JupyterHub and BinderHub Team Meeting - July¶
Goal: Continuous improvement
Date: 19th July 2018¶
Welcome to the Team Meeting¶
If you are joining the team video meeting, sign in below so we know who was here. Roll call:
name / institution / GitHub handle
min / simula / @minrk
tim / wild tree tech / @betatim
erik / sandvik code / @consideRatio
j forde / @jzf2101
Brian / Cal Poly / ellisonbg
Chris / UC Berkeley / @choldgraf
Help or discussion needed; Agenda Items for Monthly Meeting 1.1. Open PRs 2. #758, node pools, k8s 1.9+, scheduler, etc… 3.2. Open issues 4. last call for camera version of icml workshop writeup plz sign up on openreview http://openreview.net/
what if I was to receive $20000 (or more) in cloud credits, how could they go to running mybinder.org?
would require juggling billing accounts
a reason to get started with federation so we can handle this under the hood?
reasons not to take credits on other platforms is that it adds complexity that costs human effort
jzf says related: grant opportunities for additional funding
add a public demo of binderhub with more resources to convince people we can do more to encourage them to deploy their own (fancy.mybinder.org)
maybe do impressive demos like pangeo where users can spin up big clusters
Jessica submits a proposal for AWS credits in the next few days for the demo of fancy.mybinder.org
“ai.mybinder.org” in order to host fancier compute and highlight machine learning use-cases
can this grant also include some engineering time from SageMaker/AWS team?
geo.mybinder.org, yourdiscipline.mybinder.org as a way to expand this idea
create a call for hosting mybinder.org, wait till after AWS grant
if you are against this write here
like this idea of a call, but we might want to make it non-exclusive, to allow multiple vendors to participate.
how do we try out https://cloud.google.com/compute/docs/instances/preemptible to see how disruptive it would be to users and how much we would end up saving?
Tim will create an issue to follow this up
One idea: split users between two nodepools once we have the ability nodepool
Wild Tree Tech will start saying on their webpage that we offer hosted/managed JupyterHub and BinderHub instances
I’d like to get some momentum behind the “binder sustainability” conversation. Or at least to recap what we discussed @ the dev meeting.
Update on the scipy presentation: http://bit.ly/scipy-2018-binder
start using the SciPy paper once it is out instead of the eLife blog post?
what about the JOSS paper that is stalled? Waiting for 0.1
@jzf2101 should give an update on the reproducible environments paper she presented!
if you want your name on this sign up for openreview.org
Update on the PEARC conference tutorial that we’ll be giving in pittsburgh
link to agenda: https://pearc18.conference-program.com/?page_id=10&id=tut149&sess=sess103
workshop will focus on using it on Jetstream and ComputeCanada instead of commercial clouds
Cost model + projections for our $$$ burn rate
we aren’t sure how long we have before we run out of money
we give 1GB of RAM at least, 10min idle time out, max run time 8h, scaling is determined by RAM required not CPU needs
we have been using the wrong logo all along!!!11!
we should try adopting the multi-node-pool + nodeAffinity deployment strategy, which may help a lot with scale-down (harder to say without packing for Binder’s round-the-clock visitors, but works great for users in a single timezone)
ERIK: This is what #758 is about, in an optional but recommended manner.
Zach Sailer will be starting full time at Cal Poly around August 6th.
He will be working mostly full time on JupyterHub things, starting with better integration with directory services (AD, LDAP, etc.) and the idea of “projects” for JupyterHub.
Next actions (team)
News, Information, and Thanks¶
Team News and Informational items
5.1 Organization highlights
5.2 BinderHub projects
https://gitlab.com/gitlab-org/gitlab-ce/issues/48236 Tim isn’t sure what exactly they are planning but I made a suggestion for what I think would be cool.
5.3 JupyterHub projects
5.4 Related projects (repo2docker, nbgrader, others)
Thanks, Things to Celebrate, and anything else
We got some shout outs in the CERN High School teacher summer school https://indico.cern.ch/event/651996/contributions/2994003/attachments/1689835/2718654/WG5_Open_Data_Presentation.pdf and their repo https://github.com/cms-opendata-education/HST-2018
Tim will be speaking about Binder at PyconDE in October