Oracle Corporation
System and method for providing machine learning based memory resiliency

Abstract:

In accordance with an embodiment, a service, application, or other component can operate in association with a virtual machine at a computer server or device. The system can determine memory heap usage based on collected metrics, via one or more push and/or pull communications of data, and automatically adjust or tune concurrent processing of registered components, in response to such determination. In accordance with an embodiment, the system is extensible to support different low memory conditions and different components that require tuning of their processes. Datasets describing information such as heap usage, garbage collection, and stuck threads, can be captured and stored; and machine learning processes applied to those datasets, to deduce new low memory conditions. If memory pressure continues to be high, and is not controlled by tuning component processes, the system can determine likelihood of a memory leak, and restart the server.