Abstract: A request is received over a network. The request is directed to a service provided by an application of a cluster infrastructure. The cluster infrastructure includes the application and a control plane to schedule execution of the application. Responsive to receiving the request, the control plane of the cluster infrastructure is transitioned from a stopped state to an active state. The request is transferred to the cluster infrastructure.

Abstract: It is determined that data stored on a storage device associated with a computing device powered by a power source is to be transmitted to a remote computing device. A power state of the power source is determined. An encoding mechanism is selected from a plurality of different encoding mechanisms based on the power state of the power source. The data is encoded based on the encoding mechanism to generate encoded data. The encoded data is transmitted to the remote computing device.

Abstract: Computer workloads can be managed across distributed computing clusters according to some aspects of the present disclosure. In one example, a system can receive a request from a workload manager for identifying a computing cluster to which to assign a workload. The system can determine that the workload is to be assigned to a particular computing cluster among a plurality of computing clusters based on historical information about replica deployment by the particular computing cluster. The system can then transmit a response to the workload manager for causing the workload manager to assign the workload to the particular computing cluster.

Abstract: A method includes generating an optimized version of an inference serverless function using a graph compiler. The method further includes replacing a storage volume in an init container of the inference serverless function with a new storage volume comprising the optimized version of the inference serverless function.

Abstract: Computer workloads can be assigned to nodes of a distributed computing environment based on energy consumption modes employed by the nodes. In one example, a system can determine a first tag assigned to a workload. The first tag can indicate a compatibility of the workload with one or more energy consumption modes employable by one or more nodes of a distributed computing environment. The system can also determine a second tag assigned to a node of the distributed computing environment. The second tag can indicate an energy consumption mode employed by the node. The system can then determine a correspondence between the first tag and the second tag indicating that the workload is compatible with the energy consumption mode employed by the node. Based on determining the correspondence between the first tag and the second tag, the system can assign the workload to the node.

Abstract: Tracing data including a plurality of traces for a plurality of operations performed by a distributed computing system on behalf of a plurality of users of a distributed computing system during a period of time is identified. Each trace having latencies for a plurality of segments of a corresponding operation. A set of overall latencies comprising an overall latency for each segment is determined. A set of user latencies including a latency for each segment is determined for each of the plurality of users. A set of energy usage estimates including an energy usage estimate for one or more of the plurality of users is generated, by a processing device, based on the set of overall latencies and the set of user latencies using a machine learning (ML) model.

Abstract: One or more parameters corresponding to a targeted energy characteristic of a computing client device are received. A ratio of types of storage devices allocated to the computing client device is adjusted based on the one or more parameters. A storage volume for the computing client device is provisioned according to the ratio of the types of the storage devices.

Abstract: Systems and methods are disclosed that deploy software code from a dataset into a computing environment. The systems and method collect energy metrics of the software code while executing in the computing environment. The systems and methods determine a sustainability label for the software code based on the energy metrics. The systems and methods assign the sustainability label to the software code to produce a sustainability-based dataset.

Abstract: Computer-implemented methods for managing instances of serverless functions in a cloud computing system are provided. Aspects include obtaining a service level objective for a serverless function, obtaining a command queue length for a graphical processing unit disposed on each of a plurality of compute nodes in the cloud computing system, and obtaining a request queue length of the serverless function. Aspects also include calculating a number of instances of the serverless function to deploy in the cloud computing system, wherein the number of instances is determined based on the service level objective and the request queue length of the serverless function, identifying compute nodes from the plurality of compute nodes to deploy each of the number of instances of the serverless function, and creating an instance of the serverless function on each of the identified compute nodes.

Abstract: A method includes detecting, by a webhook controller, an inference serverless function invocation. The method further includes determining that the inference serverless function can be optimized. The method further includes generating an optimized version of the inference serverless function using a graph compiler, in response to the determining. The method further includes replacing, by a processing device of the webhook controller, a storage volume in an init container of the inference serverless function with a new storage volume comprising the optimized version of the inference serverless function.

Abstract: Computer-implemented methods for identifying a configuration of a virtual machine and performance selecting an application for execution on the virtual machine are provided. Aspects include executing a plurality of calibration programs on a virtual machine having an unknown configuration and collecting a plurality of performance metrics from the virtual machine during execution of the plurality of calibration programs. Aspects also include inputting the plurality of metrics into a trained machine learning model, receiving, from the trained machine learning model, a predicted configuration of the virtual machine, and executing a version of an application on the virtual machine, wherein the version is determined based at least in part on the predicted configuration.

Abstract: Systems and methods are presented to provide a first machine learning model to a collaboration platform. The systems and methods receive a second machine learning model from the collaboration platform that indicates the second machine learning model is based on the first machine learning model. The systems and methods test the second machine learning model using criteria corresponding to the first machine learning model to determine whether the second machine learning model is valid. In turn, the systems and methods publish the second machine learning model to a repository in response to determining that the second machine learning model is valid.

Abstract: A cooperative transmission method based on adaptive terminal aggregation includes arranging a base station, several mobile terminals and several cooperative nodes, and further includes the following steps: S1, judging an aggregation form by a judgment node according to the quality of the first link and the quality of the second link; S2, based on a control signaling received from the base station, transmitting business data to the base station by a mobile terminal at the first-time point and within a specific resource; and S3, if the aggregation form is the cooperative node transmission, monitoring the business data transmitted via the mobile terminal at the first-time point in S2 by one or several cooperative nodes and decoding the business data. By dynamically adjusting the values of the first regulatory factor and the second regulatory factor, the adaptive terminal aggregation form with the lower energy-delay product for each user is selected.

Abstract: An example method for data auditing for object storage public clouds includes a service broker receiving a request to store data in public object storage, where the request includes user information or a container image. The service broker, based on either the user information or the container image, determines that data auditing is necessary. The service broker creates a storage unit, in public object storage, and a storage proxy. The method further includes the storage proxy storing data, and a data auditor retrieving data from the storage proxy. The data auditor determines a data qualification for the data, and notifies the storage proxy of the data qualification.

Abstract: Energy consumption associated with users of a distributed computing system can be monitored using tracing according to some examples described herein. In one such example, a system can execute tracing software configured to track usage of a software service and generate tracing information. The tracing information cant attribute the usage of the software service to a group of users. The system can analyze the tracing information to determine a portion of the usage to attribute to a given user. The system can then determine an energy consumption estimate for the user, for example based on the portion of the usage associated with the user and an energy consumption amount associated with the software service. A graphical user interface can be generated indicating the energy consumption estimate for the user.

Abstract: A system and method for a tiered cloud storage for different availability and performance requirements includes a gateway, a block store configured to cache data, and an object store configured to persistently store data. The gateway, the block store, and the object store are in a compute zone. The gateway may receive from a user application a file access call and process the file access call. The gateway may also send the file access call to the block store. Then, the gateway may determine to store data in the object store and flush the data from the block store to the object store.

Abstract: A method includes initiating, by a device manager associated with a cluster manager proxy, a connection with a cluster of computing devices, wherein initiating the connection includes providing first credentials to the cluster of computing devices to access the cluster manager proxy. The method further includes receiving, at the cluster manager proxy, a first request to register the cluster of computing devices with a cluster manager, the first request including the first credentials to access the cluster manager proxy and sending, from the cluster manager proxy to the cluster manager, a second request to register the cluster of computing devices with the cluster manager, the second request including second credentials to access the cluster manager.

Abstract: Computer-implemented methods for deploying workloads in a cloud computing system based on energy efficiency are provided. Aspects include obtaining an energy efficiency metric for a plurality of compute nodes in the cloud computing environment and classifying the plurality of compute nodes into energy efficiency groups based on the energy efficiency metrics. Aspects also include creating a partition of nodes including one compute node selected from each of the energy efficiency groups, deploying a replica of a workload to each compute node in the partition, and monitoring an energy consumption and a computing performance of each of compute node in the partition during a probing period. Aspects further include identifying, based on the energy consumption and performance, a selected energy efficiency group from the plurality of energy efficiency groups and deploying the workload to one or more of the plurality of compute nodes in the selected energy efficiency group.

Abstract: Computer-implemented methods for adaptively training a machine learning model for estimating energy consumption in a cloud computing system are provided. Aspects include detecting an update event in the cloud computing system and collecting energy consumption data for the cloud computing system for a time period after the occurrence of the update event. Aspects also include retraining a power consumption model based at least in part on the data collected during the time period and storing the power consumption model in the cloud computing system.

Abstract: Energy consumption associated with users of a distributed computing system can be monitored using tracing according to some examples described herein. In one such example, a system can execute tracing software configured to track usage of a software service and generate tracing information. The tracing information cant attribute the usage of the software service to a group of users. The system can analyze the tracing information to determine a portion of the usage to attribute to a given user. The system can then determine an energy consumption estimate for the user, for example based on the portion of the usage associated with the user and an energy consumption amount associated with the software service. A graphical user interface can be generated indicating the energy consumption estimate for the user.