CPC G06F 16/24545 (2019.01) [G06F 16/285 (2019.01)] | 19 Claims |
1. A method, comprising:
monitoring, by a process, an access pattern of a plurality of objects in a particular database on a user-connection basis;
calculating, by the process, a wait-time metric for database sessions for the plurality of objects on the user-connection basis;
classifying, by the process and based on the access pattern, a first number of the plurality of objects in the particular database on the user-connection basis as often-used objects and a second number of the plurality of objects in the particular database on the user-connection basis as not-recently-used objects;
classifying, by the process and based on machine-learning-adjustable thresholds, the wait-time metric for the database sessions on the user-connection basis as a wait-time classification that is one of either: an acceptable wait-time metric, a warning-level wait-time metric, or a critical-level wait-time metric, wherein the machine-learning-adjustable thresholds are dynamically adjusted using an anomaly detection model that is applied to performance metrics of the particular database comprising measurements of total session time, time spent waiting for resources, and processing time; and
generating, by the process, a graphical interface that indicates, for each user-connection pairing of the particular database, an associated wait-time metric and a graphical indication of an associated wait-time classification, as well as an associated first number of often-used objects and an associated second number of not-recently-used objects.
|