hosted by an end user device

The disclosure relates to computer networks, particularly to wireless networks, particularly to mobile networks and, more specifically, to latency critical applica tions provided within those networks.

A network services exchange provider may employ a communication facility, such as a data center, in which multiple customers of the provider locate net work, server, and storage gear and interconnect to a variety of telecommunica tions and other network service provider(s). Data centers may be shared by the multiple tenants having their networking equipment located within the data cen ters.

With Information Technology (IT) and communications facilities in safe, secure hands, telecommunications, Internet, application service providers, cloud ser vice providers, content providers, and other providers, as well as enterprises, enjoy less latency and the freedom to focus on their core business. Additionally, customers may reduce their traffic back-haul costs and free up their internal networks for other uses.

However, there are some scenarios being realized by server based latency crit ical applications which require data processing near a respective end user.

Autonomous vehicles, for example, will generate a lot of data. Much of this data will be unstructured and will need to be run through powerful analytics programs to produce actionable data with any value to businesses. Edge computing archi tectures will help to prioritize what data needs to remain on the edge to be pro- cessed by the vehicle’s onboard computing power or by any computing device nearby the vehicle and what data should be relayed back to data centers for analysis.

In an autonomous vehicle, even a few milliseconds of delay can result in an ac cident and catastrophic loss of life. The stakes are simply too high to allow the vehicles’ networks to be plagued by lag. Self-driving cars need to react immedi ately to changing road conditions; they can’t simply come to a stop while waiting for instructions or recommendations from a distant cloud server analyzing data.

There are some scenarios where it is useful to locate applications directly on terminals, i.e. user equipment (UE) and, thus, to acquire a reduced latency with respect to roundtrip-time (RTT). Such applications are also called client-based application (client = end user device) in contrast to server-based applications.

However, it is still quite difficult or even impossible to provide latency critical applications via a wireless interface, i.e. a radio interface as currently used schedulers operate according to a best effort and fair schedule approach when allocating the available resources on the wireless interface which leads to an inacceptable variance in latency and jitter for any latency critical application.

As a solution to the above mentioned object, the present disclosure provides a method and a system with the features of the independent claims. Further em bodiments are provided by the dependent claims and the description.

A method is provided, the method comprising:

a) provisioning, at an end user device which is located in a cell of a wireless network, at least one latency critical application and a service layer radio application for communication with a scheduler associated with a base station of the wireless network, the base station serving the cell;

b) transferring transmission specific data in real time from the at least one latency critical application to the scheduler associated with the base sta tion and/or vice versa from the scheduler to the at least one latency criti- cal application via the service layer radio application (SLRA) which is im plemented on both, the end user device and the scheduler, and c) continuously optimizing allocation of resources in the cell by taking into account current status and operation requirements of the at least one la tency critical application and/or continuously optimizing current use of the resources in the cell by using those transmission specific data for adapt ing the at least one latency critical application in real time to current con ditions on the cell.

SLRA can also be implemented on the base station and data received via SLRA are internally forwarded to the scheduler.

Transmission is to be understood within the scope of the present disclosure in a broad sense, particularly as data transmission, i.e. as a transfer of data (a digi tal bitstream or a digitized analog signal) over a communication channel, such channels can be copper wires, optical fibers, wireless communication channels, storage media and computer buses. In the present disclosure, particularly wire less communication channels are considered.

Data can be represented as electromagnetic signals, such as radiowaves, mi crowaves, an electrical voltage or infrared signal. Data can be represented as packets, frames, bits, information considering all different communication lay ers.

Transmission specific data comprise, in the case that the transmission specific data are transferred from the at least one latency critical application to the scheduler, application specific data, such as, for example, data about a current operations status and operation requirements of the application (in the following also called context data), and in the case that the transmission specific data are transferred from the scheduler to the at least one latency critical application, data about current conditions on the cell, such as, for example, channel condi tions, available resources, throughput, packet delay, queue length, etc. "Continuously" means that changes concerning data transmission are observed in real time and whenever changes occur which have an impact on data trans mission quality the allocation of resources and/or the current use of the re sources is optimized regarding those changes.

According to an embodiment of the claimed method, step c) comprises:

d) receiving by the at least one latency critical application via the SLRA cell specific data from the scheduler of the base station and/or sending by the at least one latency critical application via the SLRA application spe cific data to the scheduler of the base station; and

e) processing, by the scheduler associated with the base station, the appli cation specific data received from the at least one latency critical applica tion and allocating, by the scheduler, available transmission resources to the at least one latency critical application taking into account the appli cation specific data and the current operation requirements of the at least one latency critical application, and/or

f) processing the cell specific data received from the scheduler of the base station and adapting the at least one latency critical application to the current conditions on the cell taking into account the application specific data.

The at least one application is specified here as a latency critical application. However, the claimed method and system and computer readable storage me dium can also be applied to near or no latency critical applications.

The scheduler is associated with the base station means that the scheduler can be co-located with the base station or it can be part of the base station or it can be located remotely from the base station for which it operates.

The decisive end-to-end latency is predominantly defined by the latencies within the radio path. Generally, up to now schedulers are optimized to data throughput for all devices which are connected to the base stations housing the respective schedulers in order to acquire a high spectrum efficiency (Bit/s/Hz), such as algorithm propor tional fair, thus, exhausting the limited and valuable resources at the best. How ever, generally, no differentiation is made between different types of use cases, such as Youtube videos which are buffered, and latency critical applications. Both, the Youtube videos and the latency critical applications have same priori ty.

This results in that the latency for an application can extremely vary in depend ence on the number of connected users, the signal quality etc. Such variance can be recognized via different mean latencies subject to location and time, and also via a respective jitter.

Nowadays, latency critical applications, such as driver assistance systems with the exchange of surroundings models, remote rendering and streaming of AR (Artificial Reality) and VR (Virtual Reality) applications, VR360 applications, of floading of computer vision algorithms for drones, vehicles, robots etc. underlie the normal scheduling. The occurring variance with respect to the mean latency and the jitter impede or even hinder the use of latency critical applications via the radio. Further, a static prioritization or static allocation of resources would reduce the spectrum efficiency and, thus, increase the costs for such latency critical applications.

According to an embodiment of the claimed method, a scheduler associated with the base station prioritizes and/or allocates the available resources to the at least one latency critical application in accordance with the present/current requirements of the at least one latency critical application.

Further, the at least one latency critical application determines in real time on the basis of a present operations status of the at least one latency critical appli cation at least some of the following context data: separated for uplink and downlink: a currently required latency, a maximum required latency, a currently required minimum throughput, maximum error rate, priority, desired duration or number of desired transmissions of this configuration. These determined con text data are then sent by the at least one latency critical application in real time via the SLRA to the scheduler associated with the base station.

According to a further embodiment, the scheduler associated with the base sta tion dynamically and in real time takes into account the determined context data received from the at least one latency critical application hosted by the end user device when allocating, prioritizing and reserving the available resources (e.g. resource blocks) for the at least one latency critical application in order to achieve an efficient scheduling (spectrum efficiency with Bits/s/Hz) compared to a static reservation/constant prioritization/reservation of resources for the at least one latency critical application, and in parallel to fulfil the requirements of the at least one latency critical application.

In order to guarantee latencies or maximum latencies, resources have to be prioritized or blocked. In the case of a radio network those resources are re source blocks or resource elements. According to the present invention, the at least one latency critical application hosted by the end user device provides the base station, i.e. the scheduler in real time with context sensitive parameters about the at least one latency critical application, such as a current mean laten cy, the application needs at a present point in time, a maximum latency, a data rate the application needs at the present point in time and further parameters which are useful for the scheduler as reference point for the current state of the at least one latency critical application. These parameters are provided besides other conventional used parameters such as channel condition, historical throughput, packet delay, queue length, etc.

First, the end user device which can be a smartphone or any other suitable ter minal, provisions the at least one latency critical application. Due to the proximi ty of the end user device to the scheduler, information can be transmitted be tween the at least one latency critical application hosted by the end user device and the scheduler in real time. Thereby, the at least one latency critical applica- tion at the end user device is logically connected with the radio scheduler asso ciated with the base station via the service layer radio application (SLRA). This enables the scheduler to receive context sensitive information/parameters about and from the at least one latency critical application via this service layer. The at least one latency critical application determines in real time dependent on its current status the mean latency needed at this current point in time, the maxi mum latency, the data rate needed at this current point in time and further sta tus dependent parameters. The scheduler considers these context sensitive parameters when allocating the available resources: for example, the scheduler provides a higher maximum latency at a point in time t, a lower maximum laten cy at a point in time t+1 , a low maximum latency with a lower data rate at a point in time t+2, a higher maximum latency with a higher data rate at a point in time t+N, etc.

Thus, only resources are prioritized or blocked at a specific point in time for the at least one latency critical application which are really required by the at least one latency critical application at this specific point in time. Thereby, an opti mized usage of the spectrum of the radio frequencies is reached while the re quirements of the at least one latency critical application are fulfilled.

Thus, for example, a vehicle driving with a low velocity (or in stillstand) can transmit periodically data such as a surroundings model with a higher maximum latency without compromising the driving safety. Conversely, the maximum la tency should be reduced when the vehicle is driving with a higher velocity in order to adapt the reaction times within the global system accordingly and to guarantee the driving security.

Particularly, applications with a high data throughput are relevant for the adap tation of the data rate, such as video based applications: a remote rendering of an AR application requires the streaming of video data. Thereby, the data rate can be reduced or has to be increased as a function of the number of rendered objects, the size of the image area, the required resolution for the depth of detail etc. Due to the provision of the context sensitive parameters/requirements of the at least one latency critical application in real time with respect to maximum laten cy and data rate, the scheduler can dynamically optimize the distribution and allocation of the radio resources.

Thus, by the knowledge of the current requirements of the at least one latency critical application with regard to maximum latency and data rate (throughput), the scheduler gets more flexible when allocating and reserving the limited re sources of the radio spectrum. Thus, the spectrum can be used/exploited more efficiently as by algorithms without knowledge of a current status of the at least one latency critical application. Such conventional algorithms do not at all take latency critical applications into account and new approaches as indicated be low consider latency requirements only statically and with a lowest latency or with a mean latency which leads to the above mentioned disadvantages such as unsatisfactory and insufficient exploitation of the radio spectrum.

There are new approaches in the 5G standardization to provide a class of appli cations with a defined quality of service (QoS), for example by means of net work slicing and/or URLLC (Ultra Reliable Low Latency Communication). Net work slicing allows a network operator to provide dedicated virtual networks with functionality specific to a service over a common network infrastructure. Net work slicing is a form of virtual network architecture and allows multiple virtual networks to be created on top of a common shared physical infrastructure. The virtual networks are then customised to meet specific needs of applications.

However, all these approaches are based on the assumption of stat ic/unchanging latency and data throughput for a specific application. Thus, the allocation of resources to any application is also static, resulting in an inefficient utilisation of resources, particularly in case of critical cell conditions, such as a full cell, bad signal conditions, end user device which is to be provided with a latency critical application being situated at the edge of the cell, etc. Due to the assumption of static latency and throughput for a specific application, non- latency critical applications are supplied, particularly in case of critical cell condi tions, worse with resources according to those approaches compared to the method described herein.

According to a further embodiment of the claimed method, the scheduler asso ciated with the base station determines in real time available resources for the at least one latency critical application as a function of a current status of the cell served by the base station, taking into account all connected devices with their applications, latency critical and/or non-latency critical applications, and their requirements and the overall available resources in the cell, derives there from a currently attainable individual data throughput and/or a currently attaina ble individual latency and/or a current individual packet loss for the at least one latency critical application, and sends the currently attainable individual data throughput and/or the currently attainable individual latency and/or the current individual packet loss to the at least one latency critical application via the SLRA and the at least one latency critical application adapts in real time accordingly.

The application adapts to the present status of the base station by adopting it self an appropriate operations status in real time. In that case it is possible that the transmission specific data are only transferred/sent in one direction, namely from the scheduler to the at least one latency critical application.

So far, an application only gets implicitly to know about a change of scheduling. Due to a new allocation of resources at the wireless/radio interface the through put for an application, for example, can be reduced instantaneously without providing the application with a possibility to adapt itself accordingly. Instead, packets can be queued/buffered or even discarded. This leads to undesired effects within the application and may require repetitions. When using TCP (Transmission Control Protocol) the data rate which is sent, is in the medium term adapted to the new transport conditions. All this leads to jitter and to unde sired effects or errors of the respective application. Instead of an indirect tuning of the application the provided method enables the at least one latency critical application to adapt in real time to a current status of a radio cell which is served by the base station. The at least one latency critical application is running near the base station at the end user device. Thus, the scheduler within the base station can transmit in real time status information about the radio cell to the at least one latency critical application hosted by the end user device. Such status information is transmitted via the service layer radio application (SLRA) which logically connects the base station/the scheduler with the end user device and, thereby, with the at least one latency critical ap plication. The scheduler determines in real time the resources which are availa ble for the at least one latency critical application in terms of resource blocks or resource elements. Therefrom, the possible latency and the data rate on the radio for the at least one latency critical application are derivable. Further pa rameter, such as packet loss etc. can be determined. The status data“new la tency”,“new data rate”,“packet loss” etc. are transmitted via SLRA in real time to the end user device and transferred to the at least one latency critical appli cation. The at least one latency critical application can adapt now to the new status information and can capture a new appropriate operations status.

Thus, an autonomous vehicle can reduce its velocity when the latency guaran teed by the base station, i.e. the scheduler is currently increased. Conversely, the vehicle can increase its velocity as soon as the latency is reduced.

With remote rendering a higher compression rate with more image/compression artefacts can be used in order to adapt to a new data rate.

With computer games with a central game server, mechanisms such as interpo lation and extrapolation can be reactivated with higher latencies with which higher latencies can be equalized, but which can also lead to errors.

Due to the ideal co-working of scheduler, end user device and the at least one latency critical application in real time the best possible operations status of the at least one latency critical application which can be currently reached, can be realized. A continuous over-tuning of the application which can appear with the above mentioned indirect adaption mechanisms (according to the state of the art) lead ing to an uncontrollable behavior of the application can be avoided.

According to a further embodiment of the claimed method, the method further comprises the step of:

- determining, for the at least one latency critical application, different possible operating points with accordingly different quality stages;

- controlledly taking up in real time one of the possible operating points in accordance with a current status of the cell served by the base sta tion concerning required resources for all connected devices.

In a further embodiment of the claimed method each of the possible operating points of the at least one latency critical application is assigned a priority and defined by at least one of the following parameters: minimum throughput down link, minimum throughput uplink, maximum latency downlink, maximum latency uplink, maximum error rate, desired duration or number of desired transmis sions of this configuration.

It is possible that the possible operating points for the at least one latency criti cal application which are listed in a table of operating points, are sent in real time from the at least one latency critical application hosted by the end user de vice to the scheduler of the base station, the scheduler of the base station ag gregates all possible operating points of all applications which are currently run ning via the base station, allocates the available resources accordingly and sends a feedback message to the at least one latency critical application host ed by the end user device about the finally selected operating point for the at least one latency critical application.

After selection of one of the possible operating points, the scheduler provides the at least one latency critical application temporarily with specific resources, such as maximum latency, minimum throughput with a well-defined Quality of Service (QoS), i.e. a well-defined guarantee. That means that the at least one latency critical application may relay for a specific period of time that the proper ties of the data transmission correspond to those of the selected operating point. Such properties are guaranteed until a new operating point is selected, triggered by the scheduler and/or the at least one latency critical application according to the techniques/mechanisms described herein.

In order to guarantee specific properties for data transmission, the scheduler takes appropriate measures such as prioritization, reservation of resources, etc.

The scheduler and the at least one latency critical application are again logically connected via the service layer radio application.

According to the above mentioned embodiment the at least one latency critical application determines as a function of its present status different possible op erating points, each representing an allowable solution for the at least one la tency critical application. However, the different operating points are assigned different quality stages for the at least one latency critical application, for exam ple via a“graceful degradation”.

The different operating points are defined by at least some of the following re quirement parameters:

- mean latency

- maximum latency

- data rate/throughput

- maximum error rate

- desired length of the operating point (in milliseconds)

- desired number of transmissions

- priority

- further possible parameters which describe the operation status of the application The possible operating points for the at least one application are defined be forehand by a developer. The aim of such determination of operating points is to perform a substitution of uncontrollable errors and states due to changed laten cy and throughput by a controlled reduced user experience. Is an optimal oper ating point for the at least one latency critical application due to a full cell or oth er specific cell conditions not realizable, the controlled adaption of a further al lowed operating point is possible.

It is to be stated that the decision process about the operating points which are to be taken up in specific scenarios can be subdivided into three stages:

(a) the developer of the at least one latency critical application defines all possible operating points for the at least one latency critical application when developing the at least one latency critical application,

(b) the at least one latency critical application decides itself which operating points are eligible in a specific scenario, i.e. in a specific current status of the at least one latency critical application. Thus, for each scenario there is a group of operating points from which one operating point can be se lected,

(c) the scheduler finally decides in real time and dynamically, which operat ing point of the group of operating points is selected dependent on the current status of the cell.

It is further possible that the at least one latency critical application is adapted to the current status of the cell by adapting/changing its internal parameters such as compression rate, frames per second, etc. which leads to a controlled reduc tion of quality of the at least one latency critical application, such as compres sions artefacts.

The following table gives an example how operating points can be defined for the at least one latency critical application:

The possible operating points are transferred together with their respective re quirement parameters via the SLRA in real time to the scheduler. The scheduler aggregates all requirements of all latency critical applications. Then, the scheduler determines the allocation of resources for all user equip ment within the cell. Thereby, the scheduler takes into account:

- The no latency critical applications within the cell,

- The latency critical application within the cell,

- Conventional input parameters such as channel condition, historical throughput, packet delay, queue length for each user equipment,

- Requirement parameters originating from the defined operating points of the latency critical applications.

First, an allocation of resources to all connected devices (latency critical and no latency critical) with requirement parameters indicated with priority 1 of the op erating points are examined for realization. Only in the case that a resource shortfall occurs, the alternatives of the further priorities of the operating points are considered until a successful resource allocation is found. Thereby, it is possible to put all latency critical applications equally in their next lower priority stage. Further, any kind of mixed method is possible. An optimization between no latency critical and latency critical applications is possible. When a success ful resource allocation is identified the scheduler applies this resource allocation and provides a respective application in real time with a report about the finally selected operating point.

According to a further embodiment, the scheduler associated with the base sta tion and the at least one latency critical application hosted by the end user de vice exchange iteratively information for negotiating a best adjustment of the at least one application to present data transmission conditions.

There are two possible alternatives for starting such iterative process.

According to a first alternative, the scheduler associated with the base station sends, as starting point of the negotiation, an initial configuration comprising at least a maximum latency and a data rate to the at least one latency critical ap plication hosted by the end user device. Thereby, it is possible that the sched- uler sends further parameters. Such initial configuration is to be understood as a suggestion on the basis of which further negotiation is to be pursued.

According to a second alternative, the at least one latency critical application hosted by the end user device sends, as starting point of the negotiation, an initial configuration comprising at least an operating point or a list of possible operating points for the at least one application as starting point of the negotia tion to the scheduler associated with the base station. Such initial configuration is to be understood as a suggestion on the basis of which further negotiation is to be pursued.

When receiving the initial configuration the respective party, i.e. the scheduler associated with the base station or the at least one latency critical application hosted by the end user device, examines the initial configuration and accepts the initial configuration or transmits a proposal for changes to the respective other party, i.e. to the at least one latency critical application hosted by the end user device or to the base station, particularly to the scheduler associated with the base station. Particularly, the scheduler is often forced to perform iterations as the scheduler which has to control and administrate a plurality of latency crit ical applications, has to determine the correct operation status, i.e. all of the plurality of latency critical applications have to be supported, i.e. supplied with resources at the best. As the allocation of resources has to consider all latency critical applications simultaneously, usually more cycles between scheduler and the at least one latency critical application provisioned by the end user device are necessary. This scenario requires a higher reliability of the real time con nection (realized by the SLRA) between the scheduler and the at least one la tency critical application hosted by the end user device because more than one communication processes can follow each other. This has an influence on the time the scheduler needs to determine an optimal allocation of the resources.

It is possible that real time events occur which lead to changes in the cell. Such changes can cause changing operation conditions for the at least one latency critical application. In case of such changes in the cell the scheduler associated with the base station covering the cell sends in real time an information about such changes and/or such changing operation conditions via SLRA to the at least one latency critical application hosted by the end user device. The at least one latency critical application decides about a new subset of operating points which can be adopted by the latency critical application in order to adapt to the changing operation conditions in the cell at best. The at least one latency critical application sends the scheduler a message via SLRA about the new subset of operating points and the scheduler selects one operating point of the new sub set of operating points.

After selection of one of the possible operating points, the scheduler provides the at least one latency critical application temporarily with specific resources, such as maximum latency, minimum throughput with a well-defined Quality of Service (QoS), i.e. a well-defined guarantee. That means that the at least one latency critical application may relay for a specific period of time that the proper ties of the data transmission correspond to those of the selected operating point. Such properties are guaranteed until a new operating point is selected, triggered by the scheduler and/or the at least one latency critical application according to the techniques/mechanisms described herein.

In order to guarantee specific properties for data transmission, the scheduler takes appropriate measures such as prioritization, reservation of resources, etc.

Such exchange of transmission specific data between the scheduler and the at least one latency critical application happens in real time at each change in the cell and/or the at least one latency critical application.

Thus, a real time adaption of the parties concerned, such as scheduler/cell and/or application, is reached.

The above described communication between the scheduler and the at least one latency critical application which is directly realized between the scheduler associated with the base station and the at least one latency critical application hosted by/implemented on an end user device/a user equipment (UE), is partic ularly important within an uplink based communication.

The present disclosure also refers to a system comprising at least an end user device which is located in a cell of a wireless network, the cell being served by a base station of the wireless network,

wherein the end user device is configured to:

a) provision at least one latency critical application which is to be provided to at least one end user of the end user device in the cell via the base station serving the cell,

b) provision a service layer radio application for communication with a scheduler associated with the base station which serves the cell, c) enable a transfer of transmission specific data in real time from the at least one latency critical application to the scheduler associated with the base station and/or vice versa from the scheduler to the at least one la tency critical application via the service layer radio application (SLRA) which is implemented on both, the end user device and the scheduler.

The claimed system is configured to execute a method according to any one of the above described embodiments.

The wireless network is generally deployed and managed by a network provid er.

The present disclosure further refers to a computer readable storage medium comprising instructions that when executed cause one or more processors of a mobile network to:

A) provision, at an end user device which is located in a cell of a wireless network, at least one latency critical application and a service layer radio application for communication with a scheduler associated with a base station of the wireless network, the base station serving the cell, B) transfer transmission specific data in real time from the at least one la tency critical application to the scheduler associated with the base station and/or vice versa from the scheduler to the at least one latency critical application via the service layer radio application (SLRA) which is imple mented on both, the end user device and the scheduler, and

C) continuously optimize allocation of resources in the cell by taking into ac count current status and operation requirements of the at least one la tency critical application and/or continuously optimize current use of the resources in the cell by using those transmission specific data for adapt ing the at least one latency critical application in real time to current con ditions on the cell.

The details of one or more examples of the techniques are set forth in the ac companying drawings and the description below. Other features, objects, and advantages of the techniques will be apparent from the description and draw ings, and from the claims.

Brief description of drawings

Figure 1 is a schematic diagram illustrating an example system for provisioning of services/applications by a mobile network operator in accordance with tech niques described herein.

Figure 2 is a detailed view of figure 1 illustrating schematically a data transfer between user equipment, base station and edge computing system in accord ance with techniques described herein.

Figure 3 shows schematically an allocation of resources to a plurality of latency critical applications in accordance with techniques described herein. Detailed description of drawings

Figure 1 is a schematic diagram illustrating an example system 100 incorporat ing an embodiment of a claimed system. The system 100 comprises a network 1 10 of a mobile network operator (MNO), an access layer 120 and end user devices 130 which can access the mobile network 1 10 via the access layer 120. Further the devices 130 can access the Internet 140 via the access layer 120 and the mobile network 1 10 and, thus, benefit from all services provided by or via the Internet 140.

The mobile network 1 10 comprises a plurality of access nodes 1 1 1 , such as a MSAN (Multi-Service Access Node) and/or a cell tower (an antenna support with mobile antennas), a plurality of edge computing systems 1 12 and a back bone 1 13 interconnecting the mobile network 1 10 with the Internet 140.

A mobile network operator (MNO), also known as a wireless service provider is a provider of wireless communications services that owns or controls all the el ements necessary to sell and deliver services to an end user including radio spectrum allocation, wireless network infrastructure, back haul infrastructure, provisioning computer systems, etc.

In general, edge devices are normally routers that provide authenticated access (most commonly PPPoA and PPPoE) to faster, more efficient backbone and core networks. The edge computing systems 1 12 provided here are made smart, so that the edge computing systems 1 12 are not only routers, but can include Quality of Service (QoS) and multi-service functions to manage different types of traffic and provision applications, particularly latency critical application. Each of the plurality of edge computing systems 1 12 is located nearby at least one base station 1 1 1 and, thus, nearby a scheduler co-located or at least asso ciated with the at least one respective base station 1 1 1 . Due to the physical proximity of each of the edge computing systems 1 12 to at least one respective base station 1 1 1 , information between the edge computing systems 1 12 and the respective neighbouring base stations 1 1 1 can be transferred in real time. This enables the scheduler associated with a respective base station 1 1 1 to allocate in real time available resources to applications which are provisioned by the respective neighbouring edge computing systems 1 12.

Considering a latency critical application which is to be provided to an end user device 130, the latency critical application can be directly hosted by the respec tive end user device 130 or it can be provisioned by a selected edge computing system 1 12 which is nearby the base station 1 1 1 serving the end user device 130. The scheduler associated with the base station 1 1 1 is provided in real time via the selected edge computing system 1 12 and/or the end user device 130 with application specific data/parameters when allocating resources to the la tency critical application. Such application specific parameters are, for example, context sensitive parameters of the latency critical application, such as mean latency, maximum latency and data rate which are currently needed by the la tency critical application. The scheduler is provided with further input parame ters, such as“channel condition”,“historical throughput”,“packet delay”,“queue length”, i.e. with parameters reflecting a current status on/in the cell. As those parameters are provided in real time, the scheduler can consider dynamically requirements of the application which change with time. Thus, the scheduler only blocks resources which are necessary for fulfilling the current requirements of the application in real time. An optimal usage of the spectrum is reached while the latency requirements of the latency critical application are fulfilled at any time.

Figure 2 shows a detailed view of figure 1 illustrating schematically a data trans fer between a user equipment 230 and a base station 21 1 , and further a data transfer between the base station 21 1 and an edge computing system 212 which is located nearby the base station 21 1 in accordance with techniques de scribed herein. The user equipment 230 is currently served by the base station 21 1 because the user equipment 230 is currently positioned within a coverage area of a cell which is served, i.e. defined by the base station 21 1 as indicated by double arrow 260. Thus, data are transferred between the user equipment 230 and the base station 21 1 as indicated by double arrow 231. Applications which are desired by the user equipment 230 have to be supplied with re sources (resource blocks/resource elements) via the base station 21 1 . A scheduler associated with the base station 21 1 has to allocate available re sources to all applications and services which are running via the base station 21 1. In the case that the user equipment 230 is to be provided, for example on demand, with a latency critical application, the latency critical application can be directly implemented on the end user device 230 as client-based application indicated by reference number 234 or it can be hosted as server-based applica tion 214 by a selected edge computing system 212 that is located most nearby the base station 21 1 , as indicated by double arrow 250, for that the selected edge computing system 212 provisions the latency critical application 214. In the following, the latency critical application 234 which is implemented on the end user device 230 is considered. The end user device 230 comprises further a service layer radio application (SLRA) 235 and a control plane 236 and/or a user plane.

The end user device 230 is logically connected, as indicated by double arrow 237, with the scheduler of the base station 21 1 via the service layer radio appli cation 235. Via the service layer radio application 235, the scheduler receives in real time context sensitive parameters of the latency critical application 234, such as currently needed mean latency, currently needed maximum latency and currently needed data rate. Thus, the scheduler of the base station 21 1 can consider those context sensitive data when dynamically allocating resources to the latency critical application 234. Thus, the scheduler can allocate a higher maximum latency at a point in time t, a smaller maximum latency at a point in time t+1 , a smaller maximum latency with a smaller data rate at a point in time t+2, a higher maximum latency with a higher data rate at a point in time t+N. Thus, only resources are blocked in real time for the latency critical application 234 which are really currently needed by the latency critical application 234. Therefore, at any time, an optimal usage of the spectrum is reached while la tency requirements are simultaneously fulfilled. The base station 21 1 and the selected edge computing system 212 form part of the mobile network 1 10 of the mobile network operator, as indicated by circle 1 10. Figure 3 shows schematically an allocation of resources to a plurality of latency critical applications A_1 , A_2, A N in accordance with techniques described herein.

Figure 3 shows a scheduler 318 as part of a base station. The scheduler 318 is configured to allocate dynamically available resources to all different applica tions A_1 , A_2, ..., A N running via the base station. Each of those different applications A_1 , A_2, ..., A N is provided for a user UE1 , UE2, ..., UEN via a respective end user device 330_1 , 330_2, ...330_N, particularly each for a re spective different user. Figure 3 shows a plurality of N different end user devic es 330_1 , 330_2,... 330_N, each end user device 330_1 , 330_2,... 330_N pro visions/hosts an application A_1 , A_2, ..., A N. According to a possible embod iment of the claimed method, it is envisaged to determine for each application A_1 , A_2, ..., A N different possible operating points B1 , B2, ...., BM with ac cordingly different quality stages. Further, it is possible, that the possible operat ing points B1 , B2, ..., BM differ from application to application. Each operating point B1 , B2, ..., BM is defined by a number of parameters P1 , P2, ..., PK. For the sake of convenience, the operating points are designated for all applications identically by B1 , B2, ..., BM. The same applies to the parameters which are designated for all applications identically by P1 , P2 , ..., PK. It is now possible, that each application is taking up in real time in a controlled manner one of the possible operating points in accordance with a present status of the base sta tion, i.e. the cell served by the base station. Each of the numbers K, M, N is a positive integer.

Each of the possible operating points is assigned a priority and defined by at least one of the following parameters P1 , P2, ..., PK: throughput downlink, throughput uplink, maximum latency downlink, maximum latency uplink.

The possible operating points B1 , B2,..., BM for the different applications A_1 , A_2, ..., A N are sent in real time from the different end user devices 330_1 , 330_2, ..., 330_N, particularly from the different applications hosted by the dif- ferent end user devices 330_1 , 330_2, 330_N as input parameters, as indi cated by arrows 314 1 , 314 2, ..., 314 N to the base station, i.e. to the sched uler 318. The scheduler 318 also receives classical input parameter 316, such as channel condition, historical throughput and packet delay and queue length and other input parameters 315. The scheduler 318 of the base station aggre gates all possible operating points of all applications which are currently running via the base station, allocates the available resources accordingly and sends a feedback message to the respective end user devices 330_1 , 330_2, ..., 330_N, particularly to the respective applications hosted by the respective end user devices about the finally selected operating point for each of the respective applications. Such feedback message is indicated by arrows 331 _1 , 331 _2, ..., 331_N, respectively.

The scheduler 318 and the respective applications A_1 , A_2, ..., A N are again logically connected via the service layer radio application, respectively.

According to the above mentioned embodiment, each application determines as a function of its present status different possible operating points, which repre sent an allowable solution for the application, respectively. However, the differ ent operating points are assigned different quality stages for the application, for example via a“graceful degradation”.

The different operating points are defined by at least some of the following re quirement parameter:

- Mean latency

- Maximum latency

- Data rate/throughput

- Maximum error rate

- Desired length of the operating point (in milliseconds)

- Priority

- Further possible parameters which describe the operation status of the application The possible operating points B1 , B2, BM for each application A_1 , A_2,

A N are defined beforehand by a developer. The aim of such determination of operating points is to perform a substitution of uncontrollable errors and states due to changed latency and throughput by a controlled reduced user experi ence. In the case that an optimal operating point for a respective application, due to a full cell or other specific cell conditions, is not realizable, the controlled adaption of a further allowed operating point is possible.

The following table gives an example how the operating points B1 , B2, BM, exemplary M=7, can be defined for each application A_1 , A_2, A N:

The possible operating points B1 , B2, BM are transferred together with their respective requirement parameters via the SLRA in real time to the scheduler 318.

The scheduler 318 aggregates all requirements of all latency critical applica tions. Then, the scheduler determines the allocation of resources for all user equipment/end user devices 330_1 , 330_2, ..., 330_N within the cell, as indi cated by arrow 317. Thereby, the scheduler 318 takes into account:

- The no latency critical applications within the cell,

- The latency critical application within the cell,

- Conventional input parameters such as channel condition, historical throughput, packet delay, queue length for each user equipment,

- Requirement parameters originating from the defined operating points of the latency critical applications.

First, an allocation of resources to all devices (latency critical and no latency critical) with requirement parameters indicated with priority 1 of the operating points B1 , B2, ..., BM are examined for realization. Only in the case that a re source shortfall occurs, the alternatives of the further priorities of the operating points B1 , B2, ..., BM are considered until a successful resource allocation 317 is found. Thereby, it is possible to put all latency critical applications equally in their next lower priority stage. Further, any kind of mixed method is possible. An optimization between no latency critical and latency critical applications is pos sible. When a successful resource allocation 317 is identified the scheduler 318 applies this resource allocation and provides a respective application in real time with a report about the finally selected operating point. According to a further embodiment, the scheduler and the respective applica tions hosted by a respective end user device exchange iteratively information for negotiating a best adjustment of all applications to present data transmission conditions within the respective cell.

List of reference signs

100 system

1 1 0 wireless network

1 1 1 access node/base station

1 1 2 edge computing system

1 13 backbone

120 access layer

130 end user device/user equipment

140 Internet

21 1 base station

21 2 edge computing system

214 server-based application

230 end user device

231 double arrow

234 client-based application

235 service layer radio application

236 control plane

237 double arrow

250 double arrow

260 double arrow

315 further input parameters/data

316 classical input parameters/data

317 resource allocation

318 scheduler

314 1 , 314 2, ..., 314_N arrows

331 1 , 331 2, ..., 331 _N arrows (feedback message) A_1 , A_2, ..., A_N applications

330 1 , 330 2, ..., 330 N end user devices

B1 , B2, ..., BM operating points

R1 , R2, ..., RK parameters