Abstract: Technologies for enhanced network discovery and configuration include a network with a fabric manager and multiple network devices. A network device requests platform information from a management controller and receives the platform information via a sideband interface. The network device broadcasts a discovery message indicative of the platform information on a link layer network. The fabric manager discovers the network topology with an enhanced link layer discovery protocol and creates a vPOD in the network. The vPOD includes an application network with multiple racks. The fabric manager creates a tagged network domain for the vPOD. The fabric manager sends an out-of-band configuration command to the network device with a tag associated with the vPOD. After receiving the out-of-band configuration command, the network device receives a packet, compares domain metadata of the packet to the tag received from the fabric manager, and routes the packet. Other embodiments are described and claimed.

Abstract: Examples described herein relate to offload circuitry comprising one or more compute engines that are configurable to perform a workload offloaded from a process executed by a processor based on a descriptor particular to the workload. In some examples, the offload circuitry is configurable to perform the workload, among multiple different workloads. In some examples, the multiple different workloads include one or more of: data transformation (DT) for data format conversion, Locality Sensitive Hashing (LSH) for neural network (NN), similarity search, sparse general matrix-matrix multiplication (SpGEMM) acceleration of hash based sparse matrix multiplication, data encode, data decode, or embedding lookup.

Abstract: Various systems and methods for implementing a multi-access edge computing (MEC) based system to realize 5G Network Edge and Core Service Dimensioning using Machine Learning and other Artificial Intelligence Techniques, for improved operations and usage of computing and networking resources, and are disclosed herein. In an example, processing circuitry of a compute node on a network is used to analyze execution of an application to obtain operational data. The compute node then may modularize functions of the application based on the operational data to construct modularized functions. A phase transition graph is constructed using a machine-learning based analysis, the phase transition graph representing state transitions from one modularized function to another modularized function, where the phase transition graph is used to dimension the application by distributing the modularized functions across the network.

Abstract: In accordance with some embodiments, a cloud service provider may operate a data center in a way that dynamically reallocates resources across nodes within the data center based on both utilization and service level agreements. In other words, the allocation of resources may be adjusted dynamically based on current conditions. The current conditions in the data center may be a function of the nature of all the current workloads. Instead of simply managing the workloads in a way to increase overall execution efficiency, the data center instead may manage the workload to achieve quality of service requirements for particular workloads according to service level agreements.

Abstract: Technologies for network interface controllers (NICs) include a compute sled and an accelerator sled in communication over a network. The accelerator sled configures a virtual switch endpoint associated with a remote direct memory access (RDMA) server instance that is associated with a field-programmable gate array (FPGA) of the accelerator sled. The accelerator sled updates local software defined networking (SDN) tables with a virtual tunnel associated with the virtual switch endpoint and a remote compute sled. A virtual switch of the accelerator sled switches virtual tunnel traffic from the remote compute sled to the RDMA server instance, which transfers data to or from the FPGA. The compute sled also updates a local SDN table with the virtual tunnel, and a virtual switch of the compute sled switches virtual tunnel traffic to or from the accelerator sled. Other embodiments are described and claimed.

Abstract: Examples described herein relate to network layer 7 (L7) offload to an infrastructure processing unit (IPU) for a service mesh. An apparatus described herein includes an IPU comprising an IPU memory to store a routing table for a service mesh, the routing table to map shared memory address spaces of the IPU and a host device executing one or more microservices, wherein the service mesh provides an infrastructure layer for the one or more microservices executing on the host device; and one or more IPU cores communicably coupled to the IPU memory, the one or more IPU cores to: host a network L7 proxy endpoint for the service mesh, and communicate messages between the network L7 proxy endpoint and an L7 interface device of the one or more microservices by copying data between the shared memory address spaces of the IPU and the host device based on the routing table.

Abstract: Technologies for contention-aware cloud compute scheduling include a number of compute nodes in a cloud computing cluster and a cloud controller. Each compute node collects performance data indicative of cache contention on the compute node, for example, cache misses per thousand instructions. Each compute node determines a contention score as a function of the performance data and stores the contention score in a cloud state database. In response to a request for a new virtual machine, the cloud controller receives contention scores for the compute nodes and selects a compute node based on the contention score. The cloud controller schedules the new virtual machine on the selected compute node. The contention score may include a contention metric and a contention score level indicative of the contention metric. The contention score level may be determined by comparing the contention metric to a number of thresholds. Other embodiments are described and claimed.

Abstract: Technologies for switch link and ply management for variable oversubscription ratios include powering up and down links of one or more network plys according to bandwidth demand, desired oversubscription ratio and/or other parameters. Telemetry data representing one or more network traffic metrics of one or more switch plies is monitored to determine respective power states of the plurality of links associated with the one or more switch plies as a function of a desired oversubscription ratio calculated based on the telemetry data. The respective power state of the plurality of links is set accordingly.

Abstract: Technologies for managing burst bandwidth requirements are disclosed. In the illustrative embodiment, a software-defined network (SDN) controller monitors storage devices in a data center. If a storage device fails, the SDN controller manages the bandwidth used to replicate the data that was stored on the failed storage device. The SDN controller may allocate an initial amount of bandwidth based on one or more parameters of the storage device, and the SDN controller may increase the bandwidth in a series of discrete steps. In another embodiment, the SDN controller may predict a bandwidth burst based on sequential writes at a storage sled from several compute devices, and allocate bandwidth accordingly in a tiered manner.

Abstract: Systems and techniques for an objective driven orchestration are described herein. In an example, system is adapted to receive a request for a service workload, including a set of performance objectives for the service workload. The set of performance objectives may indicate Quality of Service (QoS) for the performance of the service workload. The system may be further adapted to determine a plan for the service workload. The plan may orchestrate a set of actions to fulfill the set of performance objectives. The system may be further adapted to initiate execution of the service workload in accordance with the set of actions of the plan. They system may be further adapted to monitor execution of the service workload for an effect associated with each action of the set of actions and compliance with the set of performance objectives.

Abstract: Technologies for allocating resources of managed nodes to workloads to balance multiple resource allocation objectives include an orchestrator server to receive resource allocation objective data indicative of multiple resource allocation objectives to be satisfied. The orchestrator server is additionally to determine an initial assignment of a set of workloads among the managed nodes and receive telemetry data from the managed nodes. The orchestrator server is further to determine, as a function of the telemetry data and the resource allocation objective data, an adjustment to the assignment of the workloads to increase an achievement of at least one of the resource allocation objectives without decreasing an achievement of another of the resource allocation objectives, and apply the adjustments to the assignments of the workloads among the managed nodes as the workloads are performed. Other embodiments are also described and claimed.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed for dynamic load balancing for multi-core computing environments. An example apparatus includes a first and a plurality of second cores of a processor, and circuitry in a die of the processor separate from the first and the second cores, the circuitry to enqueue identifiers in one or more queues in the circuitry associated with respective ones of data packets of a packet flow, allocate one or more of the second cores to dequeue first ones of the identifiers in response to a throughput parameter of the first core not satisfying a throughput threshold to cause the one or more of the second cores to execute one or more operations on first ones of the data packets, and provide the first ones to one or more data consumers to distribute the first data packets.

Abstract: Technologies for enhanced network discovery and configuration include a network with a fabric manager and multiple network devices. A network device requests platform information from a management controller and receives the platform information via a sideband interface. The network device broadcasts a discovery message indicative of the platform information on a link layer network. The fabric manager discovers the network topology with an enhanced link layer discovery protocol and creates a vPOD in the network. The vPOD includes an application network with multiple racks. The fabric manager creates a tagged network domain for the vPOD. The fabric manager sends an out-of-band configuration command to the network device with a tag associated with the vPOD. After receiving the out-of-band configuration command, the network device receives a packet, compares domain metadata of the packet to the tag received from the fabric manager, and routes the packet. Other embodiments are described and claimed.

Abstract: Technologies for load balancing a storage network include a system. The system includes circuitry to adjust routing rules in a network interface controller to deliver a packet from one of multiple uplinks to one of any physical functions, circuitry to remap, in response to a failure of a switch, a port from one physical function to another physical function, and circuitry to communicate control data between a software defined network controller and one or more agents in one or more host endpoints with a hierarchical distributed hashing table.

Abstract: Technologies for allocating resources of managed nodes to workloads to balance multiple resource allocation objectives include an orchestrator server to receive resource allocation objective data indicative of multiple resource allocation objectives to be satisfied. The orchestrator server is additionally to determine an initial assignment of a set of workloads among the managed nodes and receive telemetry data from the managed nodes. The orchestrator server is further to determine, as a function of the telemetry data and the resource allocation objective data, an adjustment to the assignment of the workloads to increase an achievement of at least one of the resource allocation objectives without decreasing an achievement of another of the resource allocation objectives, and apply the adjustments to the assignments of the workloads among the managed nodes as the workloads are performed. Other embodiments are also described and claimed.

Abstract: Technologies for contention-aware cloud compute scheduling include a number of compute nodes in a cloud computing cluster and a cloud controller. Each compute node collects performance data indicative of cache contention on the compute node, for example, cache misses per thousand instructions. Each compute node determines a contention score as a function of the performance data and stores the contention score in a cloud state database. In response to a request for a new virtual machine, the cloud controller receives contention scores for the compute nodes and selects a compute node based on the contention score. The cloud controller schedules the new virtual machine on the selected compute node. The contention score may include a contention metric and a contention score level indicative of the contention metric. The contention score level may be determined by comparing the contention metric to a number of thresholds. Other embodiments are described and claimed.

Abstract: In one embodiment, an edge computing device for performing decision tree training and inference includes interface circuitry and processing circuitry. The interface circuitry receives training data and inference data that is captured, at least partially, by sensor(s). The training data corresponds to a plurality of labeled instances of a feature set, and the inference data corresponds to an unlabeled instance of the feature set. The processing circuitry: computes a set of feature value checkpoints that indicate, for each feature of the feature set, a subset of potential feature values to be evaluated for splitting tree nodes of a decision tree model; trains the decision tree model based on the training data and the set of feature value checkpoints; and performs inference using the decision tree model to predict a target variable for the unlabeled instance of the feature set.

Abstract: Technologies for dynamically allocating resources among a set of managed nodes include an orchestrator server to receive telemetry data from the managed nodes indicative of resource utilization and workload performance by the managed nodes as the workloads are executed, generate a resource allocation map indicative of allocations of resources among the managed nodes, determine, as a function of the telemetry data and the resource allocation map, a dynamic adjustment to allocation of resources to at least one of the managed nodes to improve performance of at least one of the workloads executed on the at least one of the managed nodes, and apply the adjustment to the allocation of the resources among the managed nodes as the workloads are executed. Other embodiments are also described and claimed.

Abstract: Technologies for load balancing a storage network include a system. The system includes circuitry to adjust routing rules in a network interface controller to deliver a packet from one of multiple uplinks to one of any physical functions, circuitry to remap, in response to a failure of a switch, a port from one physical function to another physical function, and circuitry to communicate control data between a software defined network controller and one or more agents in one or more host endpoints with a hierarchical distributed hashing table.

Abstract: Techniques and apparatus for managing a distributed computing environment using event digests are described. In one embodiment, for example, an apparatus may include at least one memory, and logic for a system manager, at least a portion of the logic comprised in hardware coupled to the at least one memory, the logic to determine a workload to schedule, access an event digest associated with a plurality of compute hosts, the event digest comprising event digest values determined using out-of-band information, determine metrics from the event digest, generate at least one host weight for at least a portion of the plurality of compute hosts based on the metrics, identify at least one candidate host from the portion of the plurality of compute hosts based on the at least one host weight, and schedule the workload on the at least one candidate host. Other embodiments are described and claimed.