PetPS: Supporting Huge Embedding Models with Persistent Memory
Summary: PetPS: first production PM parameter server for huge embedding models, using a PM-specific hash index and NIC offload to curb latency and CPU cost. 1.3–1.7× throughput gains vs state-of-the-art PM indexes; 2.9–5.5× latency reductions at equal throughput; deployed in Kuaishou with ~30% TCO savings. (summarized by gpt-5-nano on Feb 09 2026)
Incoming Non-self Citations Over Time
Authors
- 1. Minhui Xie
- 2. Youyou Lu
- 3. Qing Wang
- 4. Yangyang Feng
- 5. Jiaqiang Liu
- 6. Kai Ren
- 7. Jiwu Shu
Incoming Citations (Sorted by Pagerank)
Showing 2 of 2 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 10,482 | CARINA: An Efficient CXL-Oriented Embedding Serving System for Recommendation Models | 2025 | SIGMOD | 4.1905499e-05 |
| 11,012 | Sorting on Byte-Addressable Storage: The Resurgence of Tree Structure | 2024 | VLDB | 4.1905499e-05 |
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 6 of 6 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 643 | FPTree: A Hybrid SCM-DRAM Persistent and Concurrent B-Tree for Storage Class Memory | 2016 | SIGMOD | 0.00018733394 |
| 1,823 | Dash: Scalable Hashing on Persistent Memory | 2020 | VLDB | 0.00010403355 |
| 2,678 | HET: Scaling out Huge Embedding Model Training via Cache-enabled Distributed Framework | 2022 | VLDB | 8.3224016e-05 |
| 3,050 | Viper: An Efficient Hybrid PMem-DRAM Key-Value Store | 2021 | VLDB | 7.6491863e-05 |
| 5,169 | HET-GMP: A Graph-based System Approach to Scaling Large Embedding Model Training | 2022 | SIGMOD | 5.642415e-05 |
| 5,992 | NuPS: A Parameter Server for Machine Learning with Non-Uniform Parameter Access | 2022 | SIGMOD | 5.2380905e-05 |
Previous
Page 1 / 1
Next