I was looking into some of this myself already; really helpful article and amazing results. Another interesting approach is to compile Haskell with HaLVM[1] in to a unikernel that can run on a Xen hypervisor. But this is a non-posix environment on which Haskell's "network" package does not compile. No "network" means that it is currently impossible[2] for WAI (Haskell's standard interface between web servers and applications), but also db libraries, to run in such an environment.
With going unikernel with Haskell still being a little steep (but definitely on my wish list), then the next-best would be a mini VM/container image. And 5MB sure is mini!
Yeah it's a very complete project. And we saw "Look Ma No OS"[1] today on HN about Erlang/Elixir-on-Xen with the LING unikernel. But it is Haskell that I'm personally all hyped up about lately; and OS-less-ness is to me merely a potential secondary advantage of going back to "compiled" (I come from Ruby/Python before that Java before that C++).
My main reason for going with Haskell is that I know the costs of bugs, and feel that I get to old for fighting bugs on tight deadlines.
Well, I'd suggest that OCaml probably has similar bug fighting properties as Haskell. It depends a lot on how much you enjoy writing module existence proofs :)
If I had any genuine reason for a unikernel today I'd definitely go for Mirage.
Comments
I was looking into some of this myself already; really helpful article and amazing results. Another interesting approach is to compile Haskell with HaLVM[1] in to a unikernel that can run on a Xen hypervisor. But this is a non-posix environment on which Haskell's "network" package does not compile. No "network" means that it is currently impossible[2] for WAI (Haskell's standard interface between web servers and applications), but also db libraries, to run in such an environment.
With going unikernel with Haskell still being a little steep (but definitely on my wish list), then the next-best would be a mini VM/container image. And 5MB sure is mini!
Once again, thanks FPComplete!
1: https://github.com/GaloisInc/HaLVM
2: https://github.com/GaloisInc/HaLVM/issues/43
OCaml's Mirage unikernel seems to have a lot of steam these days.
Yeah it's a very complete project. And we saw "Look Ma No OS"[1] today on HN about Erlang/Elixir-on-Xen with the LING unikernel. But it is Haskell that I'm personally all hyped up about lately; and OS-less-ness is to me merely a potential secondary advantage of going back to "compiled" (I come from Ruby/Python before that Java before that C++).
My main reason for going with Haskell is that I know the costs of bugs, and feel that I get to old for fighting bugs on tight deadlines.
1: http://slides.com/technolo-g/intro-to-unikernels-and-erlang-...
Well, I'd suggest that OCaml probably has similar bug fighting properties as Haskell. It depends a lot on how much you enjoy writing module existence proofs :)
If I had any genuine reason for a unikernel today I'd definitely go for Mirage.