Abstract: The disclosure relates to a method and apparatus including setting a memory swap size limit, the limit being lower than a memory swap size defining a maximum size of a part of said memory resources used for swap, obtaining a score for at least one running program, a high score corresponding to a low priority level, obtaining monitoring information representative of a monitored activity of the program during a time period and of a learnt user's habit of use of the program, including a number of times the program gained the focus within the time period. The disclosure also includes deriving a score delta from information with a decrement value to the score delta at each focus gained by the program, adjusting the score by adding the delta, and terminating execution when memory swap size limit is reached and the adjusted score reaches a memory swap size limit threshold.

Abstract: Generally described, one or more aspects of the present application relate to prefetching container image layers for use in a cluster of compute instances. For example, the dependencies among the individual layers within the container images stored and/or executed on a cloud provider network may be analyzed. Then, the layers that are likely to be used by a user of the cloud provider network may be identified and prefetched into the caches of the compute instances provided by the cloud provider network, before execution of the container images including such layers is requested by the user. By doing so, the latency between the time a request to execute a set of container images is received and the time the execution of the set of container images is actually initiated can be reduced, thereby providing an improved and more efficient application execution experience to the user.

Abstract: The subject technology provides for dynamic task allocation for neural network models. The subject technology determines an operation performed at a node of a neural network model. The subject technology assigns an annotation to indicate whether the operation is better performed on a CPU or a GPU based at least in part on hardware capabilities of a target platform. The subject technology determines whether the neural network model includes a second layer. The subject technology, in response to determining that the neural network model includes a second layer, for each node of the second layer of the neural network model, determines a second operation performed at the node. Further the subject technology assigns a second annotation to indicate whether the second operation is better performed on the CPU or the GPU based at least in part on the hardware capabilities of the target platform.

Abstract: An example computer-implemented method is provided that includes: determining a remote computing system associated with a task queue comprising a plurality of processing tasks, wherein the remote computing system updates, based on a hidden schedule, prerequisites for the plurality of processing tasks; initiating the plurality of processing tasks according to a test task schedule constructed to probe the hidden schedule of the remote computing system; receiving a plurality of responses respectively for the plurality of processing tasks, wherein a respective response corresponds to at least one of a success indicator or a failure indicator; estimating, based on the plurality of responses, a prerequisite processing time of the hidden schedule; and generating, based on the estimated prerequisite processing time, a main task schedule for the task queue.

Abstract: In one or more embodiments, one or more systems, one or more methods, and/or one or more processes may determine first thermal attribute values associated with multiple information handling systems (IHSs) with respect to a period of time as the IHSs execute a first workload; determine multiple variance ranges respectively associated with the first thermal attributes; periodically determine second thermal attribute values associated with the IHSs as the IHSs execute a second workload; determine that a thermal attribute value of the second thermal attribute values exceeds a respective variance range of the variance ranges as a first information handling system (IHS) of the IHSs executes the second workload; generate an alert based at least on the thermal attribute value exceeding the respective variance range; and in response to the alert, transfer at least a portion of the second workload from the first IHS to a second IHS of the IHSs.

Abstract: A computer-implemented method, system, and computer program product are provided. The approach includes popping a first task from a queue of a garbage collection thread. The approach also includes identifying a second task from the popped first task, where the second task is associated with a bit of a bitmap and the bit is located at a location within the bitmap that has not yet been passed by a finger pointer. The approach also includes pushing the second task to the queue of the garbage collection thread.

Abstract: Embodiments provide data in-flight (DIF) services to software applications such as virtual machines (VMs) at an application level without requiring modification to established storage protocols. In exemplary embodiments, a storage controller transmits an advertisement of one or more data in-flight (DIF) services supported by a storage container of the storage controller. One or more DIF services communication path is created with attributes corresponding to the DIF services supported by the storage container. The storage controller receives, over the DIF services communication path, tagged data that can include data transmitted by a virtual machine (VM) for storage in the storage container.

Abstract: A system and method include classifying and assigning virtual disks accessed from compute only nodes. The method determines, by a management processor of a virtual computing system, characteristics for a plurality of virtual disks hosted on a plurality of hyper converged nodes in a cluster of nodes in the virtual computing system. The method further classifies, by the management processor, each of the plurality of virtual disks based on the determined characteristics and identifies, by the management processor, one of the plurality of virtual disks to host data for a virtual machine on a compute only node based on the classification to spread out input-output demand in the cluster, reducing probability of input-output bottlenecks and increasing cluster-wide storage throughput. The method also assigns, by the management processor, the identified virtual disk to host data for the virtual machine located on the compute only node.

Abstract: Disclosed are various embodiments for separating creation of and executing of updated campaigns in multitenancy environments. For example, an organization can have a hierarchical structure such that a root organization includes multiple sub-organizations. A root organization administrator may create an update campaign for enforcing software policies on devices that are included in the sub-organizations. A sub-organization administrator can approve and define the execution schedule for when the devices included in the sub-organization can execute the update campaign.

Abstract: Performing predictive analysis on running batch jobs is provided. A series of batch end time predictive models is retrieved according to a sequence of milestone jobs in a batch of jobs. Retrieved batch end time predictive models are assembled into an aggregate batch end time predictive model to increase accuracy and stability of an end time prediction of the batch of jobs. The aggregate batch end time predictive model is utilized to predict an end time of the batch of jobs during running of the batch of jobs to form a predicted end time of the batch of jobs.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for scheduling operations represented as a computational graph on a distributed computing network. A method includes: receiving data representing operations to be executed in order to perform a job on a plurality of hardware accelerators of a plurality of different accelerator types; generating, for the job and from at least the data representing the operations, features that represent a predicted performance for the job on hardware accelerators of the plurality of different accelerator types; generating, from the features, a respective predicted performance metric for the job for each of the plurality of different accelerator types according to a performance objective function; and providing, to a scheduling system, one or more recommendations for scheduling the job on one or more recommended types of hardware accelerators.

Abstract: Testing methods and systems are provided for testing a resource manager of an application management system. The testing systems include a load orchestrator configured to obtain an artificial metric that is determined based on a utilization model (e.g., CPU usage, memory allocation, or disk usage, number of webserver sessions). The load orchestrator transmits the artificial metric to applications in a cluster of computing nodes. The applications transmit the artificial metric to the resource manager. In response, the resource manager generates control output for managing applications in the cluster based on the artificial metric (e.g., scaling, load balancing, application placement, failover of applications, or defragmenting data). The utilization model may include executable code for generating artificial metric values. The model may be received as a result of an API call. The load orchestrator may be instantiated in an orchestration framework or in each node of the cluster.

Abstract: A computer-implemented method, according to one embodiment, includes: receiving a data process that includes a plurality of sub-processes. A unique subset of the sub-processes is assigned to each of: a managing thread, and at least one other thread. Moreover, performance characteristics of each of the threads is evaluated while the respective subsets of sub-processes are being performed, and a determination is made as to whether the performance characteristics of each of the threads are substantially equal to the performance characteristics of each of the other threads. In response to determining that performance characteristics of each of the threads are not substantially equal, the subsets of the sub-processes are dynamically adjusted such that the performance characteristics of each of the threads become more equal. Moreover, the adjusted subsets of the sub-processes are reassigned to each of the managing thread and at least one other thread.

Abstract: Techniques for providing high-performance buffer caches for transactional input/output (I/O) systems are disclosed. The techniques include obtaining a first logical creation time of a resource to be acquired by the first transaction during a pre-commit phase of a first transaction with an I/O system. When the first logical creation time exceeds a latest logical creation time from a set of resources previously acquired by the first transaction, the first logical creation time of the resource is compared with an earliest logical termination time from the set of resources. When the first logical creation time of the resource exceeds the earliest logical termination time from the set of resources, a conflict between the resource and the set of resources is detected, and a restart of the first transaction is triggered.

Abstract: A method for allocating on-chip memory of a neural processing unit is performed by one or more processors, and includes deallocating an allocated chunk in an on-chip memory area, which is finished with the use of the memory, and converting it into a cached chunk, receiving an on-chip memory allocation request for specific data, determining whether there is a cached chunk of one or more cached chunks that is allocable for the specific data, based on a comparison between a size of the specific data and the size of the one or more cached chunks, and based on a result of determining whether there is the cached chunk that is allocable for the specific data, allocating the specific data to a specific cached chunk of the one or more cached chunks, or allocating the specific data to at least a portion of the free chunk.

Abstract: Executing a big data analytics pipeline in a storage system that includes compute resources and shared storage resources, including: receiving, from a data producer, a dataset; storing, within the storage system, the dataset; allocating processing resources to an analytics application; and executing the analytics application on the processing resources, including ingesting the dataset from the storage system.

Abstract: Methods and systems for selecting and executing optimal execution strategies for processes, wherein an optimal execution strategy may be selected from a plurality of execution strategies for example in a look-up table, based on specific process values.

Abstract: A Multi-site Integrated Session-oriented Transaction (MIST) computing system includes a MIST mapper configured to process a compiled distributed compute session request to determine, for each of at least one remote sets of data, whether to process the remote set of data with one of one or more field nodes or with a cloud node based on a cost of transporting the remote set of data from the proximate field node to the cloud node. In response to determining to process the remote set of data with the cloud node, the MIST mapper is configured to output a data retrieval request to the proximate field node. In response to determining to process the remote set of data with the field node, the MIST mapper is configured to output the computation request to the proximate field node.

Abstract: Techniques for tracking a lock involve creating an event comprising object information, and extracting the object information in the event through a thread in a thread pool. Such techniques further involve recording lock information comprising the object information according to a determination that the thread executes a lock operation. Such a lock tracking technique for a threading model can record lock information of an object when a thread performs locking and unlocking, and realize tracking of a lock operation and generating a lock analysis report.

Abstract: A control plane system can be used to manage or generated components in a shared computing resource environment. To generate a modified components, the control plane system can receive receiving configurations of a component. The configurations can include software versions and/or parameters for the component. Using the configurations, the control plane system can generate an image of a modified component, and communicate the image to a master node in the shared computing resource environment. The master node can provides one or more instances of the modified component for use based on the received image.