Welcome

Ferranti Installation Update (18th April 2024): WekaFS fully deployed. Benchmarking on 19th April. GPU Direct enabled. CEPH deployed as well. Few cables need changing. LDAP deployment ongoing. Slurm Login nodes deployed.

The ML Cloud has archival storage (CEPH) in production. To learn more about how to use and move your files there, your quota, please go to the user-guide

Galvani Storage Update 18th April 2024: 50% of lustre nodes on racks. For the rest we don't have ports and cooling capacity. Benchmarking with lmdb, laion aesthetics, laion400M, and others. Current broken cooler in TTR2 data room prevents work.

User Guides

Access the Cluster

Get started by creating an account, logging in and using the system.

System Architecture

Lean about the ML Cloud Infrastructure

Containers and the ML Cloud

Learn how to use containers within the ML Cloud

Maintenance Schedule

Change: QB in R1 will be retired and substituted with >1PB NVMe storage

Date: Easter holiday 2024
Duration: Easter holiday (details to come)
Description: QB was an adventure. Now it is time to retire and bring our new NVMe based storage with over 1PB available capacity. Delivery is expected in february, we will install, test throughout march and during the easter holidays we will retire QB, move all data to the new storage and make the system available to the entire userbase.

Issue: Openstack Maintenance - Volumes, Snapshots, Backup Affected

Date: 25 August 2023 to 30 August 2023
Description: Since we are switching Openstack volume backend to Ceph in Region R1, there will be a down-time for the VMs with attached volumes that have not been detached, volumes, snapshots and backups. This is necessitated due to the preparation to retire QB within the next months.

Training and Events

ML Cloud Town Hall

Date: October 30st, 2024, 13:00-14:00PM
Description: Learn about the ML Cloud Infrastructure, Expansion Plans, Software Services. Ask questions about the ML Cloud of interest to you.

Profiling and Energy

Date: TBD
Description: Learn about profiling your code for optimization.

TBD

Date: TBD
Description: Multi-node training and optimization with H100