NSF-MeitY: Experimental Demonstration of Application-Aware, Root-Cause Analysis for Service Disruption: A Cross-Layer Approach

About This Grant

In today’s application-centric networking, the end user’s quality of experience is of paramount importance; for example, end users do not want to experience delays in website response or interruption in video downloads or video viewing. On the other hand, they do not always provide high quality of experience. Often, this is attributed to the network layer (e.g., congested link, packet loss); however, issues may also lie in the lower fiber optic communications layer. Optical fibers carry information with high signal quality and reliability; yet, at long distances, signal quality may degrade due to various phenomena, including fiber/amplifier aging, component anomalies, etc. To improve application performance and user experience, integrated end-to-end management is needed across the application, network, and physical (optical) layers. This approach can help network operators perform layer-specific root-cause analysis of the affected services, predict probable issues, and take preemptive actions to avoid poor service. This project addresses this critical need; it investigates cross-layer monitoring and analytics to correlate the effect of the optical layer on network-layer dynamics and vice-versa. This joint NSF-MeitY project will investigate machine-learning (ML) techniques for application-aware root-cause analysis and fault prediction. First, it will develop an optical testbed, which will be utilized by the ML model for anomaly detection and classification. It will implement routers for network-layer integration and monitoring agents at both optical and network layers. Second, a cross-layer analytics platform will be developed to achieve holistic observability from application layer to optical layer. It will build a correlator to map the impact of failures in optical layer (by simulating controlled anomalies in the optical testbed) on the performance of different applications at the network layer. The cross-layer correlator will employ ML-assisted tools to explain the root cause of service degradation at network layer. Finally, efficient strategies for preemptive application-aware fault remedy (such as re-routing of certain applications) will be developed. It will also study remedial actions at optical layer such as re-routing the affected lightpath, lighting up a new lightpath, etc. Strong international collaboration between UC Davis, IIIT-Delhi, and IIT-Madras will bring expertise and shared knowledge for effective problem solving in related research areas and verify the research outcomes on the IIT-Madras testbed. In summary, this project – with its cross-layer monitoring and analytics framework – will develop new insights in the discipline of reliable networking and contribute to the development of solutions for supporting robust future services. The project will provide excellent opportunity for training of graduate students, enhancing the educated workforce in STEM disciplines. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.