nimbo VS criu-image-streamer

You should really mention / give attribution / emphasize more that this is a fork of https://spotty.cloud and you took a lot from https://github.com/nimbo-sh/nimbo as well.

Hey everyone,

I (Miguel) am an ML PhD from the University of Edinburgh and Juozas is a Software Engineer also from Edinburgh.

Together we developed Nimbo, a dead-simple CLI that wraps the AWS CLI, allowing you to run code on AWS as if you were running it locally. You can find the source code here (https://github.com/nimbo-sh/nimbo) and the docs here (https://docs.nimbo.sh).

We decided to build this because we were frustrated with how cumbersome using AWS was, and we just wanted to be able to run jobs on AWS as easily as we run them locally. At the same time, we wanted to make use of the cheap spot instances (on Nimbo, this is a single parameter). All in all, we didn't like the current user experience of working with AWS, and we believed it was possible to vastly improve it.

For this reason, we also provide many useful commands to make it faster and easier to work with AWS, such as launching notebooks on EC2, easily checking prices, logging onto an instance, or syncing data to/from S3 (you can see some useful commands at https://docs.nimbo.sh/useful-commands).

Unlike other similar services, we are solely client-side, meaning that the code runs on your EC2 instances and data is stored in your S3 buckets (we don't have a server; all the infrastructure orchestration happens in the Nimbo package). We are also open contribution, meaning that all the source code is publicly available on our GitHub, and we welcome community contribution.

We have tons of ideas for Nimbo, like adding docker support, and providing instances with preloaded datasets like ImageNet, so that you don't have to download and store it yourself - you simply spin the instance, and the dataset is available at /datasets. We are currently working on adding GCP support, so that you can use AWS or GCP with the same config file.

We are happy to receive any feedback and suggestions you have.

My friend and I just launched Nimbo, a dead-simple CLI that wraps AWS CLI, allowing you to run code on AWS as if you were running it locally. GitHub: https://github.com/nimbo-sh/nimbo. Docs: https://docs.nimbo.sh.

Cool yeah that makes sense, makes total sense for ML where you just need to run over epochs, less clear for other workloads.

After looking around I thinking more about CRIU/docker suspend. The google stars aligned and I found this https://github.com/checkpoint-restore/criu-image-streamer + https://linuxplumbersconf.org/event/7/contributions/641/atta... which actually seems perfect. I wonder how fast it is

(or, hacking on a checkpoint idea, have a daemon periodically 'checkpoint' other programs so even if it's too slow over 60 seconds, revert to the last checkpoint. Even an rsync like application where only send the changes)