Case Study: Yelp achieves a unified hybrid cloud and faster deployments with Mesosphere

How Yelp built a hybrid cloud using Mesos and Docker

Yelp faced a growing infrastructure headache: by late 2013 scaling services across its own datacenters and AWS had become complex and time‑consuming for both developers and operations. The company juggled multiple deployment systems (bare metal tooling and Asgard for AWS), slow “golden image” builds, inconsistent configurations between environments, and a slow, hard‑to‑parallelize test pipeline that took about 80 minutes to run end‑to‑end.

Yelp adopted Docker and Mesosphere’s packaged Mesos (with Marathon) to unify compute resources and built a Docker‑based PaaSTA platform that hides infrastructure differences from developers. In three months they stood up two Mesosphere clusters, launched a Seagull testing platform, and now run about 17 million tests daily (≈1 million Docker containers/day), enabling automatic provisioning and migration across datacenter and AWS, much faster deploys, better developer productivity, and cost savings through more efficient spot‑market usage.

Yelp

Eaton

Yelp