US 7577099

US Patent 7577099: Method and Apparatus for Fault Localization in a Network

Title: Method and apparatus for fault localization in a network

Assignee: AT&T Corp

Inventors: Albert G. Greenberg, Ramana Rao Kompella, Alex C. Snoeren, Jennifer Yates

Filing Date: April 6, 2006

Issue Date: August 18, 2009

Abstract: The patent details a method and apparatus for the automatic localization of network failures. This process involves receiving an "observation" of link failures from multiple data sources and querying a database for "models of risks." These risk models, stored in a Shared Risk Link Group (SRLG) database, represent links that would likely be impacted by the failure of specific network components. A Spatial Correlation Engine (SCORE) then applies an algorithm to these models and observations to calculate a "hypothesis" that explains the observed failures, thereby pinpointing their location.

Plain-Language Overview of Independent Claims:

• Claim 1 (Method for automatically localizing failures): This claim describes a step-by-step process to automatically identify the source of failures in a network. It begins by collecting monitoring data from network devices to create an "observation," which is essentially a record of detected link failures. Next, the system consults a database containing "models of risks," which are pre-defined groups of network links that are expected to fail together if a particular underlying component fails (e.g., a shared fiber optic cable). Finally, a "greedy approximation" algorithm is applied to these observations and risk models to calculate a hypothesis—the most probable cause of the failures. This algorithm aims to maximize two key metrics: a "hit ratio" (the proportion of links in a risk group that are actually observed to have failed) and a "coverage ratio" (the proportion of observed failures that can be explained by a particular risk group).

• Claim 2 (Computer-readable storage medium): This claim covers a computer storage medium (such as a hard drive or solid-state drive) that holds software instructions. When a computer's processor executes these instructions, it carries out the exact same method steps as described in Claim 1 for automatically finding and localizing network failures.

• Claim 3 (Apparatus for automatically localizing failures): This claim describes a physical system or device built to automatically locate network failures. It includes distinct "means" (which can be hardware, software, or a combination) for performing the tasks outlined in Claim 1: a "means for generating an observation" from network monitoring data, a "means for querying a database" for the risk models, and a "means for calculating a hypothesis" using the greedy approximation algorithm to maximize hit and coverage ratios and explain the observed failures.

CAFC 2026 Dockets:

As of April 26, 2026, US Patent 7577099B1 is not listed in the publicly available scheduled cases for the U.S. Court of Appeals for the Federal Circuit for May, June, or July 2026. The patent's legal status is recorded as "Expired - Fee Related," with an adjusted expiration date of April 18, 2027.