Abstract: Slices of a 5G network may be configured to implement a trust model by which network customers are provided with assurances that slice properties meet agreed-upon criteria specified by customer policy so that slices can be trusted. Illustrative slice properties may pertain to service types, geographic area of operations, and attributes associated with software, firmware, and hardware used in the infrastructure of nodes in a trusted slice. Particular values of the properties describe a slice configuration that may be measured, digested, and attested to the customer to provide assurances that the configuration conforms with the policy. The 5G slice trust model may be implemented as a two-way model in which a slice provider performs checks to verify slice properties while customers ensure that only authenticated and authorized user equipment (UE) will access a trusted slice.

Abstract: Computing resources are managed in a computing environment comprising a computing service provider and an edge computing network. The edge computing network comprises computing and storage devices configured to extend computing resources of the computing service provider to remote users of the computing service provider. The edge computing network collects capacity and usage data for computing and network resources at the edge computing network. The capacity and usage data is sent to the computing service provider. Based on the capacity and usage data, the computing service provider, using a cost function, determines a distribution of workloads pertaining to a processing pipeline that has been partitioned into the workloads. The workloads can be executed at the computing service provider or the edge computing network.

Abstract: Aspects of the present disclosure relate to allocating RAN resources among RAN slices using a machine learning model. In examples, the machine learning model may determine an optimal RAN resource configuration based on compute power needs. As a result, RAN resource allocation generation and compute power requirements may improve, even in instances with changing or unknown network conditions. In examples, a prediction engine may receive communication parameters and/or requirements associated with service-level agreements (SLAs) for applications executing at least partially at a device in communication with the RAN. The RAN may generate one or more RAN resource configuration for implementation among RAN slices. Upon a change in network conditions or SLA requirements, an optimal RAN configuration may be determined in terms of required compute power.

Abstract: Described are examples for receiving, from one or more second virtual radio access network (vRAN) workloads operating one or more second cells, an indication of a measurement of at least a first signal transmitted by a first vRAN workload operating a first cell, computing, based on measurements of at least the first signal as received from the one or more second vRAN workloads, a boundary of the first cell, and adjusting, based on the boundary of the first cell, a transmit parameter of the first vRAN workload for transmitting signals in the first cell.

Abstract: Described are examples for providing a distributed fault-tolerant state store for a virtualized base station. In an aspect, a first server at a datacenter may perform physical layer processing for at least one virtualized base station. While performing the physical layer processing, the first server may generate inter-slot physical layer state data during a first slot. The inter-slot physical layer state data is to be used in a subsequent slot. The first server may periodically transmit the inter-slot physical layer state data to one or more other servers of the plurality of servers within the datacenter. One of the other servers may take over the physical layer processing for the at least one virtualized base station based on the inter-slot physical layer state data, for example, in response to a fault at the first server or a migration of the at least one virtualized base station.

Abstract: Systems and methods are provided for offloading a task from a central processor in a radio access network (RAN) server to one or more heterogeneous accelerators. For example, a task associated with one or more operational partitions (or a service application) associated with processing data traffic in the RAN is dynamically allocated for offloading from the central processor based on workload status information. One or more accelerators are dynamically allocated for executing the task, where the accelerators may be heterogeneous and may not comprise pre-programming for executing the task. The disclosed technology further enables generating specific application programs for execution on the respective heterogeneous accelerators based on a single set of program instructions.

Abstract: In a 5G network, a profiler component of a network slice controller is arranged to dynamically observe behaviors of pre-defined types of network slices when handling current traffic. The profiler employs the observed behaviors to generate profiles of the pre-defined slice types in terms of throughput, reliability, or other suitable metrics. In response to a request from an application for admission to the 5G network for which an ID of an appropriate pre-defined network slice type is unknown, the application request and traffic is handled on a slice which is temporarily utilized while the profiler dynamically observes application behaviors to generate an application profile. The profiler identifies a pre-defined slice type having a profile that is the closest match to the generated application profile. The application may then be moved from the temporary slice to a slice of the identified pre-defined type so that optimal slice characteristics are provided for the application's traffic.

Abstract: In a 5G network, a slice controller is arranged to dynamically configure a radio access network (RAN) by allocating physical radio resources into RAN slices by making predictions of channel state information (CSI) for user equipment (UE) executing applications that make connectivity requests for admission to particular identified slices. The slice controller grants or denies admission requests based on the predicted CSI to ensure that applicable service level agreement (SLA) guarantees are satisfied for traffic across all the RAN slices. Each time new admission requests are received from applications, the slice controller determines whether a suitable RAN configuration exists that will enable SLA guarantees for the slices to continue to be satisfied for the current traffic while also meeting the SLA guarantees applicable to the new admission request.

Abstract: Described are examples for monitoring performance metrics of one or more workloads in a cloud-computing environment and reallocating compute resources based on the monitoring. Reallocating compute resources can include migrating workloads among nodes or other resources in the cloud-computing environment, reallocating hardware accelerator resources, adjusting transmit power for virtual radio access network (vRAN) workloads, and/or the like.

Abstract: In a 5G network, a slice controller operating in a radio access network (RAN) is arranged to make predictions of channel state information (CSI) for user equipment (UE) on the network using a predictive propagation model. The slice controller uses the predicted CSI to schedule subcarriers and time slots associated with physical radio resources for data transmission on slices of the 5G network between a 5G radio unit (RU) and the UE to maximize network throughput on a slice for the radio spectrum that is utilized for a given time period. In view of the CSI predictions, the slice controller controls operations of the MAC (Medium Access Control) layer functions based on PHY (physical) layer radio resource subsets to schedule the subcarrier and time slots for data transmissions on a slice over the 5G air interface from RU to UE.

Abstract: Described are examples for deploying workloads in a cloud-computing environment. In an aspect, based on a desired number of workloads of a process to be executed in a cloud-computing environment and based on one or more failure probabilities, an actual number of workloads of the process to execute in the cloud-computing environment to provide a level of service can be determined and deployed. In another aspect, a standby workload can be executed as a second instance of the process without at least a portion of the separate configuration used by the multiple workloads, and based on detecting termination of one of multiple workloads, the standby workload can be configured to execute based on the separate configuration of the separate instance of the process corresponding to the one of the multiple workloads.

Abstract: Described are examples for deploying workloads in a cloud-computing environment. In an aspect, based on a desired number of workloads of a process to be executed in a cloud-computing environment and based on one or more failure probabilities, an actual number of workloads of the process to execute in the cloud-computing environment to provide a level of service can be determined and deployed. In another aspect, a standby workload can be executed as a second instance of the process without at least a portion of the separate configuration used by the multiple workloads, and based on detecting termination of one of multiple workloads, the standby workload can be configured to execute based on the separate configuration of the separate instance of the process corresponding to the one of the multiple workloads.

Abstract: Methods and computing devices for allocating computing resources and selecting hyperparameter configurations during continuous retraining and operation of a machine learning model. In one example, a computing device configured to be located at a network edge between a local network and a cloud service includes a processor and a memory storing instructions executable by the processor to operate a machine learning model. During a retraining window, a selected portion of a video stream is selected for labeling. At least a portion of a labeled retraining data set is selected for profiling a superset of hyperparameter configurations. For each configuration of the superset of hyperparameter configurations, a profiling test is performed. The profiling test is terminated, and a change in inference accuracy that resulted from the profiling test is extrapolated. Based upon the extrapolated inference accuracies, a set of selected hyperparameter configurations is output.

Abstract: The devices and methods leverage harmonics to resolve, separate, and identify devices. The devices and methods use the harmonic patterns associated with a frequency modulating (FM) signal to discern and extract information from the FM signal using correlation learning in a crowded spectrum space where the nodes are transmitting simultaneously on multiple channels. The methods and devices leverage harmonics to resolve, separate, and/or identify wireless communication devices.

Abstract: Systems and methods for routing packet data for transmission via a plurality of communication links are described. A method may include dividing a usage cycle for the plurality of communication links into a plurality of timeslots. Packet data traffic demands for the packet data for transmission via the plurality of communication links may be received. Based on a mixed integer linear programming model, an allocation of the packet data traffic demands to the plurality of communication links during the usage cycle may be determined using binary constraints of the mixed integer linear programming model. The binary constraints may prioritize respective subsets of the plurality of timeslots for at least some of the plurality of communication links. For each of the plurality of timeslots, an allocation of the packet data traffic demands to each of the plurality of communication links may be determined using the mixed integer linear programming model.

Abstract: Examples are disclosed that relate to methods and computing devices for allocating computing resources and selecting hyperparameter configurations during continuous retraining and operation of a machine learning model. In one example, a computing device configured to be located at a network edge between a local network and a cloud service comprises a processor and a memory storing instructions executable by the processor to operate a machine learning model. During a retraining window, a selected portion of a video stream is selected for labeling. At least a portion of a labeled retraining data set is selected for profiling a superset of hyperparameter configurations. For each configuration of the superset of hyperparameter configurations, a profiling test is performed. The profiling test is terminated, and a change in inference accuracy that resulted from the profiling test is extrapolated. Based upon the extrapolated inference accuracies, a set of selected hyperparameter configurations is output.

Abstract: A method can include classifying, using a compressed and specialized convolutional neural network (CNN), an object of a video frame into classes, clustering the object based on a distance of a feature vector of the object to a feature vector of a centroid object of the cluster, storing top-k classes, a centroid identification, and a cluster identification, in response to receiving a query for objects of class X from a specific video stream, retrieving image data for each centroid of each cluster that includes the class X as one of the top-k classes, classifying, using a ground truth CNN (GT-CNN), the retrieved image data for each centroid, and for each centroid determined to be classified as a member of the class X providing image data for each object in each cluster associated with the centroid.

Abstract: A multi-hop relay network comprises a set of data nodes arranged in a relay topology, wherein the set of data nodes communicate with one another via a data plane comprising a set of high frequency data links, such as mmWave links. An edge node communicates with one or more of the set of data nodes via the data plane and communicates with the set of data nodes over a control plane comprising a set of low frequency wireless links, such as a Wi-Fi network. The edge node determines a path utilization for the set of high frequency wireless links. When one of the high frequency wireless links is over-utilized, the edge node communicates, to a data node in the set of data nodes via the control plane, a command to change the relay topology. The edge node also determines whether the relay topology is operating at a target accuracy. When it is not, the edge node adjusts a data analytics parameter for a node to improve the overall accuracy of the network.

Abstract: A multi-hop relay network comprises a set of data nodes arranged in a relay topology, wherein the set of data nodes communicate with one another via a data plane comprising a set of high frequency data links, such as mmWave links. An edge node communicates with one or more of the set of data nodes via the data plane and communicates with the set of data nodes over a control plane comprising a set of low frequency wireless links, such as a Wi-Fi network. The edge node determines a path utilization for the set of high frequency wireless links. When one of the high frequency wireless links is over-utilized, the edge node communicates, to a data node in the set of data nodes via the control plane, a command to change the relay topology. The edge node also determines whether the relay topology is operating at a target accuracy. When it is not, the edge node adjusts a data analytics parameter for a node to improve the overall accuracy of the network.

Abstract: A thin-cloud system for distributing content, for example, live streaming video content, from a broadcaster to a viewer is provided herein. The computing devices of the broadcaster can provide the multi-bitrate transcoding, of the two or more bitstreams, sent to a file server, which alleviates the need for the file server to encode the streams for a viewer. These multiple streams are received by a file server for provision to one or more viewers. The viewers can receive the streams at one of the two or more bitrates. If the viewer receives the content at a lower bitrate, the viewers can employ a machine learning (ML) co-processor that can operate as an accelerator to improve the inbound content, if that content is provided at a lower bitrate, and thus, a lower resolution. The file server can train and provide the ML models used for the acceleration.