Peixin Zeng, Hao Huang, Yanqi Pan, Wen Xia, Darong Yang, Jiahao Chen, and Nan Zhang, Harbin Institute of Technology, Shenzhen
In virtualized environments, guest memory defragmentation is essential for exploiting huge-page benefits and improving application performance. However, existing approaches directly reuse host-side defragmentation and assume a limited guest physical address (GPA) space. As a result, they rely on memory compaction to defragment this constrained space, causing throughput to drop by up to 51% and latency to rise by up to 102% in YCSB-Redis workloads.
This paper proposes InfiniDefrag, a compaction-free memory defragmentation technique for virtualization. Our key insight is that the GPA space can be regarded as (nearly) infinite by controlling the mapping between GPA and host physical address (HPA), thereby eliminating the need for guest-side memory compaction. To realize this insight, we introduce (1) Infinite Address Manager that expands the GPA space while reclaiming free fragment pages; (2) Host Memory Guard that maintains the GPA–HPA mapping and constrains HPA usage within each VM’s quota; and (3) Scalability Optimizer that scales GPA/HPA space management to multi-thread and multi-VM environments. Experiments on micro-benchmarks and real-world applications show that InfiniDefrag outperforms state-of-the-art approaches and can achieve ideal, fragmentation-free performance.
OSDI '26 Open Access Sponsored by
King Abdullah University of Science and Technology (KAUST)
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

author = {Peixin Zeng and Hao Huang and Yanqi Pan and Wen Xia and Darong Yang and Jiahao Chen and Nan Zhang},
title = {{Compaction-Free} Memory Defragmentation for Virtualization via Infinite Guest Physical Address Space},
booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation (OSDI 26)},
year = {2026},
isbn = {978-1-939133-55-7},
address = {Seattle, WA},
pages = {1731--1747},
url = {https://www.usenix.org/conference/osdi26/presentation/zeng},
publisher = {USENIX Association},
month = jul
}