Abstract: Systems and methods for offloading data storage processing tasks from a data storage device to a graphics processing unit data are described. Data storage devices may include a peripheral interface configured to connect to a host system and provide access to a host memory buffer. The data storage device may store task input data to the host memory buffer. The data storage device may notify a processor device including the graphics processing unit to initiate the storage processing task. The processor device may access the task input data from the host memory buffer and store the task output data to the host memory buffer for access by the data storage device.

Abstract: Aspects of the present disclosure relate to a dynamic resource management system. In various examples, a dynamic resource management system enables resource-consuming tasks to be assigned to resources at least according to resource profiles and resource groupings. Resources may include primary resources and secondary resources and may be grouped into resource groups such that each resource group have a selection of primary resources and secondary resources. In some instances, resource groups may be updated to have updated selections of primary resources and secondary resources. To assign tasks to resources, the dynamic resource management system may first identify multiple resource-consuming tasks associated with a resource-consuming event then assign each resource-consuming task to a resource. The task assignment may be performed based on resource information such as availability, performance, and/or experience.

Abstract: An offload analyzer analyzes a program for porting to a heterogeneous computing system by identifying code objects for offloading to an accelerator. Runtime metrics generated by executing the program on a host processor unit are provided to an accelerator model that models the performance of the accelerator and generates estimated accelerator metrics for the program. A code object offload selector selects code objects for offloading based on whether estimated accelerated times of the code objects, which comprise estimated accelerator times and offload overhead times, are better than their host processor unit execution times. The code object offload selector selects additional code objects for offloading using a dynamic-programming-like performance estimation approach that performs a bottom-up traversal of a call tree. A heterogeneous version of the program can be generated for execution on the heterogeneous computing system.

Abstract: Host node container packing is provided. A minimum number of host nodes in a container orchestration environment that satisfies software resource dependencies of each respective container is determined using a dependency graph based on a group of containers that each respective host node of the minimum number of host nodes can run using shared software resources. Each respective group of containers is scheduled to run on a corresponding host node of the minimum number of host nodes using the shared software resources needed by that particular group of containers.

Abstract: A method of constructing sub-graphs, includes receiving a directed acyclic graph (DAG), partitioning the directed acyclic graph into an at least one section, determining at least one hardware attribute, determining at least one DAG hardware limitation of the at least one section and determining a largest continuous node list of the at least one section in which the at least one hardware attribute meets the at least one DAG hardware limitation.

Abstract: There is disclosed a computer-implemented system and method of classifying a target sample, wherein the target sample is a computer object having a feature vector, the method comprising: creating n sorted containers, comprising sorting a universe of samples based on feature vector distances from the samples to n vantage points, wherein n is a positive integer; storing the n sorted containers to a computer memory; bucketizing the n sorted containers; for the n vantage points, selecting, from the n bucketized sorted containers, n meta-buckets of that the target sample belongs to; creating an intersection container, comprising samples that appear in all n meta-buckets; selecting, as a target cluster, samples from the intersection container that have a feature vector distance from the target sample less than a threshold; and acting on the target cluster.

Abstract: A computer implemented method includes monitoring resource utilization for multiple programs running on a user device. A current user interaction with the programs is detected and a usage contextual profile representing user interaction with the programs is derived. The monitored resource utilization is compared to a performance threshold and one of the multiple programs is distributed for execution elsewhere in response to the comparing to optimize user experience on the user device in accordance with the usage contextual profile.

Abstract: A statistics and feedback-based diagnostic scan framework for cluster nodes is provided, that prioritizes work items for processing by cluster nodes according to work item priority. A heatmap of work items is created and used by the diagnostic scan framework to prioritize work items for processing by the cluster nodes. The heatmap may be created from runtime IO operations, metadata access/updates, or other metrics derived from systems executing within the storage system. Additionally, as cluster nodes process allocated work items, the results of the scan operations are used to update the heat map to adjust the heat map based on the results of the previous scan operations. The heat map may prioritize individual work items or may create regions of work items and prioritize processing of work items according work item region.

Abstract: One example method includes running a workload through a trained open-set classification model, recovering, as a result of the running, a class and an open-setness score corresponding to the workload, determining, based on the class and the open-setness score, whether the workload is new, and when the workload is determined to be new, starting a new cluster that includes the workload. A response time predictor model may be used to predict a response time associated with the new workload.

Abstract: In a method for resource management, a plurality of application processes is received. At least a processor is allocated to the plurality of application processes. For each of the plurality of application processes allocated to the processor, a quantity of remaining available resources for the respective application process in current predetermined resources of the processor is determined. The quantity of remaining available resources is associated with a quantity of remaining available resources for the respective application process in previous predetermined resources of the processor. A resource allocation command for each of the plurality of application processes is determined based on the quantity of remaining available resources for the respective application process in the current predetermined resources. The resource allocation command indicates whether the respective application process is processed.

Abstract: A program executing in kernel space monitors system calls associated with access to an operating system resource. Based on the monitoring, an access attempt to the operating system resource is detected. In response to detecting the access attempt, a process configured to handle the access to the operating system resource is spawned. The process may then terminated based on an expiration of an idle period.

Abstract: Systems and methods for managing workspaces across multiple clients are described. In some embodiments, a Workspace Managing Device (WMD) may include a processor and a memory coupled to the processor, the memory having program instructions stored thereon that, upon execution, cause the WMD to: receive a request from a client Information Handling System (IHS) to access a peripheral device coupled to the WMD; identify a workspace management policy applicable to at least one of the request, the peripheral device, the client IHS, or a user of the client IHS; and manage access to the peripheral device by the client IHS based upon the workspace management policy.

Abstract: A system, includes processing circuitry; and a memory connected to the processing circuitry, wherein the memory is configured to store executable instructions that, in response to being executed by the processing circuitry, facilitate performance of operations to request deployment software for a telecommunications system from a central inventory; send the deployment software to a load balancer that distributes executable portions of the deployment software to two or more cloud adaptors (CAs) that are operably connected to the load balancer; generate worker threads based on the distributed executable portions of the deployment software; and distribute the worker threads to hybrid clusters for execution of the worker threads and implementation of the deployment software.

Abstract: A method of using a multi-cluster network is provided. The multi-cluster network has at least a plurality of clusters, where each cluster of the plurality of clusters has at least a node and a pod. The method includes collecting cluster and application information of the multi-cluster network. The cluster and application information includes at least a cluster capacity and an application performance metric. The application performance metric corresponds to at least an application. The method further includes analyzing the cluster and application information for a current pod arrangement on each cluster and adjusting at least the application performance metric based at least on the analyzed cluster and application information by taking at least a pod-based action achieving a new pod arrangement on a target cluster. An apparatus for using the multi-cluster network is also provided.

Abstract: Systems and methods for time-based resource allocation are disclosed herein. A system receives requests for intermediate allocation of a set of resources during a first period of time. Each request may include a target entity to receive a portion of the intermediate allocation and a user-linked resource token. The system determines a first subset of the set of resources for which a final allocation of resources from the user is requested when the first period is elapsed. During the first period, the system transfers the set of resources to target entities. During a second period, the system receives a message from a target entity indicating a return of at least a portion of the intermediate allocation. The system updates the set of resources and calculates an updated first subset of resources. Responsive to determining that the second period is elapsed, the system may automatically transmit a resource allocation command configured to trigger final allocation.

Abstract: Example methods and systems for elastic provisioning of container-based graphics processing unit (GPU) nodes are described. In one example, a computer system may monitor usage information associated with a pool of multiple container-based GPU nodes. Based on the usage information, the computer system may apply rule(s) to determine whether capacity adjustment is required. In response to determination that capacity expansion is required, the computer system may configure the pool to expand by adding (a) at least one container-based GPU node to the pool, or (b) at least one container pod to one of the multiple container-based GPU nodes. Otherwise, in response to determination that capacity shrinkage is required, the computer system may configure the pool to shrink by removing (a) at least one container-based GPU node, or (b) at least one container pod from the pool.

Abstract: A system and method for executing self-evolving property graphs on GPU hardware for unbounded experiential processing. The system stores a dynamic property graph comprising event and communication vertices in GPU memory. Input streams are projected to graph vertices through specialized operators. Multiple GPU-executable operator kernels transform the graph through geometric operations including diffusion, geodesic computation, and curvature analysis. These operators are captured as a directed acyclic graph that executes repeatedly without external scheduling, with execution frequency adjusted based on a logarithmic relationship with input stream density. A compression mechanism identifies and removes redundant graph elements based on geometric properties, maintaining memory growth proportional to the logarithm of processed inputs.

Abstract: A method for managing VDI environments includes: obtaining a first resource related parameter (FRRP) for a first resource (FR) and a second resource related parameter (SRRP) for a second resource (SR) from a database; assembling a set of API combinations to generate a set of VDI environments; testing, for a per user average resource utilization value (PUARUV) of the FR and SR, each VDI environment across a range of users based on the FRRP and SRRP; providing the range of users and set of VDI environments to an analyzer, in which the analyzer is instructed to generate a model that ranks the set of VDI environments to obtain a VDI environment that provides the lowest PUARUV; generating a trained model by training the model using the range of users, set of VDI environments, PUARUVs, FRRP, and SRRP; and initiating notification of an administrator about the trained model.

Abstract: Technologies for composing a managed node with multiple processors on multiple compute sleds to cooperatively execute a workload include a memory, one or more processors connected to the memory, and an accelerator. The accelerator further includes a coherence logic unit that is configured to receive a node configuration request to execute a workload. The node configuration request identifies the compute sled and a second compute sled to be included in a managed node. The coherence logic unit is further configured to modify a portion of local working data associated with the workload on the compute sled in the memory with the one or more processors of the compute sled, determine coherence data indicative of the modification made by the one or more processors of the compute sled to the local working data in the memory, and send the coherence data to the second compute sled of the managed node.

Abstract: A plurality of resource usage profiles is generated, wherein at least two of the resource usage profiles each include a corresponding plurality of resource values that quantify real-time computing resources used by an instance of an application that previously executed under two different corresponding operating conditions. It is determined that a new instance of the application is to be initiated. A particular resource usage profile from the plurality of resource usage profiles is selected. The new instance is initiated using the particular resource usage profile.