公开(公告)号：US11469983B1

公开(公告)日：2022-10-11

申请号：US16820576

申请日：2020-03-16

Applicant: Amazon Technologies, Inc.

Inventor： Stanislav Vladimirovich Malov ,  Arran McCabe ,  Nikolaos Pavlakis ,  Alan O'Leary ,  Ivan Emilov Goychev

IPC: G06F15/173 ,  H04L43/0823 ,  H04L41/0654 ,  H04L41/12 ,  H04L43/0829 ,  H04L43/091

Abstract: An exposure of a user of a network to an adverse network event can be estimated based on network topology data, network traffic associated with the user, and network performance data. A subgraph of the network topology including candidate paths through the network devices and links of the topology that were traversed by the user's network traffic can be estimated using the network topology data and the network traffic data associated with the user. Using the network performance data, an adverse network event can be mapped on the candidate paths of the subgraph to determine an impact of the adverse network event on the user's network traffic. Upon determining that the adverse network event impacts the user, the user can be notified regarding the adverse network event. Optionally, actions to reduce the impact of the network event on the user can be performed.

公开(公告)号：US12259799B1

公开(公告)日：2025-03-25

申请号：US18126944

申请日：2023-03-27

Applicant: Amazon Technologies, Inc.

Inventor： Adrian John Hornsby ,  Laura Wingert Thomson ,  Warren Robert Russell ,  Alan O'Leary ,  Serafin Antonio Sedano Arenas

IPC: G06F11/00 ,  G06F11/26 ,  G06F11/263

Abstract: Disclosed are various embodiments for identifying fault injection impact zones. A test for an application can be received from a test service. An application schema can then be created that represents the systems or services that are involved in the execution of the application. Applications with a similar application schema are then identified using a first machine-learning model. A subset of the plurality of applications that have experienced a similar test are then identified with a second machine-learning model. Log files associated with each of the subset of applications are then analyzed to identify errors that occurred during the similar test. Each of the identified errors are then provided to the test service in response to receipt of the test.

公开(公告)号：US11539728B1

公开(公告)日：2022-12-27

申请号：US16820584

申请日：2020-03-16

Applicant: Amazon Technologies, Inc.

Inventor： Stanislav Vladimirovich Malov ,  Arran McCabe ,  Nikolaos Pavlakis ,  Alan O'Leary ,  Ivan Emilov Goychev

IPC: H04L9/40 ,  H04L43/0811 ,  H04L41/12 ,  H04L47/2483 ,  H04L47/122 ,  H04L43/16

Abstract: Network connectivity disruptions impacting users of a network, can be detected based on patterns in user network traffic and network topology data, e.g., by a monitoring server computer. Logged network traffic data can be filtered to identify anomalous data flows. The anomalous data flows can be data flows indicating connection timeouts such as failed Secure Sockets Layer/Transport Layer Security (SSL/TLS) handshakes. Sources and destinations of the anomalous data flows can be mapped to corresponding physical locations using the network topology data, and the anomalous data flows can be grouped by source and destination, in order to determine an impact or scope of a network connectivity disruption. Users of the network can be notified regarding the network connectivity disruption, and optionally, actions can be taken to reduce the impact of the network connectivity disruption.