Continuous Delivery and Large Microservice Architectures: Reflections on Ioncannon

Short Form


Continuous delivery of a monolith is easy, just automate, automate, automate! But what challenges will you run into applying the same ideas to 300 microservices? Come and find out!


I spent 2 years as the primary designer and developer of Ioncannon, the continuous delivery system used at Gilt Groupe to move over 300 microservices from development, through testing, to production. Much has been written, and open-sourced, regarding the continuous delivery of monolithic or substantially-monolithic systems, but considerably less has considered continuous delivery in the context of microservices, and the new and unique challenges in implementing it in such an environment.

I will discuss our approach, the design and evolution of Ioncannon, what worked well, and what I would do differently. I’ll touch on build and packaging systems, tooling for microservices, configuration management, test frameworks, maintaining test environments, where databases fit in, as well the cultural and political aspects of continuous delivery.

Speaking experience

I've spoken at several conferences, user groups, and company-internal presentations. Most are listed at

I have spoken about related work previously, but this will be the first talk fully devoted to ioncannon / continuous delivery.


  • Bw head

    Kevin Scaldeferri

    New Relic


    Kevin Scaldeferri is a software developer specializing in scalable, high-performance server applications. He spent several years at Yahoo building ad serving systems, and at Gilt Groupe, maintaining the core e-commerce and order processing components, as well as development tools, infrastructure, and architecture. He currently works on distributed data storage and query engines at New Relic.

    He has spoken at OSCON, Open Source Bridge, Lambda Jam, CUFP, YAPC, several user groups, and given internal tech talks on a variety of topics such as “How to Serve a Billion Requests a Day with Perl”, “Beautiful Concurrency with Erlang”, SBT, Continuous Deployment strategies, and more.