Abstract: Computer-implemented methods for dynamically generating instance types are presented. Aspects include receiving, by a controller, a user input describing a desired workload and a user intent for the desired workload. Aspects also include generating, by a machine learning model executing on the controller, an entity vector based the user input. Aspects also include accessing, by the controller, an instance type knowledge base comprising vector representations of one or more instance types. Aspects also include calculating, by the controller, a ranking between the entity vector and the vector representations of the one or more instance types in the instance type knowledge base. Aspects further include determining a set of instance types for the desired workload based on the ranking.

Abstract: In an approach for optimizing job preemption using a scheduling mechanism that considers dynamic and static information, a processor, responsive to a pending job being selected to be run, determines that no computing hosts within a computing cluster have available resources capable of running the pending job and non-preemptive measures will not provide required resources for the pending job. A processor identifies candidate jobs for preemption and dynamic information and static information for each candidate job. A processor ranks the candidate jobs for preemption based on the dynamic information and the static information for each candidate job. A processor attempts to preempt a top N candidate jobs whose released resources in combination with the available resources of the computing cluster satisfy the required resources of the pending job. A processor, responsive to successfully preempting the top N candidate jobs, initiates the pending job.

Abstract: Some embodiments of the present disclosure are directed to systems, computer-readable media, and computer-implemented methods for dynamic computing resource management. Some embodiments are directed to identifying a computing resource common between a first workload and a second workload, replacing a label associated with the first workload on the identified computing resource with a label associated with the second workload, executing the second workload using the identified computing resource. Other embodiments may be disclosed or claimed.

Abstract: Method and systems for online training management of reinforcement learning policy serving for cloud computing systems are discloses. An example method includes controlling a cloud computing system using a first reinforcement learning (RL) model; training the first RL model to generate a second RL model in response to one or more first criteria being satisfied; and controlling the cloud computing system using the second RL model in response to one or more second criteria being satisfied.

Abstract: A computer-implemented method for labeling and managing cloud computing resources includes receiving one or more computing jobs in a job queue and obtaining resource requirements for a first one of the one or more computing jobs. Nodes are placed into a cluster for the resource requirements from one or more cloud providers and the nodes are labelled to correspond to the first one of the one or more computing jobs. The first one of the one or more computing jobs from the job queue and is executed after the labelled aggregated resources are ready.

Abstract: A method and system allocating computing resources according to a trained machine learning model that includes receiving a request to accommodate a particular workload that involves a first threshold amount of computing resources implemented by at least one computing cluster. An identified a set of computing clusters, that have computing resources available for allocating from the identified set of computing clusters to the at least one computing cluster to satisfy the first threshold amount of computing resources, is received from a database. A trained deep-reinforcement learning model is applied to generate a policy for allocating the available computing resources from the identified set of computing clusters to the at least one computing cluster. It is verified that the generated policy satisfies a threshold according to one or more predetermined criteria. The computing resources are allocated from the set of computing clusters to the at least one computing cluster.

Abstract: Disclosed embodiments provide techniques for compute job allocation in a virtualized computing environment. A first list of compute jobs that are currently executing in a virtualized environment is obtained. For each job in the first list, a job description file is obtained. An entity extraction process is performed on the job description file to extract a plurality of job entities. Multiple clusters are created that correspond to the compute jobs in the first list. A second list of compute jobs that are currently queued for execution is obtained. Compute jobs in the second list are assigned to a cluster from the plurality of clusters, and the virtualized environment is reused for execution of a compute job from the second list based on the assigned cluster.

Abstract: An embodiment for improved methods for energy efficient scaling of multi-zone container clusters is provided. The embodiment may establish a connection between an upper layer container orchestration controller associated with multiple container cluster zones and lower layer resource manager controllers corresponding to multiple datacenters. The embodiment may determine additional workers are needed to perform a task and request worker offers from the lower layer resource manager controllers. The embodiment may receive the worker offers including worker profile data at the upper layer container orchestration controller.

Abstract: Mechanisms are provided for improving performance of container images. Container image chunks are generated from a container image file and input into one or more trained machine learning (ML) computer models, trained to classify container image chunks with regard to a plurality of container image performance characteristic classifications. For each container image chunk it is determined whether the a corresponding classification is negative, and in response to the classification being negative, an entry in a knowledge base having patterns of content matching content in the container image chunk is identified to determine one or more reasons for modification of the chunk specified in the entry. A notification output is generated specifying the container image chunks, their corresponding container image performance characteristic classifications, and the reasons for modification of the chunks.

Abstract: A computer-implemented method, a computer system and a computer program product boost efficiency through energy-aware workload placement. The method includes obtaining an energy profile for a plurality of computer servers and power consumption data for each computer server in the plurality of computer servers. The method also includes determining an optimal temperature for each computer server in the plurality of computer servers based on the energy profile. The method further includes determining a target processor utilization for each computer server in the plurality of computer servers based on the optimal temperature. In addition, the method includes calculating an efficiency rank for each computer server in the plurality of computer servers based on the target processor utilization and the power consumption data. Lastly, the method includes deploying a workload on a computer server with a highest efficiency rank.

Abstract: Systems, computer-implemented methods, and computer program products to facilitate optimization of resource usage based on quota trees are provided. According to an embodiment, a system can comprise a processor that executes computer executable components stored in memory. The computer executable components comprise a determination component that determines one or more quota trees that classify a job request as inadmissible. The computer executable components further comprise an optimization component that optimizes resource usage to enable admissibility of the job request based on the one or more quota trees.

Abstract: Systems, computer-implemented methods, and computer program products to facilitate using multiple quota trees in resource scheduling are provided. According to an embodiment, a system can comprise a processor that executes computer executable components stored in memory. The computer executable components comprise an evaluation component that executes admissibility of a job request based on a scope property of one or more quota trees that apply to the job request.

Abstract: A processor may generate one or more encrypted policies associated with a policy creator. A processor may generate token metadata associated with a user utilizing the one or more encrypted policies. A processor may encrypt the token metadata to form encrypted token metadata. A processor may send the one or more encrypted policies and the encrypted token metadata to a policy evaluator. The policy evaluator may evaluate the one or more encrypted policies and the encrypted token metadata. The processor may return a response. The response may be based on the evaluation by the policy evaluator.

Abstract: A shift-left analysis system receives information regarding an application implemented by one or more microservices. The system determines a microservice performance metric based on a performance prediction model for each microservice of the application. The system outputs an application performance metric for the application based on the microservice performance metrics determined for the one or more microservices of application.

Abstract: One or more systems, computer-implemented methods, and computer program products to facilitate a process for consensus regarding proceeding with a request of a transaction are provided. A system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a services component that operates a service of a transaction, where the services component approaches consensus regarding a request of the transaction by communicating with one or more other services components that operate one or more other services immediately upstream or downstream in the transaction to the service. The services component can approach the consensus by communicating only with the one or more other services components. To approach the consensus, the services component can communicate one or more messages that include one or more requests, votes or final decisions.

Abstract: Systems, computer-implemented methods, and computer program products that can facilitate performance biased resource scheduling based on runtime performance of a certain workload type on one or more nodes are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a performance component that assigns performance points to different nodes based on execution of one or more workload types. The computer executable components can further comprise a scheduler extender component that modifies a scheduling decision to run a workload type on a node based on the performance points.

Abstract: Techniques regarding anomaly forecasting are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory, and that can execute the computer executable components stored in the memory. The computer executable components can comprise a forecast component that can determine a probability of a computer application executing an anomaly state based on a probabilistic graph that is incrementally updated while the computer application is running.

Abstract: Methods, computer program products, and/or systems are provided that perform the following operations: obtaining data indicative of a node failure; obtaining data associated with nodes and pods started on each node; generating a causation score for each pod associated with a failed node, wherein each pod associated with the failed node is designated as a candidate pod for the node failure; determining pod rescheduling for each candidate pod associated with the failed node based, at least in part, on a pod ranking of the causation score for each pod; and providing the pod rescheduling to a node cluster to restart each pod associated with the failed node.

Abstract: A processor may generate one or more encrypted policies associated with a policy creator. A processor may generate token metadata associated with a user utilizing the one or more encrypted policies. A processor may encrypt the token metadata to form encrypted token metadata. A processor may send the one or more encrypted policies and the encrypted token metadata to a policy evaluator. The policy evaluator may evaluate the one or more encrypted policies and the encrypted token metadata. The processor may return a response. The response may be based on the evaluation by the policy evaluator.

Abstract: Methods, computer program products, and/or systems are provided that perform the following operations: obtaining data indicative of a node failure; obtaining data associated with nodes and pods started on each node; generating a causation score for each pod associated with a failed node, wherein each pod associated with the failed node is designated as a candidate pod for the node failure; determining pod rescheduling for each candidate pod associated with the failed node based, at least in part, on a pod ranking of the causation score for each pod; and providing the pod rescheduling to a node cluster to restart each pod associated with the failed node.