Boosting performance: Gradient Clock Synchronisation with two-way measured links

TL;DR

This thesis enhances the Gradient Clock Synchronisation algorithm by adopting a two-way measurement model, removing previous restrictions, and significantly improving synchronization accuracy and robustness in practical deployments.

Contribution

It introduces a two-way measurement paradigm to GCS, removes link length restrictions, and provides a detailed error analysis with improved bounds.

Findings

Reduction of estimation error contribution to 0.1% of link delay

Matching upper bounds on local and global skew

Enhanced model for practical GCS deployment

Abstract

This master thesis extends the formal model of the GCS algorithm as presented by (Fan and Lynch 2004, 325), (Lenzen, Locher and Wattenhofer 2008, 510) and (F\"ugger et al. 2023) to operate under implementation-near assumptions by replacing the one-way measurement paradigm assumed in prior work by the two-way measurement paradigm. With this change of paradigm, we remove many restrictions previously enforced to allow provable performance. Most notability, while maintaining the core behaviour of GCS, we: 1. Lift the requirement for unitary link lengths and thereby create a realistic model for flexible deployment of implementations of GCS in practice. 2. Provide a formal model of frequency sources assumed in prior work. 3. Perform a fine grained distinction between the different components of the algorithm's estimation error and globally reduce its impact by multiple orders of magnitude. 4. Significantly reduce the contribution of the uncertainty to the algorithm's estimation error to be in the range of 10\% to 0,1\% of the delay per link instead of being in the oder of the delay per link as in prior work and show matching upper bounds on the local and global skew of GCS.