Workshop: 13th Annual International Workshop on Innovating the Network for Data Intensive Science — INDIS
Authors: Shashwitha Puttaswamy (George Washington University/CAAREN) and Mariam Kiran (Oak Ridge National Laboratory)
Abstract: Modern wide-area and research networks support many different types of applications, each with its own perfor- mance requirements. Because of this diversity, no single routing strategy such as shortest path, least-utilized path, or Equal- Cost Multi-Path (ECMP) can consistently deliver the best results for every application. Existing telemetry mechanisms such as IOAM and gNMI are fundamentally class-agnostic and operate at 10–500ms timescales, while current SRv6-TE systems rely on controller-driven path switching based on out-of-band statistics. In this paper, we present a class-aware, in-header congestion feedback mechanism for SRv6 that (i) encodes traffic engineering (TE) class identity in the SRH Tag field, (ii) appends per-hop, per-class congestion evidence in a Telemetry Tuned Weight -SRv6 (TTW-SRv6) record carried in-band with the SRv6 tunnel, and (iii) uses load balancing algorithms per class within a distributed architecture. A new architecture is presented where a trusted edge classifier analyzes flow behavior.
Back to 13th Annual International Workshop on Innovating the Network for Data Intensive Science — INDIS Archive Listing
Back to Full Workshop Archive Listing