# Window-based Streaming Graph Partitioning Algorithm

arXiv: https://arxiv.org/abs/1902.01543 · HTML: https://mlsanity.com/paper/1902.01543/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Feb 2019.

## Built on by (most cited first, 4 of 4)

- [More Recent Advances in (Hyper)Graph Partitioning](https://mlsanity.com/paper/2205.13202/) (2022)
- [An Experimental Comparison of Partitioning Strategies for Distributed Graph Neural Network Training](https://mlsanity.com/paper/2308.15602/) (2023)
- [Can Graph Reordering Speed Up Graph Neural Network Training? An Experimental Study](https://mlsanity.com/paper/2409.11129/) (2024)
- [Armada: Memory-Efficient Distributed Training of Large-Scale Graph Neural Networks](https://mlsanity.com/paper/2502.17846/) (2025)

## Abstract (arXiv)

In the recent years, the scale of graph datasets has increased to such a degree that a single machine is not capable of efficiently processing large graphs. Thereby, efficient graph partitioning is necessary for those large graph applications. Traditional graph partitioning generally loads the whole graph data into the memory before performing partitioning; this is not only a time consuming task but it also creates memory bottlenecks. These issues of memory limitation and enormous time complexity can be resolved using stream-based graph partitioning. A streaming graph partitioning algorithm reads vertices once and assigns that vertex to a partition accordingly. This is also called an one-pass algorithm. This paper proposes an efficient window-based streaming graph partitioning algorithm called WStream. The WStream algorithm is an edge-cut partitioning algorithm, which distributes a vertex among the partitions. Our results suggest that the WStream algorithm is able to partition large graph data efficiently while keeping the load balanced across different partitions, and communication to a minimum. Evaluation results with real workloads also prove the effectiveness of our proposed algorithm, and it achieves a significant reduction in load imbalance and edge-cut with different ranges of dataset.