MACHINE LEARNING FOR LOCAL CACHING OF REMOTE DATA IN A CLUSTERED COMPUTING ENVIRONMENT

摘要

Clustered computing system monitoring, fault prediction and remediation. The disclosed techniques are used for predicting an impending occurrence of a fault. Prediction techniques involve gathering historical data pertaining to storage I/O operations performed in a clustered computing system. Ongoing storage I/O operations that are performed when a first node accesses data in local storage device as well as ongoing storage I/O operations that are performed when the first node accesses data in a remote storage device of a second node are monitored over time. On an ongoing basis, the monitored storage I/O operations are analyzed to identify impending remediable events. Before recurrence of a predicted impending remediable event, conditions that are known to have been present at the time of a historical fault are ameliorated. In some embodiments, avoidance of an impending remediable event is accomplished by moving data stored at a remote storage device to a local storage device.