Building Applications on Mesos

Chapter 2. Getting Started with Mesos

Now that you have a high-level understanding of the kinds of problems that Mesos can solve, let’s dive in. What terminology is used by the project? How does Mesos manage resources? What structural constraints does Mesos impose on applications? By answering these questions and more, you’ll develop the vocabulary and understanding to discuss Mesos. In subsequent chapters, we’ll go into specifics about each technology, in order to develop applied skills in building applications with Mesos.

Frameworks

In the Mesos world, we assume that our distributed systems have a central controller and many workers. This is how most successful, scalable systems work: Google’s BigTable, Apache Hadoop, and Twitter’s Storm all have some sort of central controller that organizes their workers. Of course, the workers should be designed to work independently from the controller so that they’re not bottlenecked by it and don’t rely on it for their availability. Applications that run on Mesos are called frameworks. A framework has two parts: the controller portion, which is called the scheduler, and the worker portion, which are called the executors.

To run a framework on a Mesos cluster, we must run its scheduler. A scheduler is simply a process that can speak the Mesos protocol; frequently, our schedulers run on Marathon, but it’s always necessary to consider how to ensure high availability for the scheduler. When a scheduler first starts up, it connects to the ...