Obviousness

The obviousness of US patent 7577099 under 35 U.S.C. § 103 can be analyzed by combining the teachings of the cited prior art with the general knowledge of a Person Having Ordinary Skill in the Art (PHOSITA) in network management and algorithms. The core of the invention lies in using a greedy approximation algorithm with specific hit and coverage ratios for fault localization.

Independent Claims of US7577099

The independent claims (Claims 1, 2, and 3) of US7577099 describe:

• Generating an observation from monitoring data associated with network elements.
• Querying a database for a plurality of models of risks (e.g., Shared Risk Link Groups or SRLGs).
• Calculating a hypothesis from these risk models that explains the observation, specifically by applying a greedy approximation algorithm.
• This greedy approximation explicitly involves maximizing a hit ratio and a coverage ratio, defined as:
  • Hit ratio: |Gi ∩ O| / |Gi| (fraction of circuits in a group Gi that are part of observation O).
  • Coverage ratio: |Gi ∩ O| / |O| (portion of observation O explained by a group Gi).

Prior Art References

The primary prior art reference cited by the examiner in US7577099 is:

1. US20020019870A1 (International Business Machines Corporation - "Proactive on-line diagnostics in a manageable network"): This patent, published prior to the filing date of US7577099, discloses a method, system, and program product for proactive online diagnostics in a manageable network. It teaches "receiving network events and correlating the network events to determine a possible fault location within the network." It describes collecting "network events" from "monitoring network elements" using a "network event collector." Furthermore, it uses "correlation rules" stored in a "correlation rule database" to perform fault diagnosis and identify "specific physical and/or logical component(s) that are causing the network problem."

The other cited patent, WO2006005665A2, focuses on neural networks for predicting context changes and does not appear to directly teach the specific fault localization approach of US7577099 to the same extent as US20020019870A1.

Obviousness Analysis under 35 U.S.C. § 103

A person having ordinary skill in the art (PHOSITA) in network management and algorithms, at the time of the invention (before April 6, 2006), would have found the claimed invention obvious in light of US20020019870A1 combined with general knowledge of algorithms and diagnostic metrics.

1. Elements Taught by US20020019870A1:

• Method for automatically localizing failures in a network: US20020019870A1 explicitly teaches "automated network diagnostics" and "correlating the network events to determine a possible fault location within the network."
• Generating an observation from monitoring data associated with network elements in the network: This is taught by US20020019870A1's "network event collector" receiving "network events" from "monitoring network elements." These events constitute the "observation" of link failures.
• Querying a database for a plurality of models of risks: US20020019870A1 discloses a "correlation rule database" that stores "correlation rules" used for fault diagnosis. A PHOSITA would understand these correlation rules to function as models that describe relationships between network events and potential fault locations, functionally similar to the "models of risks" (e.g., SRLGs) described in US7577099, which represent links likely to be impacted by component failures. The concept of modeling network dependencies to anticipate failures was well-known in network management.

2. Elements Rendered Obvious by Combination with General Knowledge:
The distinguishing features of US7577099 are the application of a greedy approximation algorithm to calculate the hypothesis and the explicit maximization of hit ratio and coverage ratio for guiding this approximation.

• Calculating a hypothesis from said plurality of models of risks that explains said observation: US20020019870A1 broadly teaches "determining a possible fault location" by correlating events using stored rules. Improving this determination process would be a natural goal for a PHOSITA.
• Applying a greedy approximation: The problem of explaining a set of observed link failures with a minimal set of underlying causes (risk groups) is analogous to a set cover problem. Greedy algorithms were a well-known and standard heuristic for efficiently approximating solutions to set cover and similar optimization problems in computer science and operations research prior to 2006. A PHOSITA seeking to improve the efficiency and accuracy of fault localization beyond simpler correlation methods would naturally consider such established algorithmic approaches.
• Maximizing a hit ratio and a coverage ratio: The "hit ratio" (|Gi ∩ O| / |Gi|) and "coverage ratio" (|Gi ∩ O| / |O|) are straightforward mathematical expressions of how well a potential fault (risk group Gi) aligns with the observed failures (O). The hit ratio measures the precision of a proposed fault group (how many of its expected failures actually occurred), while the coverage ratio measures its recall (how much of the observed failures are explained by this group). These types of metrics are intuitive and commonly used in diagnostic, statistical, and information retrieval contexts to evaluate the effectiveness of a hypothesis in explaining observed data. A PHOSITA designing an algorithm to identify fault causes would naturally employ such quantitative metrics to guide the selection process, particularly in a greedy fashion (e.g., by selecting the risk group that explains the most unexplained observations).

3. Motivation to Combine:
A PHOSITA, aiming to enhance the "proactive on-line diagnostics" and "fault location" capabilities described in US20020019870A1, would be motivated to adopt more sophisticated algorithmic techniques. The existing manual correlation or simpler automated mechanisms might have suffered from inefficiencies or inaccuracies, as acknowledged in the background of US7577099 itself.

By combining the event collection and rule-based correlation framework of US20020019870A1 with a computationally efficient greedy approximation algorithm guided by quantitative metrics (hit and coverage ratios), a PHOSITA could achieve a more robust, systematic, and automated approach to fault localization. This combination would lead to a more effective diagnostic tool that identifies the most probable fault hypotheses by optimizing how well potential causes explain observed failures, thereby reducing downtime and increasing network efficiency.

Conclusion for Claims 2 and 3:
Since Claims 2 (computer-readable storage medium) and 3 (apparatus) recite the same method steps as Claim 1, if the method of Claim 1 is obvious, then its implementation using conventional computing hardware and software, as described in US20020019870A1 which also details a system and program product, would also be obvious to a PHOSITA.