OpenStack in TSSG
This talk is based on our experiences in deploying OpenStack and the various issues we had to overcome to get a suitable deployment. This was a 15 month project for TSSG, which just recently finished. We undertook a detailed study of end-user requirements (and needs), looked at their capabilities to interact with various solutions, and designed the system the best leverage their needs and abilities.
This will incorporate our reasons for changing (from VMWare), including the time / cost implications and the overall usability of the previous and the new system. In particular we wanted the address the issues of self-service, automation, and accountability.
We will detail the multiple implementations of OpenStack that were deployed and the particular issues that were addressed with each iteration, including removing all single points of failure in the controller, service, and network nodes.
The OpenStack view of networking was rather ‘simple’ from our perspective and required considerable work to get a dynamic, HA, and scalable solution, that minimised the points of failure.
We will detail the different user types that were identified and their varying needs and how we were able to establish common service layers, and yet provide different computing availability zones for the varying user groups.
We have looked at the IaaS and PaaS service models, with an initial focus on the IaaS model. Using the IaaS model, various PaaS models have ben deployed upon it, with the end-users often creating ‘golden images’ for others to use. This has created an internal user community that removes the service creation burden from the system administrators, and enables them to explore new technologies quickly.
Ultimately we will discuss the business requirements around a new cloud computing platform, the end-user requirements gathering and capability assessment, technical systems implementation, and an overview of how the system has preformed since deployment.