Stretch: A Fault-Driven DSM Runtime for Distributed Multithreaded Applications

on YouTube ↗25m

" summary

Stretch is a userspace, page-granularity DSM runtime that uses userfaultfd-based fault interception, centralized MSI-style coherence, and CRIU-driven thread placement to extend a process across machines. Evaluation on four CloudLab servers shows compute-heavy workloads (Matrix Multiply) gain 3.39× speedup, while fine-grained write sharing (KMeans) triggers invalidation storms and collapses performance. RDMA cuts key coherence operations by up to 4× versus TCP but loses its edge when the central server becomes CPU-bound.

AI-written summary · about

" chapters (9)

1. 0:00Introduction
2. 3:30Linux primitives for DSM
3. 7:04Stretch architecture
4. 12:23Page faults and sync
5. 15:11Performance setup
6. 16:57Homogeneous scaling
7. 19:03Heterogeneous performance
8. 20:05Key takeaways
9. 21:13Q&A

" related talks

• 23mTeRM: Extending RDMA-Attached Memory with SSDJiwu Shu, Zhe Yang, Youyou Lu, Qing Wang, Xiaojian Liao, Keji Huang · 22nd USENIX Conference on File and Storage Technologies (FAST '24)
• 25mElastic Scaling of Stateful Network FunctionsScott Shenker, Sylvia Ratnasamy, Justine Sherry, Shinae Woo, Sangjin Han, Sue Moon · NSDI '18
• 16mDRust: Language-Guided Distributed Shared Memory with Fine Granularity, Full...Shan Yu, Miryung Kim, Yifan Qiao, Haoran Ma, Shi Liu, Yiying Zhang, Qingda Lu, Jiesheng Wu, Harry Xu, Yuanjiang Ni · 18th USENIX Symposium on Operating Systems Design and Implementation (OSDI '24)
• 17mDuhu: Shared Disaggregated Memory for Distributed Data Processing FrameworksAurojit Panda, Jongryool Kim, Tao Wang, Qiutong Men, Hane (Stella) Yie, Emmanuel Amaro, Marcos K. Aguilera · 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26)
• 39mTrust Deterministic Execution to Scale & Simplify Your SystemsFrank Yu · YOW! Perth 2023
• 16mPreview: Cloud ComputingKshiteej Mahajan · OSDI '21 / USENIX ATC '21
• 32mStretch Clusters in Ceph: Algorithms, Use Cases, and ImprovementsKamoltat (Junior) Sirivadhna, Greg Farnum · Cephalocon 2023
• 15mCollie: Finding Performance Anomalies in RDMA SubsystemsZhuo Jiang, Jianxi Ye, Chuanxiong Guo, Xinhao Kong, Danyang Zhuo, Yibo Zhu, Huaping Zhou · 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22)