This post is a work in progress… I’ll keep editing it in the next couple of days :construction:.
The Right Environment
Become a proficient Linux user (why: I need you to understand how to deploy software to Linux servers, and how to maintain them). I think you already used Linux a lot. Let’s just assume from this point on that you do everything I recommend in the points below in Linux or OSX (any nix-based OS will work).
Git Started! :checkered_flag:
Check your feelings at the door for this one :door:. If you’re not using git you don’t really matter – sorry, but everyone is using git. It’s simple and works well in teams. Run through the git tutorial and make sure you understand what you just did.
Also make sure you have a Github account. Pretty much every project that matters is on Github and many developers are sharing code there and collaborating on projects. Chances are that you will have to get an account just to get cracking. Get one now :octocat:.
Monoliths and Microservices
Understand the difference between monolithic vs microservice. Normally speaking people would write complete applications in Python, Ruby, PHP or any other language for the web but the problem is that, as applications grow, they become harder to maintain. The latest trend is to develop microservices which subdivide work into smaller tasks which are managed by a single application.
Instead of a complete Netflix-like application (that handles cataloging, authentication, permission management, streaming, etc), one could build separate video decoding/encoding services that just translate video from one format to another, authentication services that just look at user credentials and determine if they get permission for a certain action yes or no, cataloging services that simply look at a user and figure out which items should be available.
Learn Rails from Hartl’s tutorial which will give you a great understanding of building applications that do everything). You will write a simple micro blogging application, but that is good enough to understand how an application works that does multiple things. For an application this small it doesn’t even make sense to try the microservices approach because that approach comes with a price, too. They’re harder to write, generally, because there are a lot of separate moving parts :wink:.
Now read why microservices matter. At this stage you have probably already played with Heroku (from the Hartl’s tutorial) so we’re just connecting dots.
To the cloud
Heroku is a great service to deploy application. Sometimes, however; you need more control. Perhaps you want to manage your own stack, perhaps there are some requirements to the infrastructure, perhaps you don’t want to lock yourself into Heroku. If you’re a startup, I don’t get why you’d have a problem with Heroku. I think Heroku is a wonderful solution to get applications live. Just focus on figuring out if the assumptions your company is based on are valid. Until you’ve validated or invalidated them and are serving people to the point that you get phone calls of overjoyed clients, you have no reason to worry too much about where your application is running and how it is deployed.
However; if you want to understand what the big fuss is about in application development, lately you need to familiarize yourself with the following… configuration management (ansible, puppet, chef), containerization (docker, rkt), orchestration (docker-swarm, mesos, kubernetes). Basically I just listed a topic and a few solutions within the parenthesis. It is unlikely you’ll be able to get a in-depth look at all of the listed options, but let’s get at least a birds-eye view of what’s possible.
With configuration management we basically simplify how we provision and configure resources. Think about
- starting a new virtual machines at your provider,
- installing base OS-es and some of the needed tools to get started and
- setting up networking between the machines.
After the machines are running you need to deploy applications to these boxes. I have used capistrano a lot to do this. With capistrano, I would basically just install the necessary tools on the VM and set up the application, however; the problem with this occurs when one starts to run multiple applications on a single machine. Somehow the applications share libraries. If application A needs an updated version of some libs while application B wasn’t yet updated to be compatible with the newer library we have problems :fire:. Containerization provides a solution by allowing us to run our applications in sandboxed environments. Every application runs in its own container. The only thing the applications share is the kernel (pretty low level). The libraries and all the other junk that is specific to the application is contained within… the container. Solutions like docker and rkt make running containers quite easy.
Whenever you have enough containers floating around, management of these suckers becomes rather tricky. Some containers you rather have running on the same machine for simplicity’s sake. Having the database and the application that relies on that database running in different containers in different machines incurs extra networking overhead, running them on the same machines makes more sense. On the other end, ramping up the number of webservices, because there is more traffic on the site sometimes requires some finesse. Perhaps you want to equally distribute the workload over all virtual machines, instead of cracking down on a single VM to the limit. Mesos, kubernetes and docker-swarm are a few tools that make this a bit easier. You could tell these tools, to ramp up certain containers on some VM’s with specific hardware by using labels (e.g.: only ramp up database containers on machines that have very fast SSD storage instead of cheaper and slower magnetic disks).
To be continued :construction: