For those of you who missed the OpenStack Summit in San Diego 2 weeks ago, one of the highlights was the Keynote presentation by our friend Reinhardt Quelle, architect of the Cisco Webex OpenStack cloud.
In this 25 minute presentation, Reinhardt makes a compelling case, based on real-world experience, for using OpenStack to run a cost-effective private cloud. The mission: delivering elastic, multi-tenancy infrastructure that streamlines putting new applications and features into operation. (It’s something we know a thing or two about).
Key capabilities that running Webex Cloud Infrastructure on OpenStack cloud provides:
- A large-scale private cloud infrastructure with thousands of CPU cores and multiple petabytes of storage, based on Cisco’s UCS server platform
- Integration with key infrastructure applications such as identity services and Splunk analytics that support the core Webex conferencing services
- Visibility into underlying machine behaviors to understand the performance and stability profile of any application
- High Availability (HA) support across all OpenStack components implemented exclusively with popular open source components
- Automated provisioning framework using Puppet manifests for ongoing management of the OpenStack cloud infrastructure
- Object storage service using Swift, including multi-datacenter support
- Application of devops methodologies to instantiate a fully programmable infrastructure, allowing the operator to manipulate the infrastructure to ready for any given application’s requirements.
- Support for distributing services between multiple global datacenters and locations
There are a few nuggets in here that are you might want to see, that reflect some of our work together with the Cisco Webex team.
- Migration from EC2 was not about API compatibility; in fact, the relevant application elements were ported in about 2 days
- Integration of monitoring and management capabilities to allow internal customers/tenants full visibility
- Cobbler and Puppet provides hands-off deployment all the way through RAID configuration
- Understanding that application availability is the responsibility of the application, not the underlying platform
- Large object upload across swift nodes
- Tempest tests that do large-span validation that can exercise production environments.
- Understanding the impact of workload placement (i.e., beyond random spin-up of workloads by the OpenStack scheduler)
It’s worth noting that some of these features have been incorporated in the Cisco Edition of OpenStack – which is based on Folsom.
So, have a look at the video. And if you haven’t already, download a case study of how Mirantis helped put OpenStack to work at Webex and others.