Singularity and Firecracker

[John Hearns]

I may have asked this before.. blame advanced idiocy.

I missed FOSDEM this year and Eaduardos talks (damn).

Firecracker looks very interesting also - is anyone doing work with Singularity and Firecracker? There is a GO API and everyfing:

https://github.com/firecracker-microvm/firecracker-go-sdk

https://fosdem.org/2019/schedule/event/containers_firecracker/attachments/slides/3188/export/events/attachments/containers_firecracker/slides/3188/Firecracker_as_a_container_runtime_FOSDEM2019_4_3.pdf

[v]

What is the use case for using firecracker (with integration for singularity?) If the ultimate goal is to run reproducible containers, adding another layer of abstraction doesn't get closer to that, but might in fact get farther from it if there is some requirement to re dump container contents in yet another rootfs. I haven't looked at it in detail, but I'd be interested to hear your thoughts!

I didn't go to FOSDEM either, maybe some day :)

--
You received this message because you are subscribed to the Google Groups "singularity" group.
To unsubscribe from this group and stop receiving emails from it, send an email to singularity...@lbl.gov.

--

Vanessa Villamia Sochat
Stanford University '16

(603) 321-0676

[John Hearns]

Good question! I guess the real answer is that Firecracker is new and shiny.

Serious answer - Firecracker is billed as micro VMs which can be spun up very quickly. So you could imagine starting up a massively parallel cluster very quickly, then collapsing it when your job ends. I guess you will ask then what the advantage of running containers on those micro VMs is.

Also I just note from the GitHub page a quote about security. Singularity folks have that one covered of course.

Can Firecracker be used with Kubernetes, Docker, or Kata containers today? Not yet. We are making Firecracker open source because it provides a meaningfully different approach to security for running containers.

[v]

Interesting! Thanks for the info. I heard of Firecracker, but as an Amazon product and my takeaway was that it was the AWS effort to get into the space (via an open source project). It also seems to be scoped for "serverless" pay as you go as opposed to a research cluster (SLURM, etc.)

I'm all for getting excited about new things! So much, in fact, that I've learned to step back and ask if said new thing is really badly needed. The answer to that question, of course, depends on the communit(ies) you put your hat on to think for, and things that you want to do that aren't afforded by existing technologies.

If anyone else has cool learnings from FOSDEM, please share on the list!