This is the intro page to developing for CloudStack. If you're looking for other topics on CloudStack, look a few pages further down on the site.
To start off with CloudStack development, you must first understand the design goals of CloudStack. And, yes, we did have some design goals from the beginning. We didn't plan on being a "five geeks in a garage operation" for very long!
Of course, there's the usual "Don't be evil", "Conquer the world", "A cloud in every data center" dribble, but that's not quite useful for you. Here, on 101, we will start with how CloudStack implemented the first three design goals.
To do this, CloudStack has three different sets of APIs: the Platform API, the Agent API and the Plugin API. The Platform API is the REST-like API through which end users and administrators control CloudStack. There are various client-side bindings to this API that allow the caller to quickly put together scripts to automate processes within CloudStack. However, this API must be very secure as the caller is not trusted. The Agent API is used for CloudStack components to talk to the ServerResource, which then translates this API to what the hardware resource understands. This API is JSON based so that the client can be written in any language and to run on any platform convenient to the developer. The Plugin API allows one to insert code directly into CloudStack deployments to add to or modify behavior of CloudStack. This is a Java API and there is a set of predefined functionality exposed through this API.
At its core, as the diagram below shows, CloudStack provides the capability to map virtual resources to various physical resources. Note that it is only the "capability" and not the actual "mapping". This is key to the design principles of CloudStack. The system administrator provides, controls, and monitors the actual mapping. CloudStack is only the tool. Use it for good or for evil. It's all up to that guy in the middle there.
Every cloud operator has hardware they're comfortable with from their experiences in running data center operations. Some, like XenServer, NetScaler, and, ahem, everything else Citrix makes, are universally loved; however, more often than not, the preferences vary. Handling various hardware has its difficulties:
To handle these types of problems, CloudStack chose to provide the following implementation designs:
Once you read the above and you go read the code, you will have the following questions. And they all can be explained by one statement, "We didn't have the time and it needs work!"
API is the window to CloudStack's soul. Since version 2.0, CloudStack has sported a flexible design for adding and modifying commands. Since version 3.0, CloudStack supports third-party vendors adding commands without modifying CloudStack's command list via the PluggableService interface. CloudStack's API designs are as follows:
REST API is broken into two parts: the End-User API and OAM&P (Operations, Administration, Maintenance, and Provisioning) or, basically, the Admin API.
Plugin API is where you can affect the most changes. The following is a picture of the CloudStack software architecture. Plugin APIs are defined by the Adapters, which expose the functionality required by CloudStack to implement cloud operations. The details of how to affect this change is more complicated and is explained in 201.
Now that you have an idea of the design principles behind CloudStack, let's take a look at the rest of the nitty gritty you should take a look at before writing code.
Setting up a CloudStack Developer environment for Windows, Mac or Linux