Skip to content

Latest commit

 

History

History
12 lines (7 loc) · 1.82 KB

realization_4.md

File metadata and controls

12 lines (7 loc) · 1.82 KB

Realization of a shitty project, using awesome tech (Part IV)

Here I am, another update, nothing has been done (almost nothing, we actually started some experiments with the OpenCV libraries to implement the hand scoring service/Android app).

Lately I spend too much time thinking an end up doing everything backwards. So this time, while looking at my "deployment requirement list" I found out that among the docker hosting providers I was looking at, DotCloud bills by allocated memory at 32MB blocks. And today this tweet happened to be in my feed (just to be clear, I do not think it's shitty advice). Not that matters, but I felt that if we count this, how the app should be divided in containers and volumes (still no clue), how is this billed and all the complexity derived from Riak, well... I'll just drop Riak.

I played with it a bit and have a rough idea of its model (very rough). But I spent more focus on it than on other things that could be more important (build automatization, ssl...). But I still didn't learn decently how to manage it (and in the middle of this TODO mess I don't want to) and super-fast or memory efficient it may be, if the app has ten freaking users and no one more, it's just overhead (no mesuraments needed). No, really, I have no SLA requirement, no need for replication, scaling or other buzzwordy stuff.

And that's it. I might also think of dropping Docker (/cry). I don't want to, but I actually have a server to deploy netronner and I'll be an idiot not using it just because there's no Docker on it.

Well that's it. I might find some more simplifications to do before going to prod, don't know yet. See ya again!