Unconference Session: Checkpoint/Restore BoF

on YouTube ↗35m

" summary

Working group prototypes an API that supports three concrete cases: suspend/resume, live migration, and warm start (prewarmed JVM/GPU images). Implementations target CRIU and gVisor; checkpoints are file archives possibly stored via CSI or a registry artifact and include kernel + GPU memory. Design uses a separate Checkpoint object and controller plus a pod-side restoreFrom field, plans core API (not CRD), and raises practical issues: storage transport, device plugins, node compatibility checks, security/permission boundaries, and coordinated group checkpoints.

AI-written summary · about

" chapters (6)

1. 0:00Checkpoint/Restore BoF intro
2. 1:48Use cases and scenarios
3. 3:16Open questions in SIGs
4. 6:45API design approach
5. 9:14Community discussion
6. 34:06Closing remarks

" related talks

• 15mUniversal Checkpointing: A Flexible and Efficient Distributed Checkpointing...Masahiro Tanaka, Olatunji Ruwase, Minjia Zhang, Xinyu Lian, Sam Ade Jacobs, Lev Kurilenko, Stas Bekman · USENIX ATC '25
• 15mDistributed CheckpointIris Zhang, Chien-Chin Huang · PyTorch Conference 2023
• 10mLightning Talk: In-Cluster Distributed Checkpointing: Optimizing Training...S. Mishra, G. Kroiz · PyTorch Conference 2025
• 17mCheckmate: Zero Performance Overhead Model Checkpointing via Network Gradient ReplicationManya Ghobadi, Adam Belay, Ankit Bhardwaj, Weiyang Wang, Jeremy Carin · NSDI '26
• 28mRundown of Flink's Checkpointsdavid visikovic · Flink Forward Global 2021
• 39mProcess Migration in the Orchestration WorldIsabel Jimenez, Kapil Arya · LinuxCon + ContainerCon North America 2016
• 55mGoing ResilientAnatole Tresch · Voxxed Days Vienna 2016
• 19mProtocol-Aware Recovery for Consensus-Based StorageEric Lee, Aishwarya Ganesan, Aws Albarghouthi, Ramnatthan Alagappan, Vijay Chidambaram, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau · USENIX ATC '19