US 12,278,747 B1
Systems and methods to analyze root cause anomaly
Rajendra Kumar Vippagunta, Issaquah, WA (US); Syed Furqhan Ulla, Kirkland, WA (US); Sunayana Vempati, Seattle, WA (US); Yekesa Srinivasa Kosuru, Westford, MA (US); Namita Das, Seattle, WA (US); Devesh Ratho, Seattle, WA (US); Shashwat Srijan, Bellevue, WA (US); Lenon Alexander Minorics, Stuttgart (DE); Patrick Bloebaum, Mountain View, CA (US); and David Kernert, Stuttgart (DE)
Assigned to Amazon Technologies, Inc., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Sep. 30, 2022, as Appl. No. 17/958,159.
Int. Cl. H04L 43/067 (2022.01); H04L 41/0631 (2022.01); H04L 43/0817 (2022.01)
CPC H04L 43/067 (2013.01) [H04L 41/0631 (2013.01); H04L 43/0817 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
obtaining a plurality of infrastructure templates for a plurality of portions of a system, wherein the plurality of infrastructure templates indicates dependencies between resources within a portion of the plurality of portions;
generating a directed graph comprising nodes that represent a plurality of resource-metric pairs of a system using the plurality of infrastructure templates, wherein a resource-metric pair of the plurality of resource-metric pairs indicates an association between a resource of a system and a metric of the resource;
obtaining time series data that indicates two or more anomalies detected from the system;
determining that a subset of nodes of the directed graph are associated with two or more anomalies, wherein each node of the subset of nodes corresponds to an anomaly of the two or more anomalies;
determining dependencies between the subset of nodes based, at least in part, on a policy that indicates data access permissions between resources associated with the subset of nodes;
determining a node from the subset of nodes based at least in part on dependencies between nodes from the subset of nodes and the time series data; and
providing an indication that an anomaly associated with the node is a root cause anomaly.