While the OpenStack community likes to present a unified front to the outside world, inside the various projects that make up the OpenStack framework, there is a lot of frustration with the Neutron networking component of OpenStack. Much of that frustration stems from the fact that after five years of effort Neutron still doesn’t scale particularly well. As such, many of the organizations that have embraced OpenStack wind up swapping in a commercial network layer of software to replace Neutron.
Against that backdrop, solution providers should be paying heed to news from Akanda, which announced that its networking software has become an official part of the OpenStack framework. Renamed Astara, Akanda will continue to provide much of the technology contributions to the Astara network orchestration project.
Perhaps just as significantly, Akanda engineers will also now have more incentive to extend the scope of their work to address the shortcomings of Neutron. In fact, Akanda CEO Henrik Rosendahl said the best thing about Astara is that it eliminates the need for complex software-defined networking (SDN) controllers, overlays and plug-ins for cloud networking because the whole networking layer is now managed at a higher level in the stack. That doesn’t necessarily eliminate the need for Neutron, but it does give organizations that want to be able to scale open-source networking services in the cloud more options.
Akanda plans to provide commercial support and other value-added services on top of Astara, Rosendahl said. In the meantime, the OpenStack community as a whole, with perhaps some exceptions, should welcome the latest addition to the OpenStack framework. After all, there’s still a log of work to be done to make OpenStack simpler to deploy and scale. One of those core problems is that many of those higher-level services are theoretically supposed to be dependent on Neutron to scale. In reality, use of Neutron within OpenStack deployments today is fairly limited.
Of course, the OpenStack community tends to gloss over the current architectural limitations of Neutron. But one day soon, hopefully, there will be another implementation of Neutron that makes the issue go away once and for all.
Michael Vizard has been covering IT issues in the enterprise for more than 25 years as an editor and columnist for publications such as InfoWorld, eWEEK, Baseline, CRN, ComputerWorld and Digital Review.