Model the connection between worker nodes and storage

[knikolla]

Motivation

There's currently no comprehensive diagram containing all the hops in the data path between OCP worker nodes and storage. Furthermore, different worker nodes of different types are in different locations.

Completion Criteria

Completed model of network connection between worker nodes and storage

Description

• Ask for necessary diagrams
• Receive all necessary diagrams
• Decide on best format to represent
• Model

Completion dates

Desired - 2024-10-02
Required - TBD

[joachimweyl]

@jtriley please provide the diagrams.

[jtriley]

I sent in the following ticket requesting the info:

Folks working on the NERC project would like to get some information about the connectivity between racks 1-C-20 (OCP-PROD worker nodes), 4-C-6 (OCP-PROD V100 GPU nodes) and the 10x NESE racks. If possible it'd be nice to know bandwidth between switches, switch models, etc. and what other kind of traffic goes through the same "pipes".
In general we're looking to get enough detail to model the flow and bottlenecks between these racks.

That said, the connectivity from the A100 racks will need to come from @hakasapl and @naved001 given that the networking is managed by MOC for that rack IIRC.

[knikolla]

@jtriley thank you. I do have the diagrams for the MOC side of things and will put everything together once I get this too.