Reduce GVisor Cold Starts with GPU Snapshotting
jono_irwin
49 points
15 comments
July 01, 2026
Related Discussions
Found 5 related stories in 1310.4ms across 14,015 title embeddings via pgvector HNSW
- Reverse-engineering Nvidia's CUDA-checkpoint for faster cold starts ilreb · 18 pts · July 09, 2026 · 64% similar
- Cutting inference cold starts by 40x with LP, FUSE, C/R, and CUDA-checkpoint charles_irl · 79 pts · May 18, 2026 · 57% similar
- Popping the GPU Bubble radq · 36 pts · June 30, 2026 · 53% similar
- The 1979 Design Choice Breaking AI Workloads za_mike157 · 24 pts · March 09, 2026 · 53% similar
- Use your Nvidia GPU's VRAM as swap space on Linux tanelpoder · 227 pts · June 02, 2026 · 49% similar
Discussion Highlights (7 comments)
nixosbestos
Started scrolling, immediately closed the page. Something is deeply wrong with a person who chooses to implement this shit on a webpage. Unusable garbage, I'm sorry, literally making me motion sick somehow.
htrp
Isn't this exactly what modal does?
gpgn_
Interesting work. How does NVIDIA Dynamo Snapshot relate?
mountainriver
How does this compare to the CRIU work? Or does it use that under the hood?
Imustaskforhelp
Are there any open source solutions or is cerebrium open-sourcing the technology behind it. Although I liked the read-up (side note: you might have to tone down the animations, as others have said, it was a bit dizzy) but overall it was a nice read but I still wish for more technical details as Cold starts are something that just is something that I am interested in. So are there any more resources that perhaps the team could point out or other resources or if there are any idea of open-sourcing it ever for more internal deeper dives as I would love to know more about it!
eperot
Wrong headline order, right? Should read "Reduce GPU Cold Starts with gVisor Snapshotting".
keynha
The number that jumped out is 9GB restoring in 2.25s from S3 but 9s from local NVMe. I'd have bet on local, so the inversion is surprising.