Abstract: Embodiments of the present disclosure provide a scheduling method and device for a cloud native orchestration and scheduling system and a storage medium. A target judgment rule plugin is determined from multiple preset judgment rule plugins included in a preset judgment rule plugin set and loaded into a scheduler of the cloud native orchestration and scheduling system; whether any instance on any node in the cloud native orchestration and scheduling system is allowed to be preempted by a to-be-deployed instance is sequentially judged by using the target judgment rule plugin for the to-be-deployed instance; a candidate to-be-preempted instance on a candidate to-be-preempted node is determined; candidate to-be-preempted nodes are sorted by using a preset comparison plugin in the scheduler to determine a target to-be-preempted node, and the to-be-deployed instance is controlled to preempt the candidate to-be-preempted instance on the target to-be-preempted node.

Abstract: The present disclosure provides a cluster-based scheduling method, a medium and an electronic device, the method includes: in response to a current scheduling, obtaining a historical scheduling snapshot generated in a previous scheduling and reading a global timing value corresponding to a cluster during the previous scheduling from the historical scheduling snapshot, where the global timing value is a latest change timing of a node in the cluster during the previous scheduling recorded in chronological order; reading a respective timing value corresponding to each node in the cluster during the current scheduling from a scheduling cache and selecting, using the global timing value as a reference, a target node whose timing value is later than the global timing value in chronological order; and updating the historical scheduling snapshot based on node information of the target node to obtain a scheduling snapshot of the current scheduling.

Abstract: A key-value cache management system processes inference requests of a generative model. An inference request requests a starting virtual memory address and a number of layers in the generative model. In response to receiving an inference request, contiguous virtual memory space is reserved for the key-value cache in accordance with the inference request by assigning the starting virtual memory address in the key-value cache and calculating memory pointers for each layer in the model and each block so that the one or more blocks are written sequentially. The generated memory pointers are outputted to the generative model so that each self-attention layer writes computed key-value pairs at physical addresses in the key-value cache specified by the memory pointers.

Abstract: An example method to manage a virtual machine deployment in a cloud environment includes generating a blueprint comprising a blueprint component corresponding to an application storage policy for all endpoints associated with an infrastructure source in the cloud environment, in response to a determination that the application storage policy comprises a first storage policy of a first endpoint among the endpoints, collecting the first storage policy to deploy the virtual machine on the first endpoint according to the first storage policy, or in response to a determination that the application storage policy does not comprise any storage policy of the first endpoint among the endpoints, creating a storage policy of the first endpoint and collecting the created storage policy to deploy the virtual machine on the first endpoint according to the created storage policy.

Abstract: A policy-driven method of migrating a virtual computing resource that is executing an application workload includes the steps of determining that at least one of multiple policies of the application has been violated by the virtual computing resource while executing the workload in a first virtual data center, and responsive to said determining, programmatically performing: (1) searching for a virtual data center to which the virtual computing resource can be migrated, (2) determining that the virtual computing resource will be able to comply with all of the policies of the application while executing the workload if the virtual computing resource is migrated to the second virtual data center, and (3) based on determining the ability to comply, migrating the virtual computing resource across clouds, namely from the first virtual data center to the second virtual data center.

Abstract: An example method to manage a virtual machine deployment in a cloud environment includes generating a blueprint comprising a blueprint component corresponding to an application storage policy for all endpoints associated with an infrastructure source in the cloud environment, in response to a determination that the application storage policy comprises a first storage policy of a first endpoint among the endpoints, collecting the first storage policy to deploy the virtual machine on the first endpoint according to the first storage policy, or in response to a determination that the application storage policy does not comprise any storage policy of the first endpoint among the endpoints, creating a storage policy of the first endpoint and collecting the created storage policy to deploy the virtual machine on the first endpoint according to the created storage policy.

Abstract: A policy-driven method of migrating a virtual computing resource that is executing an application workload includes the steps of determining that at least one of multiple policies of the application has been violated by the virtual computing resource while executing the workload in a first virtual data center, and responsive to said determining, programmatically performing: (1) searching for a virtual data center to which the virtual computing resource can be migrated, (2) determining that the virtual computing resource will be able to comply with all of the policies of the application while executing the workload if the virtual computing resource is migrated to the second virtual data center, and (3) based on determining the ability to comply, migrating the virtual computing resource across clouds, namely from the first virtual data center to the second virtual data center.

Abstract: A scheduler selects an I/O from a session of a pool and updates token buckets associated with resource limits and reserves for the session and the pool and statistics used in determining fair sharing. To select an I/O, the scheduler identifies sessions with a non-empty queue, identifies head I/Os in the queues, computes for the head I/O a deadline using session and pool reserve buckets and a release time using session and pool limit buckets, and selects a head I/O with an earliest deadline that is past the release time. If the deadline of the selected candidate head I/O is in the past, the scheduler transfers the selected head I/O to the tail of the storage device queue. Otherwise, the scheduler selects the pool with the least amount of I/O traffic according to a session fair share estimator.

Abstract: A scheduler selects an I/O from a session of a pool and updates token buckets associated with resource limits and reserves for the session and the pool and statistics used in determining fair sharing. To select an I/O, the scheduler identifies sessions with a non-empty queue, identifies head I/Os in the queues, computes for the head I/O a deadline using session and pool reserve buckets and a release time using session and pool limit buckets, and selects a head I/O with an earliest deadline that is past the release time. If the deadline of the selected candidate head I/O is in the past, the scheduler transfers the selected head I/O to the tail of the storage device queue. Otherwise, the scheduler selects the pool with the least amount of I/O traffic according to a session fair share estimator.