To meet its challenges of running at scale, Twitter has transitioned over the last few years from a monolithic Ruby on Rails application to a service-oriented architecture. Apache Mesos, a platform for building and running resource-efficient distributed systems at scale, has played a key role in this transition. Mesos acts like the "kernel" for Twitter’s data centers, making it easier, faster, and more-reliable for new services to be launched and maintained.
This talk will focus on the following topics:
Brief history of operations at Twitter and how that led to the adoption of Mesos.
Designing primitives and abstractions for today’s distributed systems.
Challenges of scaling to thousands of nodes while efficiently utilizing resources.
Reliability and high-availability at scale.
Lessons learned running production software infrastructure.
