KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference

Summary: KVDrive manages LLM KV caches holistically across GPU memory, DRAM, and SSD, rather than relying on increased sparsity. Attention-aware placement, tier coordination, and overlapped pipeline scheduling reduce data movement and stalls, delivering up to 1.74× higher throughput. (summarized by gpt-5.6-luna on Jul 26 2026)

Paper ID: 7450
Venue: SIGMOD
Year: 2026
Pagerank: 5.093636e-05
Overall Rank: 10,259 | 29.62%
DOI: 10.1145/3802077

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

1. Jian Lin (Hong Kong University of Science and Technology)
2. Jiazhi Mi (Hong Kong University of Science and Technology)
3. Zicong Hong (Hong Kong University of Science and Technology)
4. Haodong Wang (Hong Kong University of Science and Technology)
5. Qianli Liu (Hong Kong University of Science and Technology)
6. Haoyue Zhang (Hong Kong University of Science and Technology)
7. Peng Li (Xi'an Jiaotong University)
8. Song Guo (Hong Kong University of Science and Technology)

BibTeX Citation

@inproceedings{lin_sigmod26,
        title = {{KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference}},
        author = {Lin, Jian and Mi, Jiazhi and Hong, Zicong and Wang, Haodong and Liu, Qianli and Zhang, Haoyue and Li, Peng and Guo, Song},
        series = {{SIGMOD} '26},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3802077},
        url = {https://dl.acm.org/doi/10.1145/3802077},
        year = {2026}
}

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank

Outgoing Citations (Sorted by Pagerank)

Showing 2 of 2 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
4,430	PQCache: Product Quantization-based KVCache for Long Context LLM Inference	2025	SIGMOD	6.7091071e-05
6,432	RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference	2026	VLDB	5.8801533e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	10,210	CoDec: Prefix-Shared Decoding Kernel for LLMs	2026	SIGMOD
2	6,900	Apt-Serve: Adaptive Request Scheduling on Hybrid Cache for Scalable LLM Inference Serving	2025	SIGMOD
3	10,432	Beluga: A CXL-Based Memory Architecture for Scalable and Efficient LLM KVCache Management	2026	SIGMOD
4	10,191	AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System	2026	SIGMOD
5	13,291	OrbitFlow: SLO-Aware Long-Context LLM Serving with Fine-Grained KV Cache Reconfiguration	2026	VLDB
6	2,816	Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation	2025	SIGMOD
7	10,357	DepCache: A KV Cache Management Framework for GraphRAG with Dependency Attention	2026	SIGMOD
8	6,432	RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference	2026	VLDB
9	6,788	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD
10	4,430	PQCache: Product Quantization-based KVCache for Long Context LLM Inference	2025	SIGMOD