from Hacker News

Dstack: An alternative to k8s for AI/ML tasks

by shcheklein on 11/5/24, 4:56 PM with 12 comments

  • by cheptsov on 11/5/24, 5:23 PM

    Oh, excited to see dstack featured here. Founder and core contributor to dstack here. Yes, we aim to simplify container orchestration for AI and build an alternative to both K8S and Slurm - for both multi-cloud and on-prem.

    Would love to hear feedback!

  • by whinvik on 11/5/24, 8:26 PM

    Struggling to compare it to k8s but maybe that's just my lack of knowledge on k8s.

    However, to the general HN crowd my question is what would be the correct abstraction for something that wants to replace k8s to make one think that this new abstraction is simple enough.

    In my naive understanding, I would think that if I don't have to think beyond a docker-compose.yml that would be the right level of simple. To clarify, locally I work with the docker compose file to bring up my services, and I should be able to just deploy it to AWS for example.

  • by htrp on 11/5/24, 7:34 PM

    >dstack is a streamlined alternative to Kubernetes and Slurm, specifically designed for AI. It simplifies container orchestration for AI workloads both in the cloud and on-prem, speeding up the development, training, and deployment of AI models.

    Any explanation for how it simplifies orchestration.

    >dstack supports NVIDIA GPU, AMD GPU, and Google Cloud TPU out of the box.

    Any plans for trainium ?

  • by cassianoleal on 11/5/24, 6:01 PM

    It's k8s, not k8: k + 8 letters + s

  • by Muhtasham on 11/5/24, 6:26 PM

    love the lambda integration with dstack