Compute. Storage. Networking. Wouldn’t it be nice if could just dump all of our raw hardware, hide it behind a slick interface, and turn that metal into a push button operated data center?
Our applications would just work. If things got slow, we’d just slam more metal into the rack. Yeah. So awesome, dude. So, so awesome…
We aren’t there yet. But maybe we’re making progress. On today’s Datanauts show, we’re joined by Michael Hausenblas, developer and cloud advocate at Mesosphere, to speak with us about Apache Mesos, a data center operating system.
We’ll discuss the concept of a data center OS, and drill into specifics on Mesos, including its framework and use cases.
Show notes go here.
Part 1 – Level Setting
- What is a DCOS?
- Use cases for a DCOS – who is this solution right and wrong for?
- What is Mesos, specifically?
- Explain some fundamental Mesos concepts:
- Master & slave (agent in 1.0)
- Applications/frameworks (scheduler & executor)
- Resource offers
- What sorts of applications can run on Mesos (i.e. anything, or specially designed)?
Part 2 – Brownfield Deployment
- What sort of existing environment does well with Mesos?
- Which parts of the DC does Mesos abstract?
- Compute (CPU/RAM)
- Storage (local, writing to /var…)
- Networking (TCP/IP port ranges)
- How does Mesos gain the permissions needed to execute tasks?
- Does it make sense to install Mesos in a pod, and then grow? What do most people do?
Part 3 – Operational Impact
- What’s it like to deploy an application using Mesos?
- How do operators manage Mesos resources?
- Let’s say my application running on Mesos is underperforming. How do I troubleshoot?
- What’s it like to upgrade Mesos?