The OpenStack Foundation debuted its 16th milestone release today with the launch of the OpenStack Pike cloud infrastructure platform. Pike follows the OpenStack Ocata release which came out in February and had a focus on cloud federation.
Unlike Ocata, the new Pike release has a particular emphasis on enabling standalone OpenStack services, without the need for an entire set of OpenStack projects. For several years, the OpenStack community debated a definition for a common set of projects, known as Defcore that define what it is to be an OpenStack cloud. Among the projects that Pike now enables to run in a more standalone, composable approach are the Ironic bare-metal and Cinder block storage projectS.
“There are new features in Pike that make it easier to run Ironic in a standalone way, with better integration with Neutron, so you can run bare metal on fully segmented networks,” Jonathan Bryce, executive director of the OpenStack Foundation, told eWEEK. “It’s now also possible to run Cinder in a standalone way without Nova.”
Neutron is the OpenStack Software Defined Networking (SDN) project, while Nova is the core compute project. Nova was one of the original projects in OpenStack when the effort was first launched by NASA and Rackspace in July 2010. A primary element of the OpenStack Defcore definition debate was that for a cloud deployment to be called OpenStack, it had to include Nova, but that’s a model that is now changing. Mark Collier, chief operating officer of the OpenStack Foundation said that the model for OpenStack usage and deployment is continuing to evolve with Pike.
“What we are still doing with the Defcore approach is defining what it means at a basic level to say that a product or service is OpenStack-powered,” Collier said.
Collier added that Defcore is still valid to help define OpenStack cloud deployments that have a complete set of services. He noted that Defcore in part is also an effort to help ensure interoperability, so that one OpenStack-powered cloud can be interoperable with other OpenStack clouds.
“Where people are now cherry-picking individual services in a composable manner, that does fall outside of the current Defcore, OpenStack-powered cloud program,” Collier said.
Lauren Sell, VP of Marketing and Community Services at the OpenStack Foundation said that the plan is now to expand the OpenStack-powered program to have a wider definition. She added that the agenda for the OpenStack Board of Directors meeting in two weeks includes an item for discussion of vertical market model for branding and certifying what an OpenStack-powered deployment should include.
The question of defining what OpenStack includes is critically important for enabling interoperability and federation of cloud services that identify themselves as being OpenStack. Without the same set of core services, different cloud services that use the name OpenStack might not be able to work with each other. Bryce emphasized that the OpenStack Foundation is not abandoning the idea of federation, rather it is now looking at a more open model.
“People are looking less for highly prescriptive cloud infrastructure and looking for more guidance about how they can pick tools and services to build the environment that they want,” Bryce said.
As such, rather than having a single testing methodology to define OpenStack cloud interoperability, Byrce said that the idea moving forward is to have multiple interoperability testing models for specific cloud use-cases and projects.
Cells
While OpenStack is now enabling a model for deployment where the core Nova compute project is not required, the Pike release does in fact have some significant improvements for Nova users. For the past several releases of OpenStack, there have been incremental additional in Nova to support a scale-out technology called Cells.
Cells allow multiple Nova compute nodes to be aggregated and managed by a single Nova API. The first version Cells debuted in the OpenStack Grizzly release in 2013 and Cells version 2 initially debuted October 2015 with the OpenStack Liberty release. In the new Pike release, Cells version 1 support has been deprecated and Cells version 2 is now considered to be production-ready.
“With Cells v2 scaling is now just a fundamental piece of the Nova architecture,” Bryce said.
The OpenStack Pike release also marks a noteworthy milestone for one of the primary foundational technical components for the entire platform. When Rackspace and NASA first created OpenStack in 2010, the entire platform used the open-source Python 2 programming language. With the Pike release, Bryce noted that platform has been re-based such that it supports the much newer Python 3.5 release. Python 2.7 is scheduled to reach its end-of-life for support in 2020 and by moving now to Python 3.5, the goal is to help the cloud platform stick with a fully supported version of Python.
“What the community is trying to do is to stay ahead of the main Python versions that the main Linux distributions are now using,” Bryce said. “This was a big lift for OpenStack to get Python 3 series support well ahead of time to make sure that it’s all functioning correctly.”
Sean Michael Kerner is a senior editor at eWEEK and InternetNews.com. Follow him on Twitter @TechJournalist.