Abstract: Some embodiments provide a method for operating a physical server in a network. The method stores multiple copies of a virtual machine (VM) image at a network-accessible storage. The method uses a first copy of the VM image as a virtual disk to execute a VM on a hypervisor of a first physical computing device. The method uses a second copy of the VM image as a virtual disk accessible via a smart network interface controller (NIC) of a second physical computing device to execute an operating system of the second physical computing device.

Abstract: A data protection system includes a splitter configured to reduce latencies when splitting writes in a computing environment. The splitter captures a write and adds metadata to augment the write with virtual related information. The augmented data is provided to a smartNIC while the write is then processed in the IO stack. The smartNIC may have a volume only visible to the splitter. The smartNIC also includes processing power that allows data protection operations to be performed at the smartNIC rather than with the processing resources of the host.

Abstract: According to an aspect, a computer-implemented method includes intercepting a call to invoke execution of a service as part of performing a synchronous transaction. A current state of the synchronous transaction is captured and persisted in a transaction object corresponding to the synchronous transaction. Execution of the service is invoked.

Abstract: An apparatus includes a CPU core, a first cache subsystem coupled to the CPU core, and a second memory coupled to the cache subsystem. The first cache subsystem includes a configuration register, a first memory, and a controller. The controller is configured to: receive a request directed to an address in the second memory and, in response to the configuration register having a first value, operate in a non-caching mode. In the non-caching mode, the controller is configured to provide the request to the second memory without caching data returned by the request in the first memory. In response to the configuration register having a second value, the controller is configured to operate in a caching mode. In the caching mode the controller is configured to provide the request to the second memory and cache data returned by the request in the first memory.

Abstract: An artificial intelligence accelerator includes an external command dispatcher, a first data access unit, a second data access unit, a global buffer, an internal command dispatcher, and a data/command switch. The external command dispatcher receives an address and access information. The external command dispatcher sends the access information to one of the first data access unit and the second data access unit, the first data access unit receives first data from a storage device according to the access information, and sends the first data to the global buffer. The second data access unit receives second data from the storage device according to the access information, and sends the second data. The data/command switch receives the address and the second data from the second data access unit, and sends the second data to one of the global buffer and the internal command dispatcher.

Abstract: A method, an apparatus, an electronic device, and a storage medium for application launch. The method includes: reading a dependency relationship of each application in a plurality of applications from a dependency description file carried by each application; generating an application launch sequence chain according to the dependency relationship; and sequentially launching the plurality of applications according to the application launch sequence chain.

Abstract: A task dispatching method and apparatus, a computing and processing device, a computer program and a computer-readable medium. The task dispatching method includes: acquiring a plurality of examination tasks to be dispatched and one or more examining apparatuses configured for processing the plurality of examination tasks, wherein the examination tasks have different task identifications; according to the examination tasks and the examining apparatus, generating a genetic sequence, the genetic sequence includes a correspondence relation between the examination tasks and the examining apparatuses processing the examination tasks, and genes in the genetic sequence are the task identifications of the examination tasks; and performing iterative updating to the genetic sequence, till an iteration stopping condition is satisfied, and according to the genetic sequence that has been updated, dispatching the examination tasks.

Abstract: Techniques are disclosed herein for a direct launch for asynchronous actions. An example method can include receiving, by a device, a first workflow request, wherein the device comprises a first launcher in operable communication with a second launcher via a shared queue, and wherein a first duplicate of the first workflow request is stored in the shared queue is to be retrieved by the second launcher upon an elapsing of a threshold time interval. The device can transmit the workflow request to an application for executing the first workflow request, wherein the transmission is prior to the threshold time elapsing. The device can determine whether the application initiated a workflow to process the workflow request. The device can update the first duplicate of the first workflow request stored at the shared queue based at least in part on the determination and prior to the threshold time interval elapsing.

Abstract: A method for identifying a second device by a first device for establishing a communication between the first device and the second device is described here. The method includes receiving, by a processor of a first device, a voice command from a worker in a workplace. In an example, the method comprises pausing, by the processor, a workflow operation executing on the first device. The method further comprises performing, by the processor, a voice recognition to analyze the voice command of the worker. The method includes activating, by the processor, a communication module of the first device based on the voice recognition, to identify a second device in proximity to the first device. The method includes terminating, by the processor, a connection between the first device and the wearable electronic device. Thus, terminating, by the processor, a second connection of the first device with the second device.

Abstract: A neural processing device and a method for managing tasks thereof are provided. The neural processing device includes a neural core configured to perform a task and generate a completion signal for completion of the task, a core global configured to transfer task information for the task to the neural core and receive the completion signal of the task from the neural core, and a task manager configured to generate and transmit the task information to the core global, receive the completion signal from the core global, generate a completion report, and transmit the completion report.

Abstract: A task manager, a neural processing device, and a method for checking task dependencies thereof are provided. The task manager includes a task buffer configured to receive first and second tasks of different first and second types, a first queue configured to receive a first task descriptor for the first task from the task buffer, a second queue configured to receive a second task descriptor for the second task from the task buffer, a dependency checker configured to check dependencies of the first and second task descriptors, a third queue configured to receive the first task descriptor from the dependency checker, and a fourth queue configured to receive the second task descriptor from the dependency checker.

Abstract: This application discloses a task execution method including: determining a plurality of deep learning tasks to be concurrently executed and an model for implementing each deep learning task; obtaining an execution policy of each deep learning task, where the execution policy indicates a scheduling mode and a used model variant of the deep learning task, and the model variant of the deep learning task is obtained according to the artificial intelligence model for implementing the deep learning task; and executing a corresponding deep learning task according to the execution policy of each deep learning task. In this application, execution performance of a deep learning task can be improved in terms of a scheduling mode of the deep learning task, and can also be improved in terms of a model for implementing the deep learning task, to effectively improve the execution performance of the deep learning task.

Abstract: The present disclosure is directed to deployment of applications to mobile devices such as, for example, payment terminals. In some embodiments, a method includes uploading a mobile application that is for execution on a group of mobile devices; preparing the mobile application for deployment; receiving a first request with a deploy plan that specifies parameters and a subset of mobile devices for deploying the mobile application; and deploying the mobile application to the subset of mobile devices according to the deploy plan.

Abstract: A resource scheduling method, applied to the field of high-performance computing, including: obtaining a binding relationship between a process group and a processor, where the binding relationship indicates a binding relationship between at least one slave thread of a first process group in at least one process group and a plurality of processor cores in the processor, and the first process group is any process group in the at least one process group; then monitoring a working state of a first processor core, where the first processor core is any processor core in the plurality of processor cores; and when the first processor core is in an idle state, scheduling the first processor core to run a first slave thread based on the binding relationship, where the first slave thread is any slave thread that has a binding relationship with the first processor core in the first process group.

Abstract: Described herein is a network fabric including a plurality of graphical processing unit (GPU) clusters. The plurality of GPU clusters includes at least a first GPU cluster operating at a first speed and a second GPU cluster operating at a second speed that is different than the first speed. The network fabric includes a plurality of blocks, wherein each block includes: (a) one or more racks that host a GPU cluster, and (b) a plurality of switches arranged in a hierarchical structure that communicatively couple the block to other blocks included in the network fabric. Responsive to receiving a request to execute a workload, allocating one or more GPUs from the plurality of GPU clusters to execute the workload.

Abstract: A computational storage unit is described. The computational storage unit may include a first resource of a first type and a second resource of the first type. A table may map a user identifier (UID) for a user to a number of resources of the first type.

Abstract: An approach for managing workload deployment, specifically, restarting workloads that are interdependent is provided. The approach includes a mechanism by which a group of interdependent workloads in a cloud infrastructure environment can be restarted sequentially. Interdependent workloads simply mean workloads that depend on each other in some way such that if one goes down, then the cloud infrastructure environment can experience downtime. The workloads can be of the same or different types. Additionally, this mechanism can be scaled such that groups of workloads can be restarted simultaneously. Furthermore, the approach can accommodate cloud infrastructure environments with multiple types of workloads, and it allows for groups of workloads to be restarted simultaneously while also allowing the workloads within each group to be restarted sequentially.

Abstract: The technology is directed to systems and methods for adjusting computing resource quotas based on metrics associated with or corresponding to a cloud environment. The metrics corresponding to computing resources used by a cloud environment may be received. The received metrics may be input into a machine learning model, which may generate, based on the received metrics, one or more predicted quotas for the computing resources used by the cloud environment. The computing resource quotas may be adjusted based on the predicted quotas generated by the machine learning model.

Abstract: A method, computer program product, and computer system are described for providing decision instructions for problem incidents in information technology (IT) systems. The method includes categorizing incident instances as incident types, where an incident instance includes one or more events having resources identifying one or more hosts specific to the incident instance and wherein an incident type categorizes incident instances having a common underlying problem with the incident type having generalized resources. The method includes generating decision instructions for an incident type wherein the decision instructions provide generalized command steps for interacting with hosts for the resolution of incident instances of the incident type.

Abstract: In some implementations, a method including receiving, at a quota management application, resource usage information for at least a first cloud platform of a plurality of cloud platforms of a multi-cloud system; analyzing, at the quota management application, the resource usage information to determine the resource usage information of a first resource at the first cloud platform exceeds a first quota threshold for the first resource; in response to the first quota threshold being exceeded, sending, by the quota management application, a request to a workflow processor to request an increase in a quota for the first resource; and in response to receiving a response to the request indicative of a grant of the increase, sending a resource request towards the first cloud platform, wherein the resource request indicates the increase in the quota for the first resource. Related systems, methods, and articles of manufacture are also disclosed.

Abstract: Descriptors of machine learning tasks to be used to respond to analysis requests, indicating acceptable categories of runtime environments for the tasks and metrics to be collected from the tasks, are received via programmatic interfaces. In response to an analysis request, an orchestrator receives results from individual tasks as they become available, provides the results to other tasks, and causes a response to the request to be prepared using results from at least a subset of the tasks. Metrics collected from the tasks, and a visual representation of the tasks indicating their runtime environments are presented.

Abstract: In an example embodiment, rather than use a traditional database or datastore, or an ETCD database, to store the state of a cloud infrastructure management and orchestration system, a peer-to-peer network is used. The complete state of the cluster is available at every node. Nodes that are newly added to the system will fetch the current state from existing nodes. If the verification of the block is successfully performed by at least a quorum of the nodes, then the block is persisted as a new state on a node, and then distributed as a new state to every other node. The verification ensures that errors or possibly compromised nodes cannot persist incorrect changes to the system state.

Abstract: A method, computer program product, and computer system are provided for container cross-cluster capacity scaling. The method includes broadcasting local capacity information of capacity availability or capacity requirement for the local cluster and receiving broadcasts from each of one or more other clusters providing capacity information including capacity availability or capacity requirements. The method may map the received capacity information with the local capacity information and may determine a suitable cross-cluster capacity sharing when a capacity requirement of the local cluster maps to a capacity availability of another cluster or when a capacity availability of the local cluster maps to a capacity requirement of another cluster. The method may coordinate the deallocation of a node from the cluster having the capacity availability and reallocation of the node to the cluster having the capacity requirement.

Abstract: A system and a method for allocating limited resources to agents that reveal their stochastic demands on arrival over a finite horizon in a proportionally fair manner that exhausts available resource budgets are provided. The method includes: receiving, from an entity, a request for a resource for which a predetermined maximum amount of the resource is available for allocation; estimating a number of future requests and amounts of the resource to be requested within a predetermined time interval; adjusting a result of the estimating by calculating a standard deviation of an uncertainty of at least one future request; and determining, based on the estimates and the adjustment, an amount of the resource to allocated to the entity by applying an algorithm with respect to the adjusted estimates and the predetermined maximum available amount of the resources.

Abstract: The present disclosure relates to systems, methods, and computer-readable media for deploying cloud-native services across a plurality of cloud-computing platforms. For example, systems disclosed herein identify resource identifiers associated with cloud-computing services (e.g., types of services) to be deployed on one or more resources capable of executing or otherwise providing cloud-native services. The systems disclosed herein further generate resource bindings including deployment specifications that include data for deploying cloud-native services on corresponding platform resources (e.g., cloud resources, edge resources). Using the resource bindings, the systems disclosed herein can deploy cloud-native services across multiple platforms via control planes configured to manage operation of resources on the different platforms.

Abstract: Apparatuses, systems, and techniques to determine a configuration based at least in part on data stored by at least one data structure of a workload at runtime, and transform the workload into a sparse workload based at least in part on the configuration. In at least one embodiment, one or more sparse workloads (e.g., one or more sparse neural networks) are generated based at least in part on, for example, one or more workloads (e.g., one or more neural networks).

Abstract: Systems and methods for quality of service management are provided. According to one embodiment, a non-transitory computer-readable medium comprises instructions that when executed by the processing resource cause the processing resource to generate an inertial parameter that represents a resistance to moving a first volume from a first protection domain in a distributed storage system and rebalance, when the inertial parameter indicates the first volume should not be removed from the protection domain, a compute load between one or more available protection domains having capacity for additional volumes. The compute load includes one of a plurality of volumes other than the volume.

Abstract: A plurality of GPU clusters are communicatively coupled with one another via a plurality of network devices arranged in a hierarchical structure, wherein the GPU clusters includes at least a first GPU cluster operating at a first speed and a second GPU cluster operating at a second speed that is different than the first speed. A routing policy is configured for each network device, wherein the configuring includes establishing a mapping of each incoming port-link of the network device to a unique outgoing port-link of the network device. For a packet transmitted by a GPU of a host machine and received by a first network device, an incoming port-link of the first network device is determined on which the packet was received and based on the configuring, an outgoing port-link is identified that corresponds to the incoming port-link. The packet is forwarded on the outgoing port-link of the network device.

Abstract: Systems, methods, and devices for offloading network data to a datastore. A system includes a plurality of instances of a datastore node in a single networking device, each of the plurality of instances of the datastore node comprising: a datastore; a publisher independent of the datastore for publishing data to the datastore; a subscriber independent of the datastore for receiving information from the datastore; a replicator agent configured to connect to the datastore as a publisher or a subscriber; and a persistent storage agent configured to connect to the datastore as a publisher or a subscriber.

Abstract: Systems and methods for secure, collaborative, digital clipboards are provided. A method may include allocating one or more clipboard seats that are mapped to one or more of a plurality of user devices. The user devices may each be accessing a shared digital application in a collaborative digital workspace that is configured to cause an action executed to the shared digital application on one of the user devices to generalize for all the user devices. The method may include launching a shared digital clipboard that may be configured to be used by the plurality of user devices to copy and paste information in the collaborative digital workspace. The shared digital clipboard may also be configured to maintain a private copy-and-paste environment for each clipboard seat, wherein information copied to the clipboard by an entity mapped to a clipboard seat is inaccessible to any entity that is unmapped to that clipboard seat.

Abstract: One method includes receiving a user-defined function, generating a binary based on the user-defined function, copying the binary to a target host where the user-defined function will run, in response to a lookup request, received from an API gateway based on a function invocation by a user, looking up the target host and a route, and returning, to the API gateway, a target host ID and the route, so as to enable the API gateway to request the target host to respond to the function invocation.

Abstract: Apparatuses, systems, and techniques that allow a parallel processing unit to launch at least one graph including one or more operations and to perform the operation(s). In at least one embodiment, a graphics processing unit launches a graph that is uploaded from a central processing unit.

Abstract: A method and system for providing outbox services separate from a source application in a distributed application system are disclosed. The method includes performing, by the source application, an operation; writing, onto an operation store, data associated with the operation performed by the source application; writing, onto an event store, an event corresponding to the data; receiving, by an outbox transporter and from the event store, the event; and publishing, to a target and by the outbox transporter, the event received by the outbox transport processor, in which the outbox transporter is separate from the source application.

Abstract: A computer network device (such as an access point, a switch or a router) that has multiple user interfaces is described. During operation, the computer network device may execute program instructions for the user interfaces and a shared messaging module, where a given user interface includes an agent corresponding to an application. When a message associated with the application is received via a user interface in the user interfaces, the corresponding agent in the user interface may extract a command or operation from the message. Then, the shared messaging module may translate the command or operation into a common format of the application. Moreover, the shared messaging module may provide (or route) the translated command or operation addressed to the application via a single communication path associated with the application and the agents for the application in the user interfaces.

Abstract: An electronic device fetches first information from memory devices concerning errors associated with radiation effects in the memory devices and fetches second information about system performance associated with the electronic device. The electronic device monitors current parameters of the memory devices and calculates parameter values for configuration of the memory devices based on the fetched first information, the fetched second information, and the monitored current parameters, the calculating performed to adjust the parameter values to improve a metric of the system performance against the errors associated with the radiation. The electronic device implements the calculated parameter values for the configuration of the memory devices.

Abstract: An extension device is positioned within a point-to-point link to connect two devices, where the extension device includes error detection circuitry to detect a set of errors at the extension device. The extension device further includes memory to store an event register, where the extension device is to write data to the event register to describe detection of an error by the error detection circuitry. The extension device further includes a transmitter to transmit a notification signal to indicate the detection of the error and presence of the data in the event register associated with the error.

Abstract: A communication system and an operation method thereof are provided. A transmitting device transmits a data unit to a receiving device through a data channel of a communication interface. The transmitting device calculates an original verification information unit of the data unit and synchronously transmits the original verification information unit to the receiving device through a verification information channel of the communication interface based on a transmission timing of the data unit in the data channel. After receiving a current data unit and before receiving a next data unit, the receiving device verifies whether the current data unit received from the data channel has errors in real time based on a current original verification information unit corresponding to the current data unit.

Abstract: Methods and systems are disclosed herein for using anomaly detection in timeseries data of user sentiment to detect incidents in computing systems and identify events within an enterprise. An anomaly detection system may receive social media messages that include a timestamp indicating when each message was published. The system may generate sentiment identifiers for the social media messages. The sentiment identifiers and timestamps associated with the social media messages may be used to generate a timeseries dataset for each type of sentiment identifier. The timeseries datasets may be input into an anomaly detection model to determine whether an anomaly has occurred. The system may retrieve textual data from the social media messages associated with the detected anomaly and may use the text to determine a computing system or event associated with the detected anomaly.

Abstract: A computing system may determine that a change has occurred in a software application or a document object model associated with a software application. The computing system may obtain a current copy of a document object model (DOM) associated with the application and a current screenshot of a screen displaying the application (e.g., in response to determining that the change in the DOM has occurred). If the current copy of the DOM is determined to be similar to the screenshot, the computing system may save storage space by storing the current copy of the DOM and discarding the screenshot. The computing system may store the current copy of the DOM in a memory buffer. The computing system may send the DOM copies or screenshots stored in the memory buffer to the monitoring system, for example, if an error in the application is detected.

Abstract: An apparatus is provided. The apparatus comprises interface circuitry, machine-readable instructions, and processing circuitry to execute the machine-readable instructions to receive a request from a virtual machine to execute a task, receive a service-level agreement, SLA, from the virtual machine indicating a desired feature of scanning a computing resource to execute the task for errors, and scan the computing resource for errors based on the SLA.

Abstract: The present disclosure provides a content addressable memory (CAM). The CAM includes a CAM register configured to store a mapping table, N comparators coupling to the CAM register, and N multiplexers coupling to the N comparators respectively and to the CAM register. The mapping table includes first addresses and second addresses. Each one of the first addresses corresponds to one of the second addresses. Each of the N comparators is configured to compare the first addresses with one of N input signals, where N is an integer greater than 1. The N input signals come from N microcontroller units. The N multiplexers is configured to generate N output signals. At least one of the N output signals is obtained according to the mapping table and a comparison output by one comparator of the N comparators.

Abstract: A read to a wordline can cause a read disturb error on neighboring wordlines. Instead of scanning the entire memory to identify wordlines that have a read disturb problem, a localized read scan approach can be used. In this approach, the memory is organized into several zones, where each zone contains several wordlines. The number of reads in each zone is tracked, and, after a certain number of reads, the data in the zone is read. If the error rate of the data exceeds a threshold, the data is relocated to another area of the memory.

Abstract: Various embodiments provide block failure protection for a memory sub-system that supports zones, such a memory sub-system that uses a RAIN (redundant array of independent NAND-type flash memory devices) technique for data error-correction. For some embodiments, non-parity zones of a memory sub-system that are filling up at a similar rate are matched together, a parity is generated for stored data from across the matching zones, and the generated parity is stored in a parity zone of the memory device.

Abstract: A system, method, and product for flexible RAID layouts in a storage system, including: determining a reliability of an individual storage device of a plurality of storage devices, the individual storage device containing a plurality of portions of a Redundant Array of Independent Disks (RAID) stripe in a storage system, wherein the RAID stripe includes user data and inter-device parity data; detecting a change in the reliability of the individual storage device that contains the portion corresponding to the RAID stripe; and changing an amount of intra-device protection corresponding to the RAID stripe by decreasing, in the RAID stripe, an amount of space used to store the inter-device protection data.

Abstract: A method comprising: receiving a request to read data stored in an array of drives; determining that the data is stored on one or more degraded drives in the array; reconstructing the data from one or more drives in the array other than the degraded drives; providing the reconstructed data in response to the request; and after providing the reconstructed data, writing the reconstructed data to one or more drives in the array other than the degraded drives.

Abstract: A service recovery method formulation device (information processing device) 16 includes a recovery method learning unit (learning unit) 165 that recognizes a pattern of each of pieces of recovery operation content for a plurality of recovery operations for services of an application program, groups the plurality of recovery operations for each pattern of the recovery operation content to form a plurality of recovery operation groups, recognizes a pattern of each of pieces of monitoring content for a plurality of pieces of monitoring data related to the services monitored immediately before and after the plurality of recovery operations are performed, respectively, and groups the plurality of pieces of monitoring data for each pattern of the monitoring content to form a plurality of monitoring data groups.

Abstract: In an aspect of the disclosure, a method, a computer-readable medium, and a BMC are provided. The BMC receives an update package containing one or more updated software components of a firmware image of the BMC to replace corresponding existing software components running on the BMC. The BMC shuts down the existing software components and executes the one or more updated software components. The BMC calculates one or more first stability factors of the one or more updated software components. The BMC compares the one or more first stability factors with one or more second stability factors of the corresponding existing software components. The BMC determines whether to keep the updated software components based on a comparison result.

Abstract: The description relates to enhancing computer performance, such as by decreasing latency associated with storage operations. One example can include recoverable processes. Each recoverable process can be configured to periodically write log records to individual storage partitions. Each log record includes a vector timestamp that describes the processes' dependencies on log records of other recoverable processes.

Abstract: An assigned subgroup that includes a plurality of entries is traversed by a prefetcher. It is determined that an expected number of entries associated with the assigned subgroup have been traversed. In response to determining that expected number of entries associated with the assigned subgroup have been traversed, it is determined that a last read entry associated with the assigned subgroup does not correspond to a last entry associated with the assigned subgroup.

Abstract: Techniques for data management are described. A monotonically increasing time source that indicates an elapsed time since a reference event may be activated. Multiple snapshots of a computing object may be generated in accordance with a schedule for backing up the computing object, where the schedule may include a retention duration for retaining snapshots. Based on generating the snapshots, timestamps for the snapshots may be stored, where the timestamps may indicate respective values of the monotonically increasing time source. As part of an expiration job, a reference value of the monotonically increasing time source may be identified based on the retention duration and a current value indicated by the monotonically increasing time source. Also, a snapshot of the snapshots may be expired based on a timestamp of the snapshot corresponding to a value of the monotonically increasing time source that is less than the reference value.