Integrating OpenStack into an enterprise environment is challenging. A typical enterprise compute environment features tight security control; limited user privileges; and a slew of existing tools and systems for management, inventory, etc. In this talk, we will describe our modest attempt to integrate a multi-site OpenStack deployment into the compute infrastructure of a quantitative investment manager based in New York.
We will explain the overall architecture and the reasoning behind the design choices we made. In particular, our talk will highlight Kerberos integration, deployment automation via Ansible, scale-out and HA neutron networking fully integrated with physical networks, a model we have developed for integrating with other infrastructure services, as well as an easy-to-use API layer for cloud users. We have been running OpenStack in production hosting a variety of mission-critical services since 2013.