Multiscale energy network tomography and smartNIC-Accelerated In-band Network telemetry for network internal state monitoring
Abstract
Network internal performance statistics are crucial for the control, operation and management of computer networks. This research work explores multiscale wavelet energy tomography using Discrete Wavelet Transform (DWT) based on end-to-end measurements for characterizing the delay statistics of internal links. Additionally, an efficient and flexible monitoring platform using a recently developed concept of In-band Network Telemetry is carefully designed and proved to be accurate and cost-efficient for monitoring network internal nodes.
Much effort and ingenuity has been applied to develop tomographic methods to derive information concerning link-level performance statistics from relatively available end-to-end measurements. However, there has been recognition in recent years that network phenomena, including network attacks, may manifest with distinct spectral distribution present in time series of associated network measurements. For time series in general, multiscale analysis using DWT is a powerful method to extract detailed signal components across frequencies. This research showed how a tomographic analysis of the DWT of end-to-end measurements can be used to provide an unbiased estimates of the energy spectrum of the contributions to those measurements from the path intersection. It also illustrates application of the method to detect low-rate periodic attack.
In-band Network Telemetry (INT), on the other hand, provides granular monitoring of performance and load on network elements by collecting information in the data plane without requiring intervention from control plane. INT enables traffic sources to embed telemetry instructions in data packets, avoiding separate probing or infrequent management-based monitoring. INT sink nodes track and collect metrics by retrieving INT metadata instructions appended by different sources of INT information. However, tracking the INT state in packets arriving at the sink is both compute intensive (requiring complex operations on each packet), and challenging for the standard P4 match-action packet processing pipeline to maintain line-rate. This research provided an accelerated monitoring platform on monitoring INT packets using SmartNIC and also showed how to optimize the INT operations to achieve cost-efficient over networks.
Citation
Feng, Yixiao (2021). Multiscale energy network tomography and smartNIC-Accelerated In-band Network telemetry for network internal state monitoring. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /195219.
Related items
Showing items related by title, author, creator and subject.
-
Kim, Seongbae (Texas A&M University, 1991)Not available
-
Seacat, Russell Holland (Texas A&M University. Libraries, 1974)
-
D'Antonio, Benjamin (Texas A&M University. Libraries, 2019-05)Predicting the effects of Polypharmacy is a difficult task, and a great amount of money is spent annually remedying the effects of negative drug interactions arising from Polypharmacy. However, Machine Learning can be used ...