The following are the challenges:
- Customer had a mandate that any application must be deployable to any number of backend infrastructures, spanning multiple private clouds
- A specific application (with a footprint of over 10,000 servers) needed better provisioning procedures and tools to ensure the mandated goal was met
- Customer's current provisioning model required an extensive overhaul, as spin-up time was several weeks if not months, largely consisting of manual procedures
The following is the solution:
- Implemented an automated workflow-based approach using a version-controlled, software-defined infrastructure via the Business Process Management (BPM) platform/workflow engine and underlying Java services
- Leveraged Puppet architecture for build processes and packaging, and bakery workflow for images
- Provided insight into the operations via a Ruby-based console and reporting UX
- Integrating Jira into the provisioning workflow will make delivery ease-of-use from familiar tools
- Provisioned application servers and requisite number of Memcached and related instances
- The new system validates a newly-provisioned infrastructure, provides an automated cleanup of any failures, and automatically switches routing rules to serve up the new infrastructure
- Provided customer with tools and patterns necessary for repeatable operation
- Created better build procedures and processes, which yield more stable infrastructure changes
- Deployment time for the target infrastructure was decreased from weeks/months to 90 minutes for 270 servers