Show HN: NStack – Typed, composable microservices for data analytics (opens in new tab)

(github.com)

30 pointsNickPollard8y ago5 comments

5 comments

pea8y ago

Hey guys, I'm Leo - one of the founders of NStack.

We started work on NStack around a year ago, to help data-scientists productionise their models and compose them together with various data sources.

We built NStack using a mixture of low-level Linux technology, including systemd-containers, btrfs, dbus IPC, along with high-level concepts such as building a DSL, type-checker, and runtime. As a functional programming company, NStack is built using Haskell and are really excited to launch our CLI and server -- along with a sandboxed demo PaaS which you can try out now.

We'd love to hear any feedback.

starquake8y ago

Did I miss something or is there no license mentioned anywhere? Would love to know before looking into code.

EDIT: Sorry, it's in the subfolders.

adam_ellsworth8y ago

What are some good resources on better understanding composability?

NickPollardOP8y ago

We have some info on our docs here: https://docs.nstack.com/en/latest/features/index.html (See the composition section at the top).

Composition is a simple technique to build functions out of multiple simpler functions. It's heavily used in Functional programming but we think it's really useful for everyone, particularly data workers who might be gluing together lots of different modules/libraries/systems. We also use types to help check that functions can compose - this can catch schema mismatch errors early rather than waiting until you run your data pipeline to realise that the output of one function isn't _quite_ what your second function expects.

Wikipedia as some more detailed info on function composition: https://en.wikipedia.org/wiki/Function_composition_(computer...

d--b8y ago

This looks great. It seems like it's a great open source alternative to all these big software vendors like dataiku.

kudos.

j / k navigate · click thread line to collapse