As I begin my contributions to the Apache Mesos project, I decided to also understand what’s at hand. There are several projects that add-on/support Mesos, so in addition to comprehending the interworkings of Mesos itself, I began researching a couple companion projects, mainly Marathon and Aurora .
For an introduction on both of these, I started with this YouTube video of Joe Stein giving a talk on how to build and deploy applications on Apache Mesos, where he briefly discusses both Marathon and Aurora:
First, here’s what I realized: Apache Mesos takes us ‘back in time’ and basically turns the entire data center into ONE HUGE COMPUTER. In today’s IT, we understand the importance of not having a SPOF (single point of failure). Having multiple machines to run your compute workload(s) offers you several advantages:
However, something that became clear to me as I began looking into this DCOS (Data Center Operating System) concept was the problem of static partitioning. As this slidedeck explains, partitioning compute nodes into servers that have a single-purpose causes the following things to happen:
With Mesos, you get all the benefits of having multiple machines in your datacenter, but they all act in unison as one ‘common operating system’, giving you the ability to deploy services across the entire data center in one fell swoop. With Mesos, the entire datacenter acts as one large computer, allowing you to dynamically assign resources (in the form applications and services) to the datacenter you see fit, all while abstracting away the underlying infrastructure details. At the end of the day, we only care about deploying our applications and services to an environment that can dynamically grow and shrink with the needs of the business. If we can do that with a holistic approach, making it easier for developers to consume the available resources inside your datacenter, we should.
I think I get it now. So does Twitter, AirBnB, Google, and others. Not only that, I think this approach to cloud technologies kicks ass, and I wanna help.
Onward, and upward…