[Performance] Speed up Azure A10 instance creation

skypilot-org / skypilot

SkyPilot: Run AI and batch jobs on any infra (Kubernetes or 12+ clouds). Get unified execution, cost savings, and high GPU availability via a simple interface.

https://skypilot.readthedocs.io

Apache License 2.0

6.82k stars 514 forks source link

[Performance] Speed up Azure A10 instance creation #4205

Closed yika-luo closed 3 weeks ago

yika-luo commented 3 weeks ago

Use the new custom image for Azure's A10 VM creation.

Example performance change (with fixed region):	VM Type 💻	Old Provision 🕐	New Provision 🕐	% Speedup ✅
Azure GPU	21min	2min 50s	85% (7x)

Tested (run the relevant ones):

[x] Code formatting: bash format.sh
[x] Manual test: time sky launch --cloud=azure -y --gpus=A10:1 and ssh to machine to confirm NVIDIA and nvidia-docker runtime installation

Michaelvll commented 3 weeks ago

A future TODO: we need to add a smoke test for A10 on Azure