Testbed Federation funding for FOSS Projects

[onepict]

We discovered that grants like fed4fire+ have been set up to only allow Companies that are VAT/TVA registered to be eligible for funding. This means that non-incorporated FOSS projects and individuals aren’t eligible for funding testbed access.

Fed4Fire+ is also coming to the end of their Horizon funding. This means that in the future projects may have to pay for access to the labs.

Labs like Grid 5000 do offer some free accounts but time on these testbeds is limited and at the discretion of the GRID 5000 board.

Speaking to Brett Sheffield recently, I realized that there’s a need for FOSS projects to be able to access outside testbeds to test their code/app on a known testbed. This is important for the longevity of projects as testing on non-project infrastructure backed by known bodies can help with Project credibility.

Depending on the Lab, or the funding body, this can also help with project notoriety.

Brett also brought up an interesting point. Sometimes Projects and individuals need access to testbeds to check if an initial idea is viable before seeking further funding.

We need some form of germination fund for testing.

We also want to look to NLNET as an example of the fund setup, for projects to be eligible. NLNET can fund individuals as well as projects.

[dachary]

I’ve used Grid 5000 for the best part of 2021 (impressive quality of service and stability of the infrastructure) and also got funding from Fed4Fire+ (last round in september 2021) so I can confirm that it is critically important in some cases. Here are the concrete problems I see:

• a number of publicly available forges restrict or lack CI because the required infrastructure is too costly (https://forge.chapril.org/, https://codeberg.org/, etc.)
• some Free Software projects are very demanding when it comes to testing (https://ceph.com has a test lab that is roughly the size of grid5000 and running around the clock, GCC needs a compile farm with a machines of various architectures of https://cfarm.tetaneutral.net/, etc.)

To my knowledge each software project has to find a solution for themselves and when it requires funding they do no have, they have no organization to turn to and their only option is to reduce the level of testing which is harmful to the quality of the releases.

I however have no clue how that could be resolved

[onepict]

Yes, its a very tricky problem to solve. I know that we do need some high capacity testing. Grid 5000 is awesome, but the integration with jfed had some issues. I’d like to do more testing with Grid 5000 if possible.

High capacity testing, particularly with multicast wasn’t possible with jfed, as it connects with grid 5000 using global vlans. Which if you aren’t careful will knock out the backbone.

I know that GNU Taler also wanted a testbed with high scalability. Like thousands of nodes. Federation of testbeds could be the answer, but it does depend on how the CLI tools integrate together. I’m happy to email the feedback we had for fed4fire to anyone that is curious.

Because while projects often have very different needs, perhaps a testbed checklist/design for high scalability projects could be created as a starting reference point.

Of course the other part of testbed infrastructure for certain projects would be stuff like being able to set stuff like routers etc. Although then you really need design beyond the infrastructure, you need decent monitoring and orchestration tools so you can reset the lab between test runs and projects.