Dynamic cloud based alert and threshold generation

A non-transitory computer readable medium having instructions stored thereon that, when executed by one or more processors, cause the processors to predicting failures in a cloud infrastructure, the predicting comprising:

• generating a graphical representation of a plurality of features of the cloud infrastructure, the graphical representation comprising a plurality of nodes and corresponding relationships between the nodes, each node corresponding to one of the plurality of features;
• monitoring for events for the plurality of features, the events corresponding to one or more of the nodes, to generate monitored events;
• populating a graph database with the monitored events;
• classifying each of the nodes with a trained graph neural network (GNN), the classification comprises a prediction of a failure of at least one node, wherein the GNN is trained using an historical training dataset derived from the monitoring and the populating;
• based on the classifying, predicting a failure of a first failure node, wherein the first failure node does not have a corresponding defined alert;
• dynamically generating a new alert corresponding to the first failure node in response to the predicted failure of the first failure node, the dynamically generated new alert comprising a dynamically generated new threshold of the first failure node that will trigger the new alert;
• retraining the trained GNN using the new alert;
• and
• repeating the monitoring and triggering the new alert when the new threshold of the first failure node is exceeded.

A cloud infrastructure comprising:

• a graph database;
• and
• one or more processors configured to execute instructions for predicting failures in the cloud infrastructure, the predicting failures comprising:
• generating a graphical representation of a plurality of features of the cloud infrastructure, the graphical representation comprising a plurality of nodes and corresponding relationships between the nodes, each node corresponding to one of the plurality of features;
• monitoring for events for the plurality of features, the events corresponding to one or more of the nodes, to generate monitored events;
• populating the graph database with the monitored events;
• classifying each of the nodes with a trained graph neural network (GNN), the classification comprises a prediction of a failure of at least one node, wherein the GNN is trained using an historical training dataset derived from the monitoring and the populating;
• based on the classifying, predicting a failure of a first failure node, wherein the first failure node does not have a corresponding defined alert;
• dynamically generating a new alert corresponding to the first failure node in response to the predicted failure of the first failure node, the dynamically generated new alert comprising a dynamically generated new threshold of the first failure node that will trigger the new alert;
• retraining the trained GNN using the new alert;
• and
• repeating the monitoring and triggering the new alert when the new threshold of the first failure node is exceeded.