Efficient GPU-Centric Evolving Graph Processing at Scale
Talkat 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26) · 2026-07-13..07-15·
summary
POEGA is a GPU-centric framework that uses proxy graphs to keep a compact in-memory approximation, runs a fused kernel to process many snapshots concurrently with coalesced neighbor access, and applies bound-based pruning to skip redundant updates across snapshots. It adds an adaptive multi-version array (AMVA) to compact multi-snapshot vertex states on-GPU. On five real-world datasets and six algorithms POEGA reports 3.7–23.5× speedups over prior EGA systems by shifting work from host IO to parallel GPU computation.
AI-written summary · about
chapters (11)
- 0:00Evolving graph analytics intro
- 1:22Incremental graph updates
- 2:24GPU deployment challenges
- 3:50Out-of-GPU memory limits
- 4:54Proxy graphs as solution
- 6:39Concurrent execution design
- 8:46Bound-based pooling
- 9:52Multi-snapshot bottlenecks
- 10:54Adaptive multi-version array
- 11:59POG framework performance
- 13:44Summary and open-source
related talks
49mGraph Based Processing of Big Images
15mMGG: Accelerating Graph Neural Networks with Fine-Grained Intra-Kernel Communication...
17mBGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing
16mPathWeaver: A High-Throughput Multi-GPU System for Graph-Based Approximate...
18mFast Inference for Probabilistic Graphical Models
13mGNNAdvisor: An Adaptive and Efficient Runtime System for GNN Acceleration on GPUs
11mA Design for Massively Parallel Gibbs Sampling on the GPU via Static and Dynamic(…)
27mA Profile-Driven Dynamic Optimization Framework for GPU Applications