Amazon Technologies, Inc.

Andreas Louca

Mark Noel Kelly

George Georgiou

Toni Pivcevic

João André Henriques Ferreira

Goran Peretin

Filip Boltuzic

Michael Achola

Andrew Miehs

Jason Svein Ashworth

Zhiheng Gao

Yan Xu

A network event remediation service receives event logs from network computing devices to identify issues associated with these network computing devices. Based on these issues, the service generates various tasks to address the identified issues. The service obtains historical task data in order to determine whether any tasks do not comport with historical tasks performed by the network computing devices and removing such tasks. The service obtains health data from the network computing devices to calculate device scores for each device. The service uses the device scores to prioritize the tasks for each network computing device. These tasks are then dispatched to the network computing devices.

Network event automatic remediation service

Technologies are provided for correlating alert messages using placeholder correlation records. When a network alert message is received, a network component associated with the alert can be identified and a correlation record can be created that is associated with the identified network component. A model of the computer network can be used identify another network component that is related to the identified network component. If a correlation record for the related component does not exist, then a placeholder correlation record for the related component can be created. If another alert message is received that is associated with another component that is also related to the same component, then the placeholder correlation record can be used to correlate the alert messages. An event message can be generated that indicates an occurrence of a network event involving the related component associated with the placeholder correlation record. Placeholder correlation records can be used to associate alert messages with previously generated events.

Network alert correlation

Technologies are provided for correlating alert messages using correlation domains. Alert messages can be received that are related to computer network components. The alert messages can be partitioned using separate correlation domains and analyzed by separate alert processors. When an alert message is received, it can be analyzed to determine a correlation domain with which the alert message is associated. The alert message can then be correlated with other alert messages associated with the correlation domain, and alert messages that are not associated with the correlation domain can be ignored. Groups of alert messages associated with different correlation domains can be correlated independently by different alert processors. Events generated by an alert processor can include correlation domain information. Events associated with different correlation domains can be handled differently.

The following description is directed to a network monitoring service. In one example, a method can include receiving a query including a reference to a logical device of a network and an action to perform on the logical device. The logical device can include a plurality of interconnected network devices. The method can include determining whether the action can be performed on the logical device based at least on combining scores of the network devices of the logical device. The score of an individual network device can be calculated by recursively calculating scores of neighboring network devices of the individual network device. The method can include transmitting a response indicating whether the action can be performed on the logical device.

Network monitoring service

Techniques and solutions are provided for detecting the source of link errors in a cut-through forwarding network fabric. For example, a network link that is the source of the link errors can be automatically located by polling a network device a number of times to identify a network interface that has an increasing error rate and tracing network interface error rates back through one or more upstream network devices to locate the network link that is the source of the link errors.

Growing community of inventors

Seattle, WA, United States of America

Andreas Louca