BMTK 2022

[danielsf]

Anton Arkhipov and Kael Dai will be hosting a modeling workshop on July 25 and 26. They need us to stand up an AWS environment to allow the participants to run the relevant modeling software.

Below are notes from a conversation @danielsf had with them on 4/8/2022 (see also the issue #2266 which Wayne opened to encapsulate this work).

Dates
July 25 and 26

Software users must have access to*

• BMTK
• VND
• Sonata
• NEST
• If users are given access to multiple CPUs (see below), the environment will need to have access to MPI.

Data users must have access to*

• There are not datasets associated with this workshop. It is a modeling workshop, so there will be no need to interact with AllenInstitute data resources.

Hardware requirements

• Every user will need access to a GPU.
• Ideally, every user will need access to many (dozens?) CPUs. The exact number will depend on the cost estimate.
• The software can be made to work with one GPU and one CPU, but it will be slow.

User Interface requirements

• Users will need to be able to run notebooks and a CLI
• JupyterLab will probably be sufficient for these purposes (?)

Initial tasks

• Pika will estimate the cost for various configurations of hardware (deliver estimate by early May)
• Kael can provide us with a Dockerfile that installs the required environment (should be ready by 5/16)
• Once Kael has given Pika the Dockerfile, Pika will spend a week (?) building a prototype AWS environment. Once we have something that users can log on to, we will hand it off to Anton, Kael, and their collaborators at the University of Illinois to test so that they can determine if it meets their needs.