Abstract: One example method includes defining experiences for a workload that are to be analyzed at a first machine-learning (ML) model. The experiences define an association between the workload and microservices having computing resources that execute the workload. A probability of using each of the microservices of the experiences to execute the workload is generated at a second ML mode. A determination is made of which of the experiences have a probability that indicates that the experience will generate a low reward when analyzed by the first ML model. The experiences that generate the low reward are removed from the experiences to be analyzed at the first ML model. The experiences that have not been removed are analyzed at the first ML model to determine which experience includes microservices that should be used to execute the workload.

Abstract: Multi-agent reinforcement learning-based workload placement is disclosed. A placement engine is configured to use the state of a system and actual rewards to generate expected rewards that correspond to actions. Agents can take actions for corresponding workloads based on the expected rewards output by the placement engine. This allows workloads to be placed in a manner that conservers power relative to load placement policies while helping avoid service level agreement violations.

Abstract: Techniques are provided for allocating resources for one or more workloads. One method comprises obtaining a current performance of a workload; determining an adjustment to a current allocation of a resource allocated to the workload by evaluating a representation of a relationship between: (i) the current allocation of the resource allocated to the workload, (ii) a performance metric, and (iii) the current performance of the workload; and initiating an application of the determined adjustment to the current allocation of the resource for the workload. The performance metric may comprise a nominal value of a predefined service metric and the current performance of the workload may comprise a current value of a variable that tracks a given predefined service metric of the workload. An amount (or percentage) of the adjustment permitted for each iteration may be controlled. A sum of allocated resources can be constrained to an amount of available resources.

Abstract: A header matrix prepended to a machine learning model allows the relative importance of a dataset's features to be determined or inferred. The header matrix begins as an Identity matrix. Gradients associated with a backpropagation are stored in the header matrix and accumulated in an accumulation matrix. The relative importance of each feature of the dataset can be determined or inferred from the accumulation matrix.

Abstract: One example method includes determining a respective metric value for each computing resource in a group of computing infrastructures, and each metric value comprises a length of a Euclidean projection of a required resources vector in a respective vector of available resources provided by one of the computing infrastructures. Next, a list is created that includes the computing infrastructures, and the list is sorted according to one of two criteria. A restricted candidates list is created that is a subset of the list, and one of the computing infrastructures is randomly selected from the restricted candidates list. Finally, the method includes executing, with whichever computing infrastructure was randomly selected, a workload associated with the required resources vector.

Abstract: Techniques are disclosed for machine learning model management using edge concept drift duration prediction. For example, a system can include at least one processing device including a processor coupled to a memory, the at least one processing device being configured to implement the following steps: detecting a drift period in a dataset, the drift period including a start time, wherein the dataset pertains to a machine learning (ML)-based model; determining a first confidence value for a period preceding the start time and a second confidence value for a period following the start time; and predicting a drift period duration for the dataset using an ML-based drift model that is trained based on the first and second confidence values.

Abstract: Techniques described herein relate to systems and methods for workload placement based on subgraph similarity. Such techniques may include obtaining an encoded workload graph based on receiving a workload execution request; using the encoded workload subgraph to obtain encoded graphs representing previous workload executions, encoded subgraphs representing infrastructures on which the workload were executed, resource usage information, and execution metrics; using the encoded infrastructure subgraphs using subgraph similarity to identify candidate infrastructure subgraphs, using an ML model to predict an execution metric for an execution of the workload using the candidate; and selecting a best candidate infrastructure on which to execute the workload based on the predicted execution results.

Abstract: One example method includes performing the various operations concerning a model that is operable to predict resource usage and execution time of computing workloads. The operations include extracting a fingerprint associated with telemetry data, and the telemetry data was generated based on performance of one of the computing workloads, in a constrained infrastructure, checking a fingerprint catalog to determine if there is a same or similar fingerprint to the fingerprint, when the same or similar fingerprint is found in the fingerprint catalog, inferring that the model includes information about the computing workload and the model is able to predict telemetry data and execution time for the computing workload in a target infrastructure, and when the same or similar fingerprint is not found, inserting the extracted fingerprint into the fingerprint catalog, and generating a retrained model by retraining the model using the telemetry data associated with the extracted fingerprint.

Abstract: Techniques are provided for assigning workloads in a multi-node processing environment using resource allocation feedback from each node. One method comprises obtaining feedback from distributed nodes that process workloads, wherein the feedback for a given node indicates (i) an allocation of resources, and (ii) a number of executing workloads. In response to receiving a given workload to be processed, candidate nodes are identified to execute the given workload; and the given workload is assigned to a given candidate node based on an amount of available resources on each candidate node and/or a stability of resource adjustments made for each candidate node. The stability of the resource adjustments made for each candidate node can be evaluated based on a maximum resource adjustment made for a given candidate node relative to a maximum resource adjustment made for each of the candidate nodes.

Abstract: Techniques are provided for adaptive resource allocation for multiple workloads. One method comprises obtaining a dynamic system model based on a relation between an amount of a resource for multiple iterative workloads and a predefined service metric; obtaining an instantaneous value of the predefined service metric; applying to a given controller associated with a given iterative workload of the multiple iterative workloads: (i) the dynamic system model, (ii) an interference effect of one or more additional iterative workloads on the given iterative workload, and (iii) a difference between the instantaneous value and a target value for the predefined service metric. The given controller applies an adjustment to the amount of the resource for the given iterative workload based at least in part on the difference. The resource allocation for the multiple iterative workloads can be performed in a sequence substantially in parallel with an execution of the iterative workloads.

Abstract: Techniques described herein relate to a method for resource allocation using fingerprint representations of telemetry data. The method may include receiving, at a resource allocation device, a request to execute a workload; obtaining, by the resource allocation device, telemetry data associated with the workload; identifying, by the resource allocation device, a breakpoint based on the telemetry data; identifying, by the resource allocation device, a workload segment using the breakpoint; generating, by the resource allocation device, a fingerprint representation using the workload segment; performing, by the resource allocation device, a search in a fingerprint catalog using the fingerprint representation to identify a similar fingerprint; obtaining, by the resource allocation device, a resource allocation policy associated with the similar fingerprint; and performing, by the resource allocation device, a resource policy application action based on the resource allocation policy.

Abstract: One example method includes accessing I/O traces, generating parameters based on the I/O traces, and defining an autoencoder deep neural network, training the autoencoder deep neural network using the parameters, collecting and storing new I/O traces, computing an encoded features difference series using the new I/O traces, detecting breakpoints in the encoded features difference series, evaluating a utility of the breakpoints, and performing an action based on the breakpoint utility evaluation.

Abstract: Techniques are provided for predicting a time to complete a data protection operation. One method comprises obtaining metadata for (i) a given data protection appliance, and/or (ii) a cluster of similar data protection appliances comprising the given data protection appliance; evaluating first level features using the obtained metadata; evaluating a second level feature using some of the evaluated first level features; and processing one or more of the first level features, and the second level feature, using a model that provides a predicted time to complete a data protection operation with respect to data of a protected device associated with the given data protection appliance. The predicted time may comprise a tolerance based on a robustness factor. The predicted time may be based on a number of protected devices that are concurrently undergoing a data protection operation with the protected device for one or more time intervals.

Abstract: One example method includes running multiple iterations of a computing workload, for each iteration of the computing workload, for each iteration of the computing workload, using a reinforcement learning process to generate an initial infrastructure allocation for the computing workload, and a reward function of the reinforcement learning process generates a respective reward for each initial infrastructure allocation, running an accumulator map voting process to generate a total reward for each initial infrastructure allocation, and identifying the initial infrastructure allocation with the largest total reward and assigning that initial infrastructure allocation to the computing workload.

Abstract: One example method includes, in an initial iteration of a hyperparameter search process, randomly selecting, from an entire hyperparameter space, an initial set of one or more hyperparameters, wherein the hyperparameters are usable in a machine learning process, generating an initial Gaussian probability distribution around the initial set of hyperparameters, and generating a Gaussian probability distribution function by normalizing the initial Gaussian probability distribution to delimit an initial portion of the hyperparameter space, and the initial portion of the hyperparameter space is smaller than the hyperparameter space.

Abstract: One example method includes identifying a group of asset backups to be performed, and each asset backup is associated with a respective asset and has an associated backup time and RPO, selecting an asset backup to run first, and the asset backup that will run first is chosen based on a start deadline of that asset backup relative to respective start deadlines of one or more other asset backups, and the start deadline falls within a time slot, selecting a stream from a group of streams for the selected asset backup, and the selected stream is a stream with a lowest value of first available time slot, and backing up the asset at a backup server by running the selected asset backup, and backup begins at a start time that is a time when the selected stream becomes available, and the asset backup runs on the selected stream.

Abstract: A backup orchestrator for providing backup services to entities includes storage for storing recovery point objectives for the entities and a backup manager. The backup manager selects an optimization periodicity based a number of backups to be generated to meet a portion of the recovery point objectives; makes a determination that at least one of the portion of the recovery point objectives has a maximum allowable unbacked up period of time that is greater than the optimization periodicity; in response to the determination: load balances the number of backups across multiple optimization periods, based on the optimization periodicity, of a balanced backup schedule; selects a backup generation time for each of the to be generated backups in each of the optimization periods of the balanced backup schedule; and generates the number of backups using the balanced backup schedule.

Abstract: An efficient method for generating asset backup scheduling plans. Within a data protection environment, at least two service level metrics may be observed—a recovery point object (RPO) and a recovery time objective (RTO). In order to meet acceptable values for these metrics, on par with established service level agreements, infrastructure employed throughout the data protection environment, as well as the scheduling of asset backup operations, tend to grow in complexity. To address service distributions potentially emerging from the aforementioned complexities, the disclosed method proposes a heuristic approach to generating asset backup scheduling plans, which consider factors such as backup device limitations, RPO violation minimization, asset usage, and asset prioritization.

Abstract: A method for managing storage devices includes obtaining a storage device cluster request, and in response to the storage device cluster request: obtaining a set of storage device telemetry entries associated with a plurality of storage devices, performing a telemetry normalization on the storage device telemetry entries to obtain a set of normalized entries, performing a pairwise evaluation on the set of normalized entries to obtain a set of initial storage device clusters, wherein a storage device cluster in the set of initial storage device clusters comprises a portion of the plurality of storage devices, performing a cluster re-evaluation on the set of initial storage device cluster groups to obtain a set of updated storage device clusters, updating a backup policy based on the set of updated storage device cluster groups, and performing a backup operation on a storage device based on the backup policy.

Abstract: An efficient method for generating asset backup scheduling plans. Within a data protection environment, at least two service level metrics may be observed—a recovery point object (RPO) and a recovery time objective (RTO). In order to meet acceptable values for these metrics, on par with established service level agreements, infrastructure employed throughout the data protection environment, as well as the scheduling of asset backup operations, tend to grow in complexity. To address service distributions potentially emerging from the aforementioned complexities, the disclosed method proposes a heuristic approach to generating asset backup scheduling plans, which consider factors such as backup device limitations, RPO violation minimization, asset usage, and asset prioritization.