How much does it cost you to run your Julia experiments on AWS / Azure etc.?

I am in the process of scoping the budget for a project that I want to embark on which will involve large-scale 3D reconstruction from satellite images. I will need more than my desktop PC to handle the memory requirements and to the expedite computational time required to solve the large scale non-linear least squares “bundle adjustment” optimization problems associated with multiple-view image reconstruction.

I know that some of you (in particular Economists) regularly use Julia and cloud computing to conduct large-scale simulations for your experiments. I believe some of those experiments even run for a matter of days. I would be grateful if you could share some of your experiences with running Julia in the cloud, and in particular the ballpark costs associated with running your experiments.

Best wishes,

Zygmunt

1 Like

For me, Hetzner offers a great value with a variety of options, especially as an American because I’m exempt from the VAT.

They have dedicated servers starting at 34 euros per month with a recent i7 or Ryzen and 64GB RAM. Their cloud instances are also good because they charge by the hour and have a simple API for spinning them up with up to 16 virtual cores and 32GB of RAM per instance.

2 Likes

imho:

European experience:

  • Hetzner dedicated is OK ; but need an extra month payment for as a frontend … ( ZEN2 and Intel cores ; full machine) … Not ideal for high parallel ( 50 machine ) + short time experiment ( 3 days )
  • Hetzner Cloud - cheap ; but shared CPU … so not perfect for high CPU works ( my experience )

Now for personal testing I am using the extreme cheap Contabo ( Europe )

  • https://contabo.com/?show=vps 15 EUR/month 30GB ; 800GB SSD; 8cores (" VPS L SSD")
    ( my preference was a high SSD ; high RAM for the data processing )

And you have to compare/check the value with the https://www.scaleway.com/en/ they improve the service very fast pace …

10 Likes

Thank you both for the fantastic advice. I really appreciate you taking the time.