view of “amount of schedulable radio resources needed”, the UE would be allocated to the second cell rather than the first cell. However, as mentioned above, also the load of the pre-allocated amount of radio network resources of the at least one of the plurality of network slices for each of the plurality of cells needs to be taken into account when allocating the UE to cells and when allocating radio network resources of network slices to the UE, so the eventual allocation of cell and radio network resources of network slices to the UE may turn out differently.
[00039] According to yet another embodiment, the re-allocating 206 of the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the plurality of cells 150, 155 is further based on trends on the amount of schedulable radio resources needed for each cell 150, 155 to deliver radio network resources of the plurality of network slices to the plurality of UEs 140, 145. [00040] Hereby, not only trends on changes in UE needs of radio network resources of network slices are taken into consideration but also trends on changes in amount of schedulable radio resources needed for each cell to deliver the radio network resources of a network slice. As an example, if the UE need of slice A has increased over time in an area and the amount of schedulable radio resources needed to deliver an amount of radio network resources of slice A of a cell covering that area has decreased over time, a decision to reallocate radio network resources so that the amount of radio network resources of slice A for that cell is increased can be taken more quickly than when only trends on UE resource need is observed. Hereby, an even more flexible allocation of radio network resources of network slices to UEs is achieved, [00041] According to another embodiment, the allocating 204 of the first UE 140 to one or more of the plurality of cells 150, 155, comprises allocation of a Primary Cell, PCell, to the first UE 140. [00042] According to an embodiment, the allocation of a PCell to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the PCell is further based on amount of schedulable radio resources needed for a group of the plurality of cells 150, 155 to deliver the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices, wherein the group of cells consists of the cells of the plurality of cells that is/are able to provide the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices.
[00043] In case there are one or more cells here called a group of cells that can provide the first UE with all its need of radio network resources for the at least one network slice, the amount of schedulable radio resources that are needed for each of the one or more cells can be used as a criterion for selecting PCell for the first UE. The lower the amount of schedulable radio resources that is needed for a cell to deliver the need of network slice radio network resources to the first UE, the higher the probability that the cell is selected as Pcell for the first UE. [00044] According to another embodiment, the allocation of PCell to the first UE 140, when it is determined that none of the plurality of cells 150, 155 can provide the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices, is further based on amount of the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices that cannot be provided. [00045] In case none of the cells can provide the first UE with its requested radio network resources of the network slice(s) it needs, it is checked how much radio network resources that cannot be provided by each cell, and the cell that has the lowest amount of non-providable radio network resources gets the lowest so called “penalty”. In other words, the higher amount of the first UE´s need of network slice radio network resources that cannot be met by a cell, the higher the "penalty" for that cell and the lower the probability that the cell is selected as Pcell for the first UE. [00046] According to another embodiment, the method further comprises defining 202 an optimization criterion for allocation of radio network resources of the plurality of network slices in the wireless communication network 100. Further, the allocation 204 of one or more of the plurality of cells to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources to the first UE 140 is further based on the optimization criterion. By using such an optimization criterion, the operator can select how the allocation of resources should be made, i.e., if any network slices should be prioritized over other network slices or if some UEs should be prioritized over other UEs, or any other criterion that determines how the allocation should be performed.
[00047] According to another embodiment, the optimization criterion defines which of the plurality of network slices to prioritize over other of the plurality of network slices. This means that when allocating cells to the first UE and network slice radio network resources to the first UE, the allocation is to be performed so that a prioritized network slice is prioritized, i.e., UEs are firstly allocated to cells in order to fulfil their demand/need for radio network resources for the prioritized network slice. Thereafter, the first UE´s demand/need for radio network resources for other non-prioritized network slices is tried to be fulfilled. It is to be pointed out that there may be more than one optimization criterion and the optimization criteria may be co-existing. [00048] Accordig to yet another embodiment, the optimization criterion defines which of the plurality of UEs 140, 145 to prioritize over other of the plurality of UEs. This means that when allocating cells to the first UE and network slice resources to the first UE, the allocation is to be performed so that the prioritized UEs are prioritized, i.e., a prioritized UE is firstly allocated to cells in order to fulfil their demand/need for radio network resources for network slice. Thereafter, a non- prioritized UE´s demands/need for radio network resources for network slices are tried to be fulfilled. So, in case the first UE is a prioritized UE, the first UE´s need for radio network resources of network slices is fulfilled before a non-prioritized UE´s need. [00049] According to yet another embodiment, the re-allocation 206 of the pre- allocated radio network resources of the plurality of network slices of the plurality of cells 150, 155 is performed based on the optimization criterion as well as on the trends for the plurality of UEs 140, 145. [00050] Fig.3 illustrates an example allocation of UEs to cells when the network slicing concept is used. The wireless communication network has a set of network slices, Slice A, Slice B, Slice C and Slice D, configured. The wireless communication network further comprises three cells: Cell 1, Cell 2, Cell 3. The three cells each has a certain amount of resources dedicated, or pre-allocated, to the different network slices. This is illustrated by the shaded areas in the right- hand side of fig.3. The resources may be radio network resources. The cells do
not have to dedicate an equal amount of resources to all slices, but can differentiate this. In the example of fig.3, Cell 3, for example, has a much larger amount of resources dedicated to slice C than Cell 2 and Cell 1. In some embodiments, all cells must have at least some minimum amount of resources dedicated to each configured network slice. However, in other embodiments, some cells may only have dedicated resources to some network slices, or even only one of the network slices. [00051] Further, in the example of fig.3, there are four UEs: UE1, UE2, UE3 and UE4 that are allocated resources in the network from the different slices. The allocation of resources is shown in the pie charts of the left-hand side of fig.3. Note that every UE does not have allocated resources to all network slices, for example, UE2 has allocated resources of Slice B and Slice C, whereas UE4 has resources allocated for all four slices. Further, the arrows with diamond-shaped ends in fig.3 illustrate allocation of cell to the respective UE, for example allocation/selection of PCell. UE1 is allocated to Cell 2, UE2 and UE4 to Cell 3 and UE3 to Cell 1. In this illustration, one can see how the different UEs have a diverse mix of traffic flowing over the different slices. An overall purpose of embodiments proposed in this patent application is to control to which cell or cells the different UEs should be allocated and how much resources to allocate to the UEs for each slice, as well as to control how much resources each cell should dedicate, or be pre-allocated, to the different network slices. [00052] To accurately present a framework for traffic and resource control in a wireless communication network employing network slicing according to embodiments, a mathematical model and notation of the concept is presented below. Following this presentation, an embodiment of framework architecture is presented in fig.4, along with a description of each entity of the architecture and their interconnections. [00053] First, a communication network ^ = (^, ^, ^) is
where ^ = {1,2, … , ^} is the set of UEs in the network, and ^ = {1,2, … , ^} is the set of cells in the network, and ^: ^ → ^ is the mapping between the two, e.g., the
mapping of which UE is placed on which cell. In other words, the mapping ^
(^
) = ^ means that the ^-th UE is placed on the ^-th cell. Further, it is assumed that the network has a set of network slices configured, the set being defined by ^ = {1,2, … , This implies that there is a total of ^ network slices configured in the network. [00054] Here it should be noted that a network slice can be defined in a very broad sense and thus include any type of traffic or compute isolation. However, according to embodiments, a network slice is limited to a slice where the resource allocation on a cell-level is controlled through a radio resource partition framework. Naturally, the definition of a network slice could easily be extended to also include concepts such as dedicated Central Processing Units (CPUs), or virtual CPUs, virtual machines, containers, Graphics Processing Units (GPUs), memories, or other types of resources. [00055] Let us assume that the ^-th UE has a need for radio network resources of the ^-th slice given by ^
^ (^
). This radio network resource need can be specified in many possible ways, for example as bandwidth requirements, transmission rate, latency, packet loss, etc. As an example, the radio network resource need will be described in bandwidth requirements in the following. [00056] Given ^
^ (^
), the total resource need for the ^-th UE is given by: ^
^ = ∑ ^ {
^^^} ^
^(^) . A so-called cost for a cell ^ to deliver a certain amount of resources of slice ^ to a UE ^ is given by ^
^,^^^
^(^)^ = ^
^,^(^). In other words, this is the cost for the cell to provide the given UE with the amount of ^
^(^) resources for the ^-th slice. This cost is specified as schedulable radio resources. load on the ^-th slice of the ^-th cell is given by ^
^ (^
) = (
^)^^} (^) . This is simply the sum of the cost of the cell to deliver all the resources provided to all the UEs placed on this cell.
The total resource load of the ^-th cell is given by ^
^ =
∑ {^^^} ^
^(^) .
Every cell will have a certain amount of dedicated, or pre-allocated, resources to a slice. For the ^-th cell, the dedicated amount of resources for the ^-th slice is given by ^
^ ^^^ .
of radio network resources for a cell is given by ^^^^ ^ , implying that it must always hold that ^ ^^^ ^ ≤ ^
^ . [00058] It should be noted that according to embodiments, the dedicated amount of radio network resources to a certain slice is not an absolute upper bound of how much resource that particular slice can use. Instead, it can be seen as a lower bound of how much resources that slice is guaranteed to have. When using the RRP-concept to dedicate radio network resources to the different slices, it is possible for a given slice to use more than the dedicated amount. However, doing so will result in reduced priority for that slice, and thus a reduced quality-of- experience. Further, it should be noted that the total amount of radio network resources for a cell, i.e., ^^^^ ^ , is indeed a strict upper bound of the amount of resources that the cell has access to at a certain time point. However, this amount can be fluctuating over time if a cell using dynamic spectrum sharing or a fluid amount of disaggregated resources is considered. [00059] In the following, an overall architecture according to embodiments will be described with reference to fig.4. The architecture comprises a top-level so called Intent controller 402. The Intent controller 402 is the entity that allows the network operator to specify one or more governing optimization criterion which is to be propagated to two control loops of the architecture. The architecture further comprises a Traffic controller 404, which is part of the first control loop, a Resource controller 406, which is part of the second control loop and an Observer 408, which is part of the first and the second control loop. The Intent controller 402 is connected to the Traffic controller 404 via an interface a, and to the Resource controller 406 via an interface b. The first control loop comprises the Traffic controller 404, the RAN 400 of the wireless communication network and the Observer 408. The traffic controller 404 is connected to the RAN 400 via an interface e and the Observer 408 is connected to the RAN 400 via an interface g. The second control loop comprises the Resource controller 406, the Observer 408
and the RAN 400.The Resource Controller 406 is connected to the RAN 400 via an interface f, and to the Observer via an interface d. [00060] As with any proper feedback-loop the Traffic controller 404 as well as the Resource controller 406 send their respective control signals to the RAN 400 over the interfaces e and f, respectively, and receive feedback information back via the Observer 408 over interfaces g, c and g, d, respectively. It is worth noting here that the control loop of the Resource controller 406 operate on a timescale which is orders of magnitudes slower than the control loop of the Traffic Controller 404. The reason for this is to be able to separate the concerns of the two control loops, and to ensure that they operate independently without any unintended side-effects. [00061] The Intent controller 402 is the entity that allows a network operator to specify the one or more optimization criterion and the governing intent for the control of the traffic steering and resource control. These types of intents may for instance be optimization weights to specify the importance of fulfilling the UEs´ resource demands for the radio network resources of the different network slices in the network. As an example, this could imply that it is more important to fulfil the radio network resource demand for a premium slice than what it is for an economy slice. An example of an optimization criterion or intent could be that the Resource controller 406 should control the radio network resources in such a way that premium performance subscription holders are the most satisfied, i.e., that the demand/need for radio network resources of network slices of UEs of premium performance subscription holders are prioritized. In other words, the Intent controller 402 could state that the resource controller 406 should optimize the available radio network resources in such a way that the available mix of resources best matches the typical traffic mix of the UEs holding a premium performance subscription and disregarding the traffic mix of other UEs of the network. The one or more optimization criterion is sent by the Intent controller 402 to the Traffic controller 404 via interface a and to the Resource controller 406 via interface b. [00062] The Traffic controller 404 is the entity that controls which cell the different UEs should be placed on, or allocated to. In essence, the Traffic controller 404
controls how traffic, i.e., data between the UEs and the network, should flow in the different network slices via the different cells in the network. Using the mathematical notation presented above, the goal of the Traffic controller 404 is to control ^, i.e., the mapping between the UEs and the cells. To do this, the Traffic controller 404 takes as input the one or more optimization criterion from the Intent Controller 4023 received over the interface a. Along with this, the Traffic controller 404 obtains information from the Observer 408 about radio network resource need of the different network slices of the UEs (e.g., ^
^(^)), the cost to deliver those (e.g., well as the resource loads on the cells (e.g., ^
^(^) and ^
^). [00063] According to an embodiment, and as shown in fig.5, the Traffic controller 404 comprises two parts, a UE handover controller 404a and a PCell controller 404b. [00064] The purpose of the PCell Controller 404b is to find the best PCell for a given UE. Given the notation presented above, this would be described as finding the best mapping ^(^) =
where ^
∗ indicate the best, or optimal, cell in which to place the ^-th UE. Naturally, there are many ways to define what an optimal placement is, and this is where the input from the Intent controller 402 comes into place. The Intent controller 402 communicates a specific intent or optimization criterion to the PCell Controller 404b, which optimization criterion defines guidelines to what the optimal placement is. There are many possible optimization objectives or criteria that can be considered. However, if the ^-th UE is to be placed on a cell ^ that can cater for all the radio network resource needs of all the network slices of one UE,
,
where the cost of the cell to provide the resource needs of ^
^(^) for the ^-th slice. [00065] In the following, an embodiment of selecting PCell to a UE called Scenario A is described. In Scenario A there exists one or more cells that can provide all radio network resources of network slices that the UE needs. In this embodiment, the PCell Controller 404b determines a weight-vector, which
dictates the cost of a cell for providing resources to a certain slice w
^(k). An interpretation of such a cost weight-vector could be that it is more “expensive” for a cell ^ to provide resources to slice ^ than what it would be for another cell ^′ to do so. An implication of such a weight-vector would be that it can allow UEs that are using more of one type of network slice radio network resources to be co- located in a certain cell. Or in other words, the cost weight-vector could make it more expensive for economy users/UEs to use resources in the lower frequency bands, such as the 700MHz band than in higher frequency bands, such as the 3.5 GHz band. For those economy users it would be “cheaper” to use the 3.5GHz band. On the other hand, allocating resources to the premium user/UE on the 700Mhz could be cheaper. The optimization criterion for finding the best PCell could then be stated as finding the
min ^ ^
∗i∈m^ize
such that
∀ ^ = 1,2, … , ^. [00066] In the following, an embodiment of selecting PCell to a UE called Scenario B is described. In Scenario B there does not exist any cell that can provide all radio network resources of network slices that the UE needs. Naturally, there might not always be the case that there exists a cell that can cater for all the resource needs of the UE. In such a case, the PCell Controller 404b determines a penalty for each cell for not meeting the resource needs, given by p
^ (k
). Again, this is defined per cell to allow the Intent Controller 402 to drive the traffic steering behavior for different UEs and network slices to certain types of cells, along the same reasoning as provided in the comment above. Assuming that the penalty for the ^-th cell to not meet the resource needs of a network slice ^ is given by ^
^(^), then the optimization problem of finding the best min
∗im^ize
^ ∈ , ^^^^ ^^ ^^^^
such that
∑ ^^^ ^
^,^ ∗(^
) + ^
^ ∗(^) ≤ ^
^ .
where
[⋅
]^ ensures that it is a non-negative number and where ^
^,^ ∗(^
) + ^
^ ∗(^
) − ^
^ ^∗
^^(^) is the amount of the resource needs for the ^-th UE
be met by the cell. The bottom constraint in the equation above is to ensure that the total amount of resources for the cell is not exceeding its maximum amount of resources. The intuition of the above optimization objective is that it penalizes the amount of resource needs that cannot be met for the UE. It then selects the cell which yields the least amount of penalty as the cell to which to allocate the UE. [00067] Naturally, it would be possible to combine the optimization objectives of Scenario A and Scenario B into one optimization objective. A way to find the best possible PCell (i.e., ^
∗) would be to first evaluate the optimization problem under Scenario A, and if that would not yield any placement for the UE, then the PCell Controller 404b would continue and evaluate the optimization problem under Scenario B. [00068] The purpose of the UE Handover controller 404a is to monitor the mapping of the different UEs in the network to cells and to ensure that the current mapping of UEs to cells (i.e.,
) does not differ too far from an optimal UE mapping f
∗ which can be considered the ideal or best possible placement of the UEs. There are many reasons why the current mapping of the UEs to the cells might deviate from the ideal one, even though all UEs may be placed using the optimization objectives presented in the previous section. One reason is that the UEs arrive to the network in a sequence, and since they are then also mapped to cells in a sequence it may lead to an overall suboptimal mapping of UEs to cells. Another reason could be that the UEs move around, and so the cost of providing a UE with a certain amount of radio network resources will change over time, even though their resource needs does not change. A third reason could be that the dynamic and will change over time. In other words,
since ^
^ , , all change dynamically over time, so will an optimal UE-to-cell mapping ^
∗. [00069] To ensure that the network operates in an efficient manner, and to ensure that the current UE-to-cell mapping ^ remains close to the optimal UE-to-
cell mapping ^
∗, the UE Handover Controller 404a will monitor the network and trigger handover events to specific UEs in the network. This can be expressed as the desire to ensure that
|^
∗ − ^
| < ^ℎ^^^ℎ^^^ where the given threshold could be specified by the Intent Controller 402. Naturally, it is difficult to derive a concrete metric for
|^
∗ − ^
| so instead the UE Handover Controller 404a may monitor other metrics, such as: ^
^ ⋅ ^
^
c) ^
^ > ^^^^ ^ − threshold. In the options above, metric (a) would monitor the weighted relative cost of providing resource needs. This can be useful in order to catch a UE which is draining an unnecessary amount of resources from a cell. In such a scenario, there may be a different cell that could fulfill these needs to a lower cost, for instance because it may have a lower path loss. If so, the UE Handover Controller 404a would trigger handover from this UE to the cell that can fulfil the UE’s radio network resource needs to a lower cost. Metric (b) would monitor the resource usage for the particular slices in a cell. Should this grow larger than a given threshold, then the UE handover Controller 404a can check which UE is relatively the most expensive using metric (a), and trigger handover of this UE. Metric (c) would monitor the overall resource usage of the entire cell. Should the cell load grow too close to its maximum capacity, then using (b) and (a), the UE Handover Controller 404a can identify the most loaded slice as well as the UE which has the highest relative cost and trigger a handover event for this UE. Therefore, with the above three metrics, the UE Handover Controller 404a can ensure that the UE-to- cell mapping remains close to the ideal UE-to-cell mapping. [00070] The Resource Controller 406 of fig.4 is the entity that controls the amount of dedicated resources to the different network slices in the different cells. Using the mathematical notation from above, the goal of the Resource Controller 406 is to control ^
^ ^^^(^
) for the different network slices in the different cells. To do this, the Resource Controller 406 takes as input the optimization criterion from the Intent Controller 402. Along with this, the Resource Controller 406 also obtains information from the Observer 408 about long-term trends of the UEs´ resource
needs and optionally also the costs for the cells to deliver those, cost in terms of amount of schedulable radio resources needed for each cell to deliver radio network resources of the plurality of network slices to the plurality of UEs. [00071] In the following, an embodiment of a resource allocation algorithm used by the Resource Controller 406 is described. The resource allocation algorithm can be seen essentially as an offline vector scheduling problem in d dimensions. [00072] Considering the mathematical notation provided above, the input that the Resource Controller 406 obtains from the Observer 408 ^
^̂ . The
former (i.e., ^
^̂(^)) is defined as the long-term trend for the i-th “typical” UEs resource needs on the k-th slice. The latter ) is defined as the long-term
trend for the j-th cell to provide the i-th UE with the resources it needs. Here it is worth commenting that a “typical UE” with a “typical traffic mix” is a statistical notion and can be created and interpreted in multiple ways. However, one embodiment of such a notion would be to let the set ^ consist of UEs that are the most probable ones to co-exist in the network at the busiest 10% of the hours. ^
^̂(^) would then capture the long-term trend for the set ^. Further, the notion of long-term trend is also a statistical notion and can be embodied in many different ways. One such embodiment would be by using a sliding-window approach or a K- means approach combined with a low-pass filter of the resource needs of the different UEs. This is discussed at greater extent in the section describing the Observer 408. [00073] As stated above, finding an optimal resource allocation to the different network slices on the different cells can be seen as an offline vector scheduling problem in ^ dimensions over ^ bins, where ^ is the number of Cells in the network. This resource allocation is given by finding the best
resource mapping → such a mapping has been found, the optimal for the ^-th slice on the ^-th cell can be derived by ^
^ ^^^( ^ ) =
∑ {^:^(^)^^} ^
^,^(^) . Once this optimal resource allocation has been found, it will be realized by communicating the desired resource allocation to the wireless communication system, i.e., the corresponding base stations in the corresponding
cells, over the interface f in fig.4. In order to find the optimal resource mapping, the optimization criterion needs to be stated first. It is assumed that this will be provided by the Intent Controller 402 over the interface b. One example of such an optimization criterion could be to minimize the maximum resource cost of a slice in a cell in terms of amount of schedulable radio resources needed to deliver the need of radio network resources. The advantage of such a criterion would be to spread the resource load of the UEs evenly across the different cells. This objective could be specified as: minimize
where || ⋅ |
| ^ is the infinity-norm, given by {^∈ m^,^a∈x^} ^
^(^). Another possible optimization objective would be to instead minimize the relative cost of providing resources: minimize ∑
{^∈^} ∑
{^:^(^)^^} ∑ ^ ^
^^ ^
^̂,^(^)⁄ ^
^̂(^) . Finally, there can also be a weighted objective criterion, to make it more costly for certain cells to provide certain resources to certain slices. As an example, it could be expensive for a lower-band cell, e.g., a 700Mhz cell, to provide resources to an economy slice. On the other hand, it might be cheap to do so for a higher-band cell, e.g., a 3.5Ghz cell. Moreover, for the lower-band cell it might be cheaper to provide resources to the premium slice than to the economy slice. This way, it is possible for the Intent Controller 402 to control the high-level behavior of the network, e.g., which slice-traffic should be flowing on which type of cells. An example of such a criterion could be: minimize
) which is simply a weighted version of the relative cost of providing resources. Finding the solutions to these optimization problems can be done through standard methods such as a linear program or a genetic algorithm. [00074] While the previous section illustrated how the resource allocation, or resource re-allocation, is done, the question remains when to do the resource allocation. This action can be triggered in many different ways. The simplest option is to trigger it periodically. In other words, to trigger a resource re-allocation once
every x hours, days, or weeks. A more advanced option would be to instead have an event-based trigger. This could be embodied by checking if current resource allocation has deviated too far, i.e., above a certain threshold, from what would be an optimal allocation given the current ^
^̂(^) and (^). Naturally, a final option is to also involve a Machine Learning (ML)- or
(AI)-based solution to identify when the current resource allocation has deviated to far from the optimal resource allocation. [00075] The final component of fig.4, the Observer 408, is the entity that obtains, for example pulls, needed information from the radio part of the wireless communication network 400, analyzes the information, and compiles the information into a format needed by the Traffic Controller 404 and the Resource Controller 406. Using the mathematical notation above, the Observer 408 would send information to the Traffic Controller 404 about ^
^ (^
), ^
^,^ (^
),
. To the Resource Controller 406 it would send information about ^
^̂(^) and ^
^̂,^(^) where the “^” indicates estimations of the long-term trend or average of the “typical” UEs in the network, their “typical” resource mix, as well as the “typical” cost of delivering resources to those UEs. The Observer 408 communicates with the Traffic Controller 404 over the interface c and with the Resource Controller 4067 over the interface d. [00076] The observation of the UE resource need is relatively straight forward, however requires some care. With the notation above, this is the observation of ^
^(^), which means the ^-th UEs need for radio network resources of the ^-th slice. Note again that these resources can be denoted in any entity, but in this proposed solution the focus will be on bandwidth and/or transmission rate, but it could just as well be denoted in latency, or contention-delay, or CPU-cycles, or GPU-cycles, or FPGA-configuration and cycles. [00077] To identify and observe this metric the Observer 408 will interface with the radio access network and the core network of the wireless communication network in order to establish measurements of the data rate being sent from/to the UE. It should be noted that this example does not discuss any use of carrier
aggregation (CA). However, should CA be a factor, it is easy to extend the proposed mechanisms to such a scenario as well. In a case CA is used by a UE, the UE will communicate with the network over multiple cells simultaneously. In such a case, the mapping ^: ^ → ^ will be of type one-to-many, i.e., ^
(^
) = {^
^, ^
^, ^
^, … } should the ^-th UE send traffic over cells ^
^, ^
^, and ^
^. The resource need of the ^-th UE can then be observed and expressed as: ^
^ = ,
the resource on the ^-th cell.
be on the resource needs for the downlink, but the same principles, mechanisms, and methodology also applies to the uplink. [00078] The observation of the cost for a cell ^ to provide a certain amount of radio network resources ^
^ (^
) to a UE, i.e., to observe ^
^,^(^), is a bit more intricate. The main reason for this is because, at least when it comes to catering to the bandwidth needs, the signal quality of the radio link between the UE and the base station matters. Therefore, this would need to be estimated according to a function similar to
, where CSI
^,^ is the Channel State Information (CSI)-report that contains information about the signal quality, Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), etc. Should the radio network resource need ^
^ (^
) instead be related to a latency requirement, the function to estimate the resource cost would also have to consider the other UEs sending traffic on the cell. In other words, it would have to consider queue-sizes of the traffic flows of the other UEs. [00079] The observation of the resource load of a cell ^ for a slice ^ is performed by aggregating the observation of the resource cost for all the UEs sending traffic over the ^-th slice:
=
(^)^^} . These observations when the considered resources are bandwidth, would correspond to aggregating the symbols or the physical resource blocks that the cell provides to the ^-th slice. Should the considered resource be of a different
type, for instance latency, then the cost could be measured through the volume of resources that the cell has to increase the priority for. Should the considered resources be for CPU-cycles, or GPU-cycles then it could simply be an aggregate of the number of cycles provided to the different slices. [00080] When it comes to estimating the long-term averages of the resource need and the cost to deliver those resource ^
^̂ and ^
^̂,^(^),
respectively, then the proposed mechanisms to use standard statistical methods for generating long-term trends through a combination of low-pass filters and sliding-window filters. This could possibly also be combined with K-means regression to allow the different “archetype” or “cohorts” of UEs/users to be collected. [00081] Fig.6, in conjunction with fig.1, describes one or more network entities 600 configured for controlling allocation of radio network resources of a plurality of network slices configured in a wireless communication network 100 comprising a plurality of cells 150, 155. The one or more network entities 600 comprises a processing circuitry 603 and a memory 604. Said memory contains instructions executable by said processing circuitry, whereby the one or more network entities 600 is operative for allocating, to a first UE 140 of a plurality of UEs 140, 145 residing in the wireless communication network 100, one or more of the plurality of cells 150, 155, each of the plurality of cells 150, 155 having pre-allocated an amount of radio network resources for each of the plurality of network slices, and allocating to the first UE 140 at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the one or more cells. The allocation of one or more of the plurality of cells to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources to the first UE 140 is based on need of the first UE 140 for radio network resources of the at least one of the plurality of network slices and load of the pre- allocated amount of radio network resources of the at least one of the plurality of network slices for each of the plurality of cells 150, 155. The one or more network entities 600 is further operative for re-allocating the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the
plurality of cells 150, 155 based on trends of need for radio network resources of each of the plurality of network slices for the plurality of UEs 140, 145. [00082] The one or more network entities 600 may be realized at or in one of the network nodes 130, i.e. base stations, that handles one of the involved cells. Alternatively, the one or more network entities 600 may be arranged at or in any other network node of the wireless communication network 100, such as in any other radio access network (RAN) node or in any core network node 160. Alternatively, the one or more network entities 600 may be realized as a group of network nodes, wherein functionality of the one or more network entities 600 is spread out over the group of network nodes. The group of network nodes may be different physical, or virtual, nodes of the network. The group of network nodes may be arranged in the cloud network 170. Still alternatively, in case the wireless communication network comprises an Open Radio Access Network (O-RAN), the one or more network entities may be implemented in a node of the O-RAN, such as an O-RAN Central Unit, an O-RAN Distributed Unit or an O-RAN Radio Unit, or in a cloud network connected to the O-RAN. [00083] According to an embodiment, the one or more network entities 600 is operative for performing the allocating of one or more of the plurality of cells 150, 155 to the first UE 140 and the allocating of at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices to the first UE 140 more often than the re-allocating of the amount of radio network resources pre-allocated to each of the plurality of network slices of each of the plurality of cells 150, 155. [00084] According to an embodiment, the one or more network entities 600 is operative for the allocating of the first UE 140 to one or more of the plurality of cells 150, 155 and the allocating of at least a part of the amount of pre-allocated radio network resources of at least one of the plurality of network slices of the one or more cell to the first UE 140 further based on amount of schedulable radio resources needed for each of the plurality of cells 150, 155 to deliver the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices.
[00085] According to another embodiment, the one or more network entities 600 is operative for the re-allocating of the amount of radio network resources pre- allocated to each of the plurality of network slices of each of the plurality of cells 150, 155 further based on trends on the amount of schedulable radio resources needed for each cell 150, 155 to deliver radio network resources of the plurality of network slices to the plurality of UEs 140, 145. [00086] According to another embodiment, the one or more network entities 600 is operative for the allocating 204 of the first UE 140 to one or more of the plurality of cells 150, 155 by allocation of a Primary Cell, PCell, to the first UE 140. [00087] According to yet another embodiment, the one or more network entities 600 is operative for the allocation of a PCell to the first UE 140 and the allocation of at least part of the pre-allocated amount of radio network resources of at least one of the plurality of network slices of the PCell further based on amount of schedulable radio resources needed for a group of the plurality of cells 150, 155 to deliver the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices, wherein the group of cells consists of the cells of the plurality of cells that is/are able to provide the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices. [00088] According to yet another embodiment, the allocation of PCell to the first UE 140, when it is determined that none of the plurality of cells 150, 155 can provide the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices, is further based on amount of the need of the first UE 140 for radio network resources of the at least one of the plurality of network slices that cannot be provided. [00089] According to still another embodiment, the one or more network entities 600 is further operative for defining an optimization criterion for the allocation of radio network resources of the plurality of network slices in the wireless communication network 100. Further, the one or more network entities 600 is operative for the allocating of one or more of the plurality of cells to the first UE
140 and the allocating of at least part of the pre-allocated amount of radio network resources to the first UE 140 further based on the optimization criterion. [00090] According to one embodiment, the optimization criterion defines which of the plurality of network slices to prioritize over other of the plurality of network slices. [00091] According to another embodiment, the optimization criterion defines which of the plurality of UEs 140, 145 to prioritize over other of the plurality of UEs. [00092] According to yet another embodiment, the one or more network entities 600 is operative for the re-allocating of the pre-allocated radio network resources of the plurality of network slices of the plurality of cells 150, 155 based on the optimization criterion as well as on the trends for the plurality of UEs 140, 145. [00093] According to other embodiments, the one or more network entities 600 may further comprise conventional means for communication with other network nodes of the wireless communication network 100. In case the one or more network entities is arranged in the network node 130, the communication unit 602 comprises conventional means for wireless communication with the UEs 140,145, such as a transceiver for wireless transmission and reception of signals in the communication network. The instructions executable by said processing circuitry 603 may be arranged as a computer program 605 stored e.g. in said memory 604. The processing circuitry 603 and the memory 604 may be arranged in a sub- arrangement 601. The sub-arrangement 601 may be a micro-processor and adequate software and storage therefore, a Programmable Logic Device, PLD, or other electronic component(s)/processing circuit(s) configured to perform the methods mentioned above. The processing circuitry 603 may comprise one or more programmable processor, application-specific integrated circuits, field programmable gate arrays or combinations of these adapted to execute instructions. [00094] The computer program 605 may be arranged such that when its instructions are run in the processing circuitry 603, the instructions cause the one
or more network entities 600 to perform the steps described in any of the described embodiments of the one or more network entities 600 and its method. The computer program 605 may be carried by a computer program product connectable to the processing circuitry 603. The computer program product may be the memory 604, or at least arranged in the memory. The computer program product may be called a computer-readable storage medium 606. The memory 604 may be realized as for example a Random-access memory (RAM), Read-Only Memory (ROM) or an Electrical Erasable Programmable ROM (EEPROM). In some embodiments, a carrier may contain the computer program 605. The carrier may be one of an electronic signal, an optical signal, an electromagnetic signal, a magnetic signal, an electric signal, a radio signal, a microwave signal, or computer readable storage medium. The computer-readable storage medium 606 may be e.g., a CD, DVD or flash memory, from which the program could be downloaded into the memory 604. Alternatively, the computer program 605 may be stored on a server or any other entity to which the one or more network entities 600 has access via the communication unit 602. The computer program 605 may then be downloaded from the server into the memory 604. [00095] Although the description above contains a plurality of specificities, these should not be construed as limiting the scope of the concept described herein but as merely providing illustrations of some exemplifying embodiments of the described concept. It will be appreciated that the scope of the presently described concept fully encompasses other embodiments which may become obvious to those skilled in the art, and that the scope of the presently described concept is accordingly not to be limited. Reference to an element in the singular is not intended to mean "one and only one" unless explicitly so stated, but rather "one or more." All structural and functional equivalents to the elements of the above- described embodiments that are known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed hereby. Moreover, it is not necessary for an apparatus or method to address each and every problem sought to be solved by the presently described concept, for it to be encompassed hereby. In the exemplary figures, a broken line generally signifies that the feature within the broken line is optional.
The total resource load of the ^-th cell is given by ^^ = ∑ {^^^} ^^(^) .
Every cell will have a certain amount of dedicated, or pre-allocated, resources to a slice. For the ^-th cell, the dedicated amount of resources for the ^-th slice is given by ^^ ^^^ .
of radio network resources for a cell is given by ^^^^ ^ , implying that it must always hold that ^ ^^^ ^ ≤ ^^ . [00058] It should be noted that according to embodiments, the dedicated amount of radio network resources to a certain slice is not an absolute upper bound of how much resource that particular slice can use. Instead, it can be seen as a lower bound of how much resources that slice is guaranteed to have. When using the RRP-concept to dedicate radio network resources to the different slices, it is possible for a given slice to use more than the dedicated amount. However, doing so will result in reduced priority for that slice, and thus a reduced quality-of- experience. Further, it should be noted that the total amount of radio network resources for a cell, i.e., ^^^^ ^ , is indeed a strict upper bound of the amount of resources that the cell has access to at a certain time point. However, this amount can be fluctuating over time if a cell using dynamic spectrum sharing or a fluid amount of disaggregated resources is considered. [00059] In the following, an overall architecture according to embodiments will be described with reference to fig.4. The architecture comprises a top-level so called Intent controller 402. The Intent controller 402 is the entity that allows the network operator to specify one or more governing optimization criterion which is to be propagated to two control loops of the architecture. The architecture further comprises a Traffic controller 404, which is part of the first control loop, a Resource controller 406, which is part of the second control loop and an Observer 408, which is part of the first and the second control loop. The Intent controller 402 is connected to the Traffic controller 404 via an interface a, and to the Resource controller 406 via an interface b. The first control loop comprises the Traffic controller 404, the RAN 400 of the wireless communication network and the Observer 408. The traffic controller 404 is connected to the RAN 400 via an interface e and the Observer 408 is connected to the RAN 400 via an interface g. The second control loop comprises the Resource controller 406, the Observer 408
and the RAN 400.The Resource Controller 406 is connected to the RAN 400 via an interface f, and to the Observer via an interface d. [00060] As with any proper feedback-loop the Traffic controller 404 as well as the Resource controller 406 send their respective control signals to the RAN 400 over the interfaces e and f, respectively, and receive feedback information back via the Observer 408 over interfaces g, c and g, d, respectively. It is worth noting here that the control loop of the Resource controller 406 operate on a timescale which is orders of magnitudes slower than the control loop of the Traffic Controller 404. The reason for this is to be able to separate the concerns of the two control loops, and to ensure that they operate independently without any unintended side-effects. [00061] The Intent controller 402 is the entity that allows a network operator to specify the one or more optimization criterion and the governing intent for the control of the traffic steering and resource control. These types of intents may for instance be optimization weights to specify the importance of fulfilling the UEs´ resource demands for the radio network resources of the different network slices in the network. As an example, this could imply that it is more important to fulfil the radio network resource demand for a premium slice than what it is for an economy slice. An example of an optimization criterion or intent could be that the Resource controller 406 should control the radio network resources in such a way that premium performance subscription holders are the most satisfied, i.e., that the demand/need for radio network resources of network slices of UEs of premium performance subscription holders are prioritized. In other words, the Intent controller 402 could state that the resource controller 406 should optimize the available radio network resources in such a way that the available mix of resources best matches the typical traffic mix of the UEs holding a premium performance subscription and disregarding the traffic mix of other UEs of the network. The one or more optimization criterion is sent by the Intent controller 402 to the Traffic controller 404 via interface a and to the Resource controller 406 via interface b. [00062] The Traffic controller 404 is the entity that controls which cell the different UEs should be placed on, or allocated to. In essence, the Traffic controller 404
controls how traffic, i.e., data between the UEs and the network, should flow in the different network slices via the different cells in the network. Using the mathematical notation presented above, the goal of the Traffic controller 404 is to control ^, i.e., the mapping between the UEs and the cells. To do this, the Traffic controller 404 takes as input the one or more optimization criterion from the Intent Controller 4023 received over the interface a. Along with this, the Traffic controller 404 obtains information from the Observer 408 about radio network resource need of the different network slices of the UEs (e.g., ^^(^)), the cost to deliver those (e.g., well as the resource loads on the cells (e.g., ^^(^) and ^^). [00063] According to an embodiment, and as shown in fig.5, the Traffic controller 404 comprises two parts, a UE handover controller 404a and a PCell controller 404b. [00064] The purpose of the PCell Controller 404b is to find the best PCell for a given UE. Given the notation presented above, this would be described as finding the best mapping ^(^) =
where ^∗ indicate the best, or optimal, cell in which to place the ^-th UE. Naturally, there are many ways to define what an optimal placement is, and this is where the input from the Intent controller 402 comes into place. The Intent controller 402 communicates a specific intent or optimization criterion to the PCell Controller 404b, which optimization criterion defines guidelines to what the optimal placement is. There are many possible optimization objectives or criteria that can be considered. However, if the ^-th UE is to be placed on a cell ^ that can cater for all the radio network resource needs of all the network slices of one UE,
,
where the cost of the cell to provide the resource needs of ^^(^) for the ^-th slice. [00065] In the following, an embodiment of selecting PCell to a UE called Scenario A is described. In Scenario A there exists one or more cells that can provide all radio network resources of network slices that the UE needs. In this embodiment, the PCell Controller 404b determines a weight-vector, which
dictates the cost of a cell for providing resources to a certain slice w^(k). An interpretation of such a cost weight-vector could be that it is more “expensive” for a cell ^ to provide resources to slice ^ than what it would be for another cell ^′ to do so. An implication of such a weight-vector would be that it can allow UEs that are using more of one type of network slice radio network resources to be co- located in a certain cell. Or in other words, the cost weight-vector could make it more expensive for economy users/UEs to use resources in the lower frequency bands, such as the 700MHz band than in higher frequency bands, such as the 3.5 GHz band. For those economy users it would be “cheaper” to use the 3.5GHz band. On the other hand, allocating resources to the premium user/UE on the 700Mhz could be cheaper. The optimization criterion for finding the best PCell could then be stated as finding the
min ^ ^∗i∈m^ize
such that
∀ ^ = 1,2, … , ^. [00066] In the following, an embodiment of selecting PCell to a UE called Scenario B is described. In Scenario B there does not exist any cell that can provide all radio network resources of network slices that the UE needs. Naturally, there might not always be the case that there exists a cell that can cater for all the resource needs of the UE. In such a case, the PCell Controller 404b determines a penalty for each cell for not meeting the resource needs, given by p^ (k). Again, this is defined per cell to allow the Intent Controller 402 to drive the traffic steering behavior for different UEs and network slices to certain types of cells, along the same reasoning as provided in the comment above. Assuming that the penalty for the ^-th cell to not meet the resource needs of a network slice ^ is given by ^^(^), then the optimization problem of finding the best min∗im^ize
^ ∈ , ^^^^ ^^ ^^^^
such that ∑ ^^^ ^^,^ ∗(^) + ^^ ∗(^) ≤ ^^ .
where [⋅]^ ensures that it is a non-negative number and where ^^,^ ∗(^) + ^^ ∗(^) − ^^ ^∗^^(^) is the amount of the resource needs for the ^-th UE
be met by the cell. The bottom constraint in the equation above is to ensure that the total amount of resources for the cell is not exceeding its maximum amount of resources. The intuition of the above optimization objective is that it penalizes the amount of resource needs that cannot be met for the UE. It then selects the cell which yields the least amount of penalty as the cell to which to allocate the UE. [00067] Naturally, it would be possible to combine the optimization objectives of Scenario A and Scenario B into one optimization objective. A way to find the best possible PCell (i.e., ^∗) would be to first evaluate the optimization problem under Scenario A, and if that would not yield any placement for the UE, then the PCell Controller 404b would continue and evaluate the optimization problem under Scenario B. [00068] The purpose of the UE Handover controller 404a is to monitor the mapping of the different UEs in the network to cells and to ensure that the current mapping of UEs to cells (i.e.,
) does not differ too far from an optimal UE mapping f ∗ which can be considered the ideal or best possible placement of the UEs. There are many reasons why the current mapping of the UEs to the cells might deviate from the ideal one, even though all UEs may be placed using the optimization objectives presented in the previous section. One reason is that the UEs arrive to the network in a sequence, and since they are then also mapped to cells in a sequence it may lead to an overall suboptimal mapping of UEs to cells. Another reason could be that the UEs move around, and so the cost of providing a UE with a certain amount of radio network resources will change over time, even though their resource needs does not change. A third reason could be that the dynamic and will change over time. In other words,
since ^^ , , all change dynamically over time, so will an optimal UE-to-cell mapping ^∗. [00069] To ensure that the network operates in an efficient manner, and to ensure that the current UE-to-cell mapping ^ remains close to the optimal UE-to-
cell mapping ^∗, the UE Handover Controller 404a will monitor the network and trigger handover events to specific UEs in the network. This can be expressed as the desire to ensure that |^∗ − ^| < ^ℎ^^^ℎ^^^ where the given threshold could be specified by the Intent Controller 402. Naturally, it is difficult to derive a concrete metric for |^∗ − ^| so instead the UE Handover Controller 404a may monitor other metrics, such as: ^^ ⋅ ^^
c) ^^ > ^^^^ ^ − threshold. In the options above, metric (a) would monitor the weighted relative cost of providing resource needs. This can be useful in order to catch a UE which is draining an unnecessary amount of resources from a cell. In such a scenario, there may be a different cell that could fulfill these needs to a lower cost, for instance because it may have a lower path loss. If so, the UE Handover Controller 404a would trigger handover from this UE to the cell that can fulfil the UE’s radio network resource needs to a lower cost. Metric (b) would monitor the resource usage for the particular slices in a cell. Should this grow larger than a given threshold, then the UE handover Controller 404a can check which UE is relatively the most expensive using metric (a), and trigger handover of this UE. Metric (c) would monitor the overall resource usage of the entire cell. Should the cell load grow too close to its maximum capacity, then using (b) and (a), the UE Handover Controller 404a can identify the most loaded slice as well as the UE which has the highest relative cost and trigger a handover event for this UE. Therefore, with the above three metrics, the UE Handover Controller 404a can ensure that the UE-to- cell mapping remains close to the ideal UE-to-cell mapping. [00070] The Resource Controller 406 of fig.4 is the entity that controls the amount of dedicated resources to the different network slices in the different cells. Using the mathematical notation from above, the goal of the Resource Controller 406 is to control ^^ ^^^(^) for the different network slices in the different cells. To do this, the Resource Controller 406 takes as input the optimization criterion from the Intent Controller 402. Along with this, the Resource Controller 406 also obtains information from the Observer 408 about long-term trends of the UEs´ resource
needs and optionally also the costs for the cells to deliver those, cost in terms of amount of schedulable radio resources needed for each cell to deliver radio network resources of the plurality of network slices to the plurality of UEs. [00071] In the following, an embodiment of a resource allocation algorithm used by the Resource Controller 406 is described. The resource allocation algorithm can be seen essentially as an offline vector scheduling problem in d dimensions. [00072] Considering the mathematical notation provided above, the input that the Resource Controller 406 obtains from the Observer 408 ^^̂ . The
former (i.e., ^^̂(^)) is defined as the long-term trend for the i-th “typical” UEs resource needs on the k-th slice. The latter ) is defined as the long-term
trend for the j-th cell to provide the i-th UE with the resources it needs. Here it is worth commenting that a “typical UE” with a “typical traffic mix” is a statistical notion and can be created and interpreted in multiple ways. However, one embodiment of such a notion would be to let the set ^ consist of UEs that are the most probable ones to co-exist in the network at the busiest 10% of the hours. ^^̂(^) would then capture the long-term trend for the set ^. Further, the notion of long-term trend is also a statistical notion and can be embodied in many different ways. One such embodiment would be by using a sliding-window approach or a K- means approach combined with a low-pass filter of the resource needs of the different UEs. This is discussed at greater extent in the section describing the Observer 408. [00073] As stated above, finding an optimal resource allocation to the different network slices on the different cells can be seen as an offline vector scheduling problem in ^ dimensions over ^ bins, where ^ is the number of Cells in the network. This resource allocation is given by finding the best
resource mapping → such a mapping has been found, the optimal for the ^-th slice on the ^-th cell can be derived by ^^ ^^^( ^ ) =
∑ {^:^(^)^^} ^^,^(^) . Once this optimal resource allocation has been found, it will be realized by communicating the desired resource allocation to the wireless communication system, i.e., the corresponding base stations in the corresponding
cells, over the interface f in fig.4. In order to find the optimal resource mapping, the optimization criterion needs to be stated first. It is assumed that this will be provided by the Intent Controller 402 over the interface b. One example of such an optimization criterion could be to minimize the maximum resource cost of a slice in a cell in terms of amount of schedulable radio resources needed to deliver the need of radio network resources. The advantage of such a criterion would be to spread the resource load of the UEs evenly across the different cells. This objective could be specified as: minimize
where || ⋅ || ^ is the infinity-norm, given by {^∈ m^,^a∈x^} ^^(^). Another possible optimization objective would be to instead minimize the relative cost of providing resources: minimize ∑{^∈^} ∑{^:^(^)^^} ∑ ^ ^^^ ^^̂,^(^)⁄ ^^̂(^) . Finally, there can also be a weighted objective criterion, to make it more costly for certain cells to provide certain resources to certain slices. As an example, it could be expensive for a lower-band cell, e.g., a 700Mhz cell, to provide resources to an economy slice. On the other hand, it might be cheap to do so for a higher-band cell, e.g., a 3.5Ghz cell. Moreover, for the lower-band cell it might be cheaper to provide resources to the premium slice than to the economy slice. This way, it is possible for the Intent Controller 402 to control the high-level behavior of the network, e.g., which slice-traffic should be flowing on which type of cells. An example of such a criterion could be: minimize
) which is simply a weighted version of the relative cost of providing resources. Finding the solutions to these optimization problems can be done through standard methods such as a linear program or a genetic algorithm. [00074] While the previous section illustrated how the resource allocation, or resource re-allocation, is done, the question remains when to do the resource allocation. This action can be triggered in many different ways. The simplest option is to trigger it periodically. In other words, to trigger a resource re-allocation once
every x hours, days, or weeks. A more advanced option would be to instead have an event-based trigger. This could be embodied by checking if current resource allocation has deviated too far, i.e., above a certain threshold, from what would be an optimal allocation given the current ^^̂(^) and (^). Naturally, a final option is to also involve a Machine Learning (ML)- or
(AI)-based solution to identify when the current resource allocation has deviated to far from the optimal resource allocation. [00075] The final component of fig.4, the Observer 408, is the entity that obtains, for example pulls, needed information from the radio part of the wireless communication network 400, analyzes the information, and compiles the information into a format needed by the Traffic Controller 404 and the Resource Controller 406. Using the mathematical notation above, the Observer 408 would send information to the Traffic Controller 404 about ^^ (^), ^^,^ (^),
. To the Resource Controller 406 it would send information about ^^̂(^) and ^^̂,^(^) where the “^” indicates estimations of the long-term trend or average of the “typical” UEs in the network, their “typical” resource mix, as well as the “typical” cost of delivering resources to those UEs. The Observer 408 communicates with the Traffic Controller 404 over the interface c and with the Resource Controller 4067 over the interface d. [00076] The observation of the UE resource need is relatively straight forward, however requires some care. With the notation above, this is the observation of ^^(^), which means the ^-th UEs need for radio network resources of the ^-th slice. Note again that these resources can be denoted in any entity, but in this proposed solution the focus will be on bandwidth and/or transmission rate, but it could just as well be denoted in latency, or contention-delay, or CPU-cycles, or GPU-cycles, or FPGA-configuration and cycles. [00077] To identify and observe this metric the Observer 408 will interface with the radio access network and the core network of the wireless communication network in order to establish measurements of the data rate being sent from/to the UE. It should be noted that this example does not discuss any use of carrier
aggregation (CA). However, should CA be a factor, it is easy to extend the proposed mechanisms to such a scenario as well. In a case CA is used by a UE, the UE will communicate with the network over multiple cells simultaneously. In such a case, the mapping ^: ^ → ^ will be of type one-to-many, i.e., ^(^) = {^^, ^^, ^^, … } should the ^-th UE send traffic over cells ^^, ^^, and ^^. The resource need of the ^-th UE can then be observed and expressed as: ^^ = ,
the resource on the ^-th cell.
be on the resource needs for the downlink, but the same principles, mechanisms, and methodology also applies to the uplink. [00078] The observation of the cost for a cell ^ to provide a certain amount of radio network resources ^^ (^) to a UE, i.e., to observe ^^,^(^), is a bit more intricate. The main reason for this is because, at least when it comes to catering to the bandwidth needs, the signal quality of the radio link between the UE and the base station matters. Therefore, this would need to be estimated according to a function similar to
, where CSI^,^ is the Channel State Information (CSI)-report that contains information about the signal quality, Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), etc. Should the radio network resource need ^^ (^) instead be related to a latency requirement, the function to estimate the resource cost would also have to consider the other UEs sending traffic on the cell. In other words, it would have to consider queue-sizes of the traffic flows of the other UEs. [00079] The observation of the resource load of a cell ^ for a slice ^ is performed by aggregating the observation of the resource cost for all the UEs sending traffic over the ^-th slice:
= (^)^^} . These observations when the considered resources are bandwidth, would correspond to aggregating the symbols or the physical resource blocks that the cell provides to the ^-th slice. Should the considered resource be of a different
type, for instance latency, then the cost could be measured through the volume of resources that the cell has to increase the priority for. Should the considered resources be for CPU-cycles, or GPU-cycles then it could simply be an aggregate of the number of cycles provided to the different slices. [00080] When it comes to estimating the long-term averages of the resource need and the cost to deliver those resource ^^̂ and ^^̂,^(^),