Ceph: Operations & Deployment

A typical Ceph installation will have tens to thousands of servers. Managing systems at such a large scale requires significant automation. In this talk, we will walk through how the Ceph project plans to handle the lifetime of a cluster: networking setup, initial software deployment, growing & shrinking a running cluster, handling software & hardware failures, monitoring and performance and usage metrics.

We will explore several core Ceph concepts that enable a Single Point of Failure (SPOF)-free architecture in every aspect, including initial deployment. We will show how to enable efficient data center operations for handling hardware and software failures. We will introduce product functionality that integrates with various devops tools, such as Chef, Puppet and Crowbar, allowing each to deploy Ceph according to their individual strengths.

Presented on 2012-09-19 at SUSECon.