Abstract: Techniques are described for error log anomaly detection. In an implementation, error logs from an application are processed to generate training data. The error logs, for instance, are processed to remove personal information and other data such as numerical strings. The processed error logs are converted into embeddings to generate the training data. The training data is utilized to train an anomaly detection model. For instance, as part of training the anomaly detection model, an anomaly threshold is defined based on a loss value determined from output of the anomaly detection model. Further error logs from the application are then processed by the trained anomaly detection model to determine which of the further error logs are error anomalies, such as based on comparing loss values for the further error logs to the anomaly threshold.

Abstract: Capacity-based scaling of queue-based resources is described. Initially, a scaling system measures capacity of service processors that are instantiated at a cloud computing service system to provide a service on behalf of a service provider, and also measures a load on these processors. In contrast to conventional scaling systems—which base scalings on a number of queued messages which the instantiated service processors process to provide the service—the scaling system measures the load in terms of not only the number of messages held in a queue but also an input rate of the messages to the queue. The described scaling system then determines whether and by how much to scale the instantiated processors based on this number of messages and input rate. Given this, the scaling system instructs the cloud computing service system how to scale the instantiated service processors to provide the service.

Abstract: In implementations of anomaly detection for incremental application deployments, one or more computing devices of a system implement an evaluation module for anomaly detection. An error record of a legacy version of a software application is generated, as well as an error record of an incremental version of the application. Data in the error record of the legacy version of the application is organized and grouped into one or more groups, each group having a membership criterion. Data in the error record of the incremental application is organized and compared to the one or more groups of the legacy application error record data based on the membership criterion for each group. If an organized data of the incremental application does not belong to any of the groups, then the organized data is identified as an anomaly.

Abstract: In implementations of anomaly detection for incremental application deployments, one or more computing devices of a system implement an evaluation module for anomaly detection. An error record of a legacy version of a software application is generated, as well as an error record of an incremental version of the application. Data in the error record of the legacy version of the application is organized and grouped into one or more groups, each group having a membership criterion. Data in the error record of the incremental application is organized and compared to the one or more groups of the legacy application error record data based on the membership criterion for each group. If an organized data of the incremental application does not belong to any of the groups, then the organized data is identified as an anomaly.

Abstract: Capacity-based scaling of queue-based resources is described. Initially, a scaling system measures capacity of service processors that are instantiated at a cloud computing service system to provide a service on behalf of a service provider, and also measures a load on these processors. In contrast to conventional scaling systems—which base scalings on a number of queued messages which the instantiated service processors process to provide the service—the scaling system measures the load in terms of not only the number of messages held in a queue but also an input rate of the messages to the queue. The described scaling system then determines whether and by how much to scale the instantiated processors based on this number of messages and input rate. Given this, the scaling system instructs the cloud computing service system how to scale the instantiated service processors to provide the service.