SIGMA: A Versatile Streaming Graph Partitioner for Vertex- and Edge-Balanced Distributed GNN Training
Barbara Hoffmann, Shai Dorian Peretz, Adil Chhabra, Ahmet Kadir Yalcinkaya, Ruben Mayer, Christian Schulz · 2026
Read SIGMA: A Versatile Streaming Graph Partitioner for Vertex- and Edge-Balanced Distributed GNN Training by Barbara Hoffmann, Shai Dorian Peretz, Adil Chhabra, Ahmet Kadir Yalcinkaya, Ruben Mayer, Christian Schulz (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03519.txt.
Description
Hoffmann and co-authors present SIGMA, a streaming partitioner for distributed graph neural network training. How a graph is split across machines drives training cost. Older partitioners pick one side only, vertex or edge. They also chase one goal, like edge cut, under one balance rule. SIGMA handles both split styles in one frame.
Vertex and edge balance are tracked at the same time as cut quality. A clustering step adds global structure before the streaming pass. That step lifts quality while keeping the speed of streaming methods. Six benchmark graphs test the system across domains and scales. Two training systems run the trials, one edge-partitioned and one vertex-partitioned. SIGMA performs strongly in both settings. It often beats rival streaming tools and rivals in-memory partitioners. Gains are largest for edge-partitioned training. Teams get one tool for very different training setups.
Catalog record
| CREATOR | Barbara Hoffmann, Shai Dorian Peretz, Adil Chhabra, Ahmet Kadir Yalcinkaya, Ruben Mayer, Christian Schulz |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 63 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03519