TIP v3 - osparc-devteam-tasks

[SCA-ZMT]

Tasks

Preview Give feedback

1. make only 1 unit default at a time, all others need to be undefaulted #1439
   +0
   
2. enhancement: optimize docker image size #1623
   +0
   
3. Enhancement: Lock a EC2 machine for a specific project osparc-simcore#5669
   a:autoscaling a:director-v2 a:webserver +3
   
4. enhancement: optimize pricing plans @sanderegg @mguidon #1516
   +0
   
5. Optional: Migrade deployment to different AWS account #1548
   +0
   

For the applet-developers tasks, see: #1155

Eisbock

Preview Give feedback

1. request functionalities of osparc python api client for ti-simu #1611
   7 of 14
   +0
   

   
2. Use initialized EBS storage instead of buffer machines osparc-simcore#5864
   17 of 17
   a:autoscaling a:infra+ops +2
   
3. tip.science: Migrate data from tip.itis.swiss osparc-ops-environments#632
   blocked / paused p:high-prio t:infra-ops +3
   
4. Deprecate old services in the tip.itis.swiss DB #1549
   +0
   
5. tip-product on aws-master and aws-staging #1624
   4 of 4
   +0
   
6. Show time spent on a service in the Credits Summary view #1622
   2 of 2
   HARD LAUNCH a:frontend s4l:web type:enhancement +4
   
7. Update osparc-client version in all TIP services #1618
   +0
   

   
8. Update s4l based services #1450
   1 of 1
   +0
   
9. Release osparc client #1517
   +0
   
10. ✨ Enable tip.science v1 and v2 dynamic and comp autoscaling osparc-ops-environments#645
    8 of 8
    blocked / paused p:high-prio +2
    

Tasks

Preview Give feedback

No tasks being tracked yet.

[mguidon]

Resource Requirements:

• The pipeline consist of 5 dynamic nodes: electrode selector, personalizer, simulator, tip-postpro, s4l-postpro
• All services should be based on the same jupyter-smash version (good for caching) and simphony base image
• personalizer, tip-postpro and s4l-postpro require a GPU for rendering
• None of the services can fit on the currently buffered machine types
• The simulator will execute one remote job on a high RAM instance

Based on this requirements, the tool will be quite expensive and slow. I see two options:

• we prepare one node with all the software installed (@pcrespov we discussed this over coffee, see Architecture: One machine per User with Apps osparc-simcore#5593)
• we create a proper all-in-one service with a dedicated UI (based on s4l-postpro)

[mguidon]

• "Fix" license server IP address
• Create s4l-python runner with iSolve
• a fake or real billing center?

[drniiken]

[GitHK]

20240507 TIP - > AWS Meeting
Attendees: Niels, Nik, Melanie, Dustin, Manuel, Odei, Matus

Meeting Minutes

• In the new tip.science UI:
  • The credits are the only difference compared to the current tip.itis.swiss. (perception to
    the user).
  • We need to play through and refine asap the billable scenario (free + payable tiers).
  • Startup time might be a bit slower.
• Melanie is currently testing the integration for V3 o The new pipeline consists of five services (machines).
  • Release is envisioned for "end of May" (V3, personalization, AWS).
  • Demo (today) showed still some details/issues to address. Ideally we have a 1st/beta
    version very soon to play around with and refine/iterate.
  • UX still needs various refinements.
• Decision:
  • We will pay for everything except the simulations.
  • Cost on us: about 60k pa.
  • We might be able to save some more cost by "choosing/optimizing" the right machines.
  • Personalization: simulation cost is charged to the users.
• Billing will likely be set up to have a zero burn rate for all services, with only the simulation being billable.
  • Find a good communication, perception of the "billing aspect" for the users.
• Decision:
  • We will not spend one+ month of development time migrating TIP to sim4lie.io deployment

[mrnicegyu11]

Linking related cases:

• various tip improvements TIP Improvements #1345
• tip v3 TIP v3.0 #1155

[pcrespov]

MartinKippenberger Sprint

• Improvements on the osparc client (used in TIP)
  • enhanced documentation and learning resources
  • improved release workflow (pre-release available for developers)
  • released new API features (v0.7)
  • new enhanced osparc client including pydantic models(coming soon)