Abstract: Example implementations relate to determining and implementing a feasible resource optimization plan for public cloud consumption. Telemetry data over a period of time is obtained for a current deployment of virtual infrastructure resources within a current data center of a cloud provider that supports an existing service and an application deployed on the virtual infrastructure resources. Information regarding a set of constraints to be imposed on a resource optimization plan is obtained. Indicators of resource consumption relating to the currently deployed virtual infrastructure resources during the period of time are identified by applying a deep learning algorithm to the telemetry data. A resource optimization plan is determined that is feasible within the set of constraints based on a costing model associated with resources of an alternative data center of the cloud provider, the indicators of resource consumption and costs associated with the current deployment.

Abstract: Workload profiling can be used in a distributed computing environment for automatic performance tuning. For example, a computing device can receive a performance profile for a workload in a distributed computing environment. The performance profile can indicate resource usage by the workload in the distributed computing environment. The computing device can determine a performance bottleneck associated with the workload based on the resource usage specified in the performance profile. A tuning profile can be selected to reduce the performance bottleneck associate with the workload. The computing device can output a command to adjust one or more properties of the workload in accordance with the tuning profile to reduce the performance bottleneck associated with the workload.

Abstract: The embodiments herein describe hybrid parallelism techniques where a mix of data and model parallelism techniques are used to split the workload of a layer across an array of processors. When configuring the array, the bandwidth of the processors in one direction may be greater than the bandwidth in the other direction. Each layer is characterized according to whether they are more feature heavy or weight heavy. Depending on this characterization, the workload of an NN layer can be assigned to the array using a hybrid parallelism technique rather than using solely the data parallelism technique or solely the model parallelism technique. For example, if an NN layer is more weight heavy than feature heavy, data parallelism is used in the direction with the greater bandwidth (to minimize the negative impact of weight reduction) while model parallelism is used in the direction with the smaller bandwidth.

Abstract: A method of migrating a virtual machine having a virtual device that is backed by direct passthrough hardware, from a source host to a destination host, includes the steps of determining whether or not the destination host has direct passthrough hardware that can back the virtual device, and upon determining that the destination host has direct passthrough hardware that can back the virtual device, determining if a version of the direct passthrough hardware at the source host matches a version of the direct passthrough hardware at the destination host. If the versions do not match, the steps further include quiescing the virtual device, deleting data structures relating to the virtual device, and then migrating the virtual machine from the source host to the destination host. If the versions match, the virtual machine is migrated without quiescing the virtual device and without deleting the data structures relating to the virtual device.

Abstract: A system to facilitate infrastructure management is described. The system includes one or more processors and a non-transitory machine-readable medium storing instructions that, when executed, cause the one or more processors to execute an infrastructure management controller to receive first monitoring data indicating a first infrastructure condition occurring at an on-premise infrastructure controller, determine a first load state of the on-premise infrastructure controller based on the first infrastructure condition and adjust a consistency level of the on-premise infrastructure controller to a first level of the consistency based on the first state.

Abstract: A system and method for executing multi-stage distributed computing operations initiates an operation workflow for a multi-stage distributed computing operation in response to a request to execute the multi-stage distributed computing operation. The operation workflow includes tasks of the multi-stage distributed computing operation that are executed by a plurality of service compute nodes, which are monitored to detect any failures. When a failure of the operation workflow for the multi-stage distributed computing operation is detected, a rollback workflow for the multi-stage distributed computing operation is initiated, which includes rollback tasks that correspond to the tasks of the operation workflow to roll back executed tasks of the operation workflow. The rollback workflow is an independent workflow from the operation workflow.

Abstract: Concepts and technologies directed to dedicated optical distribution of computing resources in virtualized environments are disclosed herein. In various aspects, a system can include a processor and memory storing instructions that, upon execution, cause performance of operations. The operations can include receiving a virtual machine creation request that includes a virtual processing requirement and a virtual memory requirement for a virtual machine. The operations can include accessing a physical host infrastructure map that identifies remainder resources from physical host servers within a datacenter. The operations can include creating a simulation test routine and assembling a candidate resource set from the remainder resources. The operations can include establishing a dedicated processing path and a dedicated memory path for the candidate resource set.

Abstract: Embodiments of the present disclosure relate to requirement-based resource sharing in a computing environment. In an embodiment, a computer-implemented method is disclosed. According to the method, it is determined whether a plurality of services to be provisioned in a computing environment match with each other in resource consumption based on respective resource requirements of the plurality of services. In response to determining that the plurality of services match with each other, a resource pool is allocated in the computing environment to be shared by the plurality of services. The resource pool has a resource quota determined based on at least one of the resource requirements. The plurality of services is caused to be provisioned in the computing environment using the resource pool. In other embodiments, a computer system and a computer program product are disclosed.

Abstract: Mechanisms and techniques are employed for managing the allocation and load balancing of storage system resources for the containerized, distributed execution of applications on a storage system. A control component executing on a processing component of the storage system may control reserving the necessary resources on one or more processing components to implement an application, and control a container management module to create, deploy and/or modify one or more containers on one or more processing components of the storage system. The one or more containers then may be executed to implement the application. Multiple processing components of the storage system may have a resource management module executing thereon. The control component may exchange communications with the one or more resource management modules of each processing component to determine the resources available within the processing component; e.g.

Abstract: An overall access rating for each user in a plurality of users for accessing a computing resource of a set of computing resources is generated. Reduced performance of the computing resource is identified. Access metrics associated with each user in the plurality of users who are accessing the computing resource during the reduced performance of the computing resource are determined. The generated overall access ratings based on the determined access metrics are modified. Access to the computing resource is granted based on a ranking of the modified overall access ratings.

Abstract: A scheduler manages execution of a plurality of data-collection jobs, assigns individual jobs to specific forwarders in a set of forwarders, and generates and transmits tokens (e.g., pairs of data—collection tasks and target sources) to assigned forwarders. The forwarder uses the tokens, along with stored information applicable across jobs, to collect data from the target source and forward it onto an indexer for processing. For example, the indexer can then break a data stream into discrete events, extract a timestamp from each event and index (e.g., store) the event based on the timestamp. The scheduler can monitor forwarders' job performance, such that it can use the performance to influence subsequent job assignments. Thus, data-collection jobs can be efficiently assigned to and executed by a group of forwarders, where the group can potentially be diverse and dynamic in size.

Abstract: Methods, systems, and other embodiments for protecting a shared resource in a data processing system are described herein. In one embodiment, when responding to a request that requires use of the shared resource, determine whether a previously established sleep period for the shared resource has expired. If so, determine a health status of the shared resource based on one or more of applying an asymptotic transform to a percentage of time that a processing unit is idle and applying a linear transform to a session count for the shared resource. Then, determine whether the health status is sufficient to permit responding to the request; and if not, (i) determining an appropriate sleep period for the shared resource, (ii) initiating a sleep state for the shared resource for a period substantially equal to the determined appropriate sleep period; and (iii) exiting the sleep state and responding to the request.

Abstract: A computer implemented system is described for assigning executable jobs to pipeline sets, whereby the jobs may be network based computer jobs. The assigning includes generating a weight for each pipeline set of multiple pipeline sets to obtain multiple weights. Generating a weight includes obtaining duty cycle metrics for pipeline software threads in the pipeline set. The duty cycle metrics include a measure of an amount of time that a corresponding pipeline thread is executing and actively processing data. Generating the weight further includes determining the weight for the pipeline set based at least in part on the duty cycle metrics. The method further includes assigning a job request to a target pipeline set selected from the pipeline sets according to a weighted random algorithm, wherein the weighted random algorithm uses the weights.

Abstract: A system to facilitate infrastructure management is described. The system includes one or more processors and a non-transitory machine-readable medium storing instructions that, when executed, cause the one or more processors to execute an infrastructure management controller to automatically balance utilization of infrastructure resources between a plurality of on-premise infrastructure controllers.

Abstract: A snapshot handler is registered with an event monitoring subsystem of a process virtual machine, the registering specifying a trigger event of the snapshot handler, wherein the trigger event comprises execution of a specified portion of an application executing in the process virtual machine, the trigger event specified externally from a source code of the application. Responsive to the event monitoring subsystem detecting an occurrence of the trigger event, the snapshot handler is executed, storing data of an execution state of the process virtual machine at a time of occurrence of the trigger event.

Abstract: A storage system has a QOS recommendation engine that monitors storage system operational parameters and generates recommended changes to host QOS metrics (throughput, bandwidth, and response time requirements) based on differences between the host QOS metrics and storage system operational parameters. The recommended host QOS metrics may be automatically implemented to adjust the host QOS metrics. By reducing host QOS metrics during times where the storage system is experiencing high volumes of workload, it is possible to throttle workload at the host computer rather than requiring the storage system to expend processing resources associated with queueing the workload prior to processing. This can enable the overall throughput of the storage system to increase. When the workload on the storage system is reduced, updated recommended host QOS metrics are provided to enable the host QOS metrics to increase. Historical analysis is also used to generate recommended host QOS metrics.

Abstract: A performance manager (400, 500) and a method (200) performed thereby are provided, for managing the performance of a logical server of a data center. The data center comprises at least one memory pool in which a memory block has been allocated to the logical server. The method (200) comprises determining (230) performance characteristics associated with a first portion of the memory block, comprised in a first memory unit of the at least one memory pool; and identifying (240) a second portion of the memory block, comprised in a second memory unit of the at least one memory pool, to which data of the first portion of the memory block may be migrated to apply performance characteristics associated with the second portion. The method (200) further comprises initiating migration (250) of the data to the second portion of the memory block.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating customized recommendations for environmentally-conscious cloud computing frameworks for replacing computing resources of existing datacenters. One of the methods involves receiving, through a user interface presented on a display of a computing device, data regarding a user's existing datacenter deployment and the user's preferences for the new cloud computing framework, generating one or more recommendations for environmentally-conscious cloud computing frameworks based on the received data, and presenting such recommendations through the user interface for the user's review and consideration.

Abstract: Systems and techniques are provided for monitoring and managing the performance of services accessed by sites on a computing platform. When a performance issue is identified, a service is monitored to determine if calls to the service exceed a threshold completion time. If so, a resource available to call the service is adaptively throttled by the platform.

Abstract: A method, implemented by a computer system comprising a trusted execution environment (TEE) and a rich execution environment (REE) includes creating, by the TEE, a plurality of sub-threads preparing to implement sub-functions of a trusted application (TA), for each sub-thread, triggering, by the TEE, the REE to generate a shadow thread, where running of the shadow thread will cause a core on which the shadow thread runs to enter the TEE, and scheduling the created sub-thread to the entered core for execution.