Turing membership of the International Binder Federation
Kirstie Whitaker
Yes - turing-mybinder
$25,000
A BinderHub (https://binderhub.readthedocs.io) is a computational infrastructure that requires resources which are readily available as a service on the Azure cloud platform (see the "Computational Requirements" section for a detailed description). A different computing platform would likely not have these services in place.
mybinder.org is already hosted by four cloud platforms: Microsoft Azure, Google Cloud, OVH.com (a European cloud provider), and an on-prem service GESIS Notebooks hosted by the Leibniz Institute. This diversity of hosts consolidates Project Binder's status as a cloud-neutral, open source project. We are propagating knowledge related to maintaining a BinderHub on Azure into the wider Binder community and we are contributing to a sustainable and resilient open infrastructure by decentralising resources and knowledge throughout the Federation.
The Binder Federation (https://blog.jupyter.org/the-international-binder-federation-4f6235c1537e) is an experiment in open infrastructure for data science and beyond. It is a community of scientists and open source developers dedicated to scientific reproducibility, communicating ideas through interactive computing, and providing real-time access to results and analyses. The Federation running mybinder.org has grown to hosting ~140k user sessions per week which range in purpose from research, to teaching and workshops, and beyond. Continuing to be a member of the Federation promotes all three of the Institute's goals: advancing world-class research on real-world problems in academic, government and industry settings; providing interactive training materials for the next generation of data science leaders; and leading the public conversation on modular, interoperable, community owned and run infrastructure. Our shared goal is to democratise access to artificial intelligence for all.
This increase to $25k in Azure credits will allow the Turing to scale up the proportion of mybinder.org we support and facilitate continuous contributions to the improvement, reliability and promotion of the service.
Unchanged from previous setup.
The following breakdown was estimated using the Azure Pricing Calculator.
Service type | Region | Description | Estimated monthly cost per unit ($) | Number of units | Estimated monthly cost ($) | Estimated allocation cost ($) |
---|---|---|---|---|---|---|
Azure Kubernetes Service (AKS) | West Europe | D4s (4 vCPU, 16 GB RAM) nodes | 167.90 | 10 | 1,679.00 | 21,827.00 |
Virtual Network | 500 GB data transfer within West Europe region | 10.00 | 1 | 10.00 | 130.00 | |
Azure DNS | West Europe | 1 hosted DNS zone | 0.90 | 1 | 0.90 | 11.70 |
Key Vault | West Europe | 0.00 | 1 | 0.00 | 0.00 | |
Subtotal | 21,968.70 | |||||
Additional budget for autoscaling and testing | 13.8% | 3,031.30 | ||||
Total | 25,000.00 |
Request for clarity from RCWG:
What is the plan for sustainability? What happens after the $25k of Azure credits has been used up? How will the contributions and improvements to the mybinder.org be sustained (without coming back for more Azure credits)?
Response:
To clarify, without these credits we cannot continue to run the infrastructure set up as part of the Binder Federation. I think the important nuance that may have been lost in the application is that the Turing's donation of these credits/infrastructure is acknowledged on the mybinder.org homepage and the Federation page (https://binderhub.readthedocs.io/en/latest/federation/federation.html#members-of-the-binderhub-federation) which in turn grants us certain privileges both internally and externally.
Externally, the Turing will be seen as a major contributor to open infrastructure, reproducible research and decentralisation of knowledge. We also gain ourself a seat at the table in terms of involvement in decision-making regarding the roadmap of mybinder.org. Internally, we can learn skills from an established community on large-scale project management and workflows for reproducible research that we can leverage to benefit other Turing projects.
Simon Li You've already done loads of work promoting Turing, the goals of Turing, and mybinder, through running workshops, hanging out on forums, etc. Mention this? It's evidence you've already made good use of the previous proposal and you're going to build on it! I don't know how Turing is funded, but if it's UK research councils they're always keen on public outreach so this would help them get future funding? Sarah Our Azure credits are a gift from Microsoft and I'm not sure that the research council model of "doing outreach things to get money" translates unfortunately :-/ The whole reason that there's a process for applying for credits is because we're not sure if/when Microsoft will renew that gift (and we got that gift because a previous Turing CEO came from Microsoft Research) Simon Li Might be worth stating the proportion of the 140k weekly sessions hosted by turing? This both emphasizes the contribution of Turing to the federation, and also shows the costs are shared amongst all members. Sarah I think our pod quota was only 80 which doesn't scream "major contributor"
I'd be more comfortable with a "we've completed the proof-of-concept run, now lets go big league" approach if we're going to explicitly mention proportions Simon Li Do you have any concrete examples of Turing Institute users using mybinder in their research or for running their workshops? Sarah NoImage Not Showing Possible ReasonsLearn More →
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Though a colleague does want to publish his paper results in a BinderImage Not Showing Possible ReasonsLearn More →
- The image file may be corrupted
- The server hosting the image is unavailable
- The image path is incorrect
- The image format is not supported
Since Kirstie and I are core contributors to Project Binder, we can still support the sustainability of mybinder.org in terms of day-to-day operations of the Google cluster (which is also currently trying to find credits for the next year!) and leadership roles within the project. However, I don't think that's as strong a show of support from the Turing as donating computational resources would be.
Simon Li Could go further, and emphasize by joining the federation Turing gets a strong influence on the future development on mybinder and JupyterHub, and this closely aligns with <some stuff related to what else Turing is doing>. Ideally mention something that turing will have to spend money on, that can be partially solved by binderhub/jupyter, and in it's absence they'd have to spend money on some other solution anyway?
I hope this helps answer your questions.