V1.5.0 Release Plan

[yukirora]

Endgame

• Code freeze: Dec. 26
• Bug Bash date:
• Testing start date:
• Release date:

Main Features

Platform Features & Stability

1. • LTP rest-server data auth (Ruigao/list disappeared vc jobs #112 )
2. • LTP prometheus data auth
3. • Containers losing access to NVIDIA GPUs
4. • local storage ssh key gen consistency enhancement
5. • Job exporter support on arm64 architecture GPUs

Inference

1. • statistic about inference usage
2. • Inference user data auth design and implementation

Security

1. • Add support of Azure email communication service

TODO

Automatic Endpoint Deployment/Upgrade CI/CD
Distributed prometheus design
Job Reliability & Monitoring
Automatic Failure Detector
Interactive gracefully job exit/retry
Hardware Failure Detector
Software Failure Detector
Cluster and Job utilization metrics
LTP-Megatron Support for Project Level Metrics
Project Level Metrics Monitoring