2020-11-25 Benchmarking for Industry Partners Meeting notes

Public molecule submission for QCArchive calculations

Gary Tresadern

• GT: We have requested partners give us molecules for the public test of the protocol and get those to Openff team soon around 1500 each max. Think about interesting cases and chemistry this will be public and please start sharing soon. We will then begin to run this as soon as we can.

• AG: was the file I sent okay?
  yes but it was small do you have more, will look for more as they are struggling to workout what is already in the public domain.

• A few partners are waiting on final say on molecule selections.

Demo of the progress so far on the protocol

David Dotson

• Link to survey: https://docs.google.com/forms/d/e/1FAIpQLSe-DP0J10-DD7XWJWx02zNGdcLwYASU-x-W_XGxfx3A9E_glQ/viewform

• DD: Please fill it out as detailed as possible

• DD will email out the form as some users can not access google forms

In-house calculations/security issues

• PV: There might be issues setting up QCArchive server for in-house calculations (IT sees it as a security issue).

  • DD: Concern of long-running service that talks to the network?

  • PV: You need to open ports, IT sees it as a security risk.

  • TF: BI might also have this issue

  • PV: Is it possible to wrap it in bash scrips?

Presentation of project plan

David Dotson

• Geometry Optimization Benchmarking for Industry Partners

• We are on track for testing at Janssen from Dec. 1

• More detailed plan in software approach: Benchmarking Software Approach

• 4 developers are working on this, you can follow the activities here:

Demo of CLI

David Dotson

• DD: Walks through the layout of the CLI starting with starting up a fractal server and managers using the CLI points we have built. Explains how each of the components work.

• We could have the server stand up in a submitted job so everything is ran locally within the job no outgoing connections, this would work with users that can not have a long standing QCArchive instance.

• DD: Explains molecule identifier and how to submit molecules to the local server.

• AG: The manager is controlling the tasks? - DD: Yes

• AG: Do the jobs have to come online on time? - DD: The manager handles submission of jobs. That requires the manager to be online on the head node for a while

• DD: we could submit the manager into the queue to generate new jobs rather than stand up the job on the head node

• PV: That would not work for us. Easiest would be to have a list of molecules and submit them to the queue. The presented cli looks like you have to do a lot before even one molecule is calculated.

• DD: We are trying to “free” the user from tasks as much as possible, as well as giving us the possibility to help. If there is manual work, we cannot help that much if something goes wrong.

• PV: I would not mind a low-level submission approach. Some scripting is ok for me

• DD: Thanks for letting us know. Let me think about this. There are some challenges. We try to find a solution which works for everybody.

Anything else?

All

• AG: Can we start working on / trying out something?
  DD: No, I would wait. We will share the workflow as soon as ready.