The present invention provides an image forming apparatus comprising: at least one first module and at least one second module configured to execute functions corresponding respectively to the at least one first module; a first control unit configured to notify, to a corresponding second module, a request accepted by the at least one first module; and a second control unit configured to control in accordance with the notification from the first control unit, the corresponding second module, wherein the at least one first module is activated at all times from when the image forming apparatus is activated, and the at least one second module is generated as a container of an execution environment that is independent of an operating system and whose activation state is controlled by an instruction from the second control unit.

Resource allocation problems involve identification of resource, selection by certain criteria and offering of resources to the requester. Identification of required resources may involve matching the type of resource, selecting based on user requirements and policy criteria, and offering the resource through an assignment system. An apparatus and a method are provided that enable identification and selection of resources. The method includes receiving a resource allocation request for the allocation of a resource, the resource allocation request specifying a set of user requirements. The method includes receiving an operator policy associated with the resource, the operator policy including one or more policy requirements. The method includes synthesizing a resource request based on the resource allocation request and the operator policy. Synthesizing the resource request based on the resource allocation request and the operator policy comprises combining the user requirements with the one or more of the policy requirements.

A host processing device (“host”) instructs a plurality of data processing (DP) accelerators to configure themselves for secure communications. The host generates an adjacency table of each of the plurality of DP accelerators (“DPAs”). The host is communicatively coupled to the plurality of DPAs via a switch. The host transmits, to the switch, a list of the DPAs and instructs the switch to generate an adjacency table of the DPAs that includes a unique identifier of each DPAs and a communication port of the switch associated with the DPA. The host establishes a session key communication with each DPA and sends the DPA a list of other DPAs that the DPA is to establish a session key with, for secure communications between the DPAs. The DPA establishes a different session key for each pair of the plurality of DPAs. When all DPAs have established a session key for communication with other DPAs, the host can assign work tasks for performance by a plurality of DPAs, each communicating over a separately secured communication channel.

In some examples, a system generates a neural network comprising logical identifiers of compute resources. For executing the neural network, the system maps the logical identifiers to physical addresses of physical resources, and loads instructions of the neural network onto the physical resources, wherein the loading comprises converting the logical identifiers in the neural network to the physical addresses.

A system enabling a hypervisor to assign processor resources for specific tasks to be performed by a virtual machine. An example method may comprise: receiving, by a hypervisor running on a host computer system, a virtual processor (“vCPU”) assignment request from a virtual device driver running on a virtual machine managed by the hypervisor, assigning a vCPU for executing a task associated with the assignment request, and causing the virtual device driver to execute the task using the vCPU.

This disclosure describes techniques for improving allocation of computing resources to computation of machine learning tasks, including on massive computing systems hosting machine learning models. A method includes a computing system, based on a computational metric trend and/or a predicted computational metric of a past task model, allocating a computing resource for computing of a machine learning task by a current task model prior to runtime of the current task model; computing the machine learning task by executing a copy of the current task model; quantifying a computational metric of the copy of the current task model; determining a computational metric trend based on the computational metric; deriving a predicted computational metric of the copy of the current task model based on the computational metric; and, based on the computational metric trend, changing allocation of a computing resource for computing of the machine learning task by the current task model.

Operations include halting a runaway query in response to determining that a performance metric of the query exceeds a performance threshold. The runaway query halting system receives a query execution plan associated with a query and divides the received execution plan into one or more components. For each component, the system determines a predicted resource usage associated with executing the component. The system further determines a predicted resource usage associated with the query execution plan based on the predicted resource usage associated with each component. The system executes the query associated with the received query execution plan and compares the predicted resource usage associated with the query to a resource usage threshold. In response to determining that the predicted resource usage of the query execution plan exceeds the resource usage threshold, the system halts execution of the query associated with the query execution plan.

A cloud access method of an internet of things (IoT) device and devices performing the cloud access method are disclosed. The cloud access method using a cloud proxy function includes receiving a first resource retrieval request of a client device from a cloud, extracting, from the first resource retrieval request, a device identification (ID) of a device including a resource for which a resource retrieval is requested, and transmitting a second resource retrieval request of the client device to the device based on the device ID.

The present disclosure relates to a method performed by a process control node (210) configured to allocate resources shared by a plurality of tenant applications, wherein each tenant application comprises a selection of non real-time processes and real-time processes, the method comprising receiving a first resource request, from a tenant application, indicative of resources requested to be allocated, by the process control node, for one or more real-time processes of the tenant application, evaluating a scheduling test to determine if the set of processing resources can be allocated from the shared resources by determining if resources requested by the first resource request can be allocated, and if it is determined that the requested resources can be allocated from the shared resources, the method further comprises performing the steps starting the one or more real-time processes of the tenant application within a resource partition of the tenant application, calculating updated resource quotas and priorities for non real-time processes comprised by the tenant application, transmitting a first resource response to the tenant application.

A ratio of prediction liable to result in user requirement violation is reduced by adjusting results of resource design even if it is highly likely that the user requirement violation will incur a heavy penalty. There is provided a requirement specifying functional unit (11) that specifies a user requirement for a service of interest, and a resource design unit (12) that predicts, by machine learning, performance achievable at a plurality of resource settings in performing the service of interest and selects a resource setting that satisfies the specified user requirement, based on results of the prediction, wherein the resource design unit (12) generates a P model as a model for use to predict performance, the P model using a P-mode loss function obtained by adding a function to an N model that uses an existing N-mode loss function, the added function taking a finite value when actual performance is lower than predicted performance.