OpenStack Summit 2015 Tokyo: OVH welcomes project evolution towards Public Cloud use
The biannual OpenStack gathering was held in Japan on October 27-30, 2015. The technical team in charge of the Public Cloud offer at OVH.com took part in the event as an “Ops”, a term used for operators of the open source project. Review the trends with Jean-Daniel Bonnetot, system administrator, specializing in OpenStack at OVH.
The time where it’s necessary to introduce OpenStack is over. Today this project is known worldwide, as evidenced by the diversity of its contributors and users. Here is the message delivered by Jonathan Bryce, executive director of the OpenStack foundation at the opening of the Summit.
Today, more and more SMEs as well as large enterprises are adopting OpenStack for their projects in production. The early adopters are not alone. Five years after the birth of OpenStack, we can say that its use is widespread.
Thanks to Big Tent, OpenStack is being enriched very quickly
The acceleration during the last months owes a lot to the opening of all the open source initiatives gravitating around OpenStack within “Big Tent”. Under this appellation are gathered all the peripheral projects at the core of OpenStack which are approved by the Technical Committee according to several criteria: alignment with the OpenStack mission, the existence of a team supporting the project, conformity with the OpenStack workflow, interoperability with other projects, and respect to the 4 “Open” rules: Open Source, Open Design, Open Community, Open Development.
We find among the projects enriching OpenStack’s features an interesting initiative called CloudKitty. It’s a module, developed by Objectif Libre, which can help calculate project usage costs. We also find within the scope of Big Tent, projects such as Magnum or Kuryr, devoted to containerization (we’ll come back to this later).
OpenStack Liberty: the project has reached adulthood
Each summit is also the occasion to celebrate the latest release, the fruit of the previous six months. Liberty, the twelfth version of OpenStack, brings with it 4 million lines of code (according to the infographic), placing it between that of an American military drone and PhotoShop CS6). This release includes a lot of new features, but – it must be noted – they are fewer than in previous releases. This is for good reason, the current watchwords are stability and optimization. This is a turning point in the history of OpenStack, which enters into the age of maturity. We see within the community, a desire to better coordinate the different actors surrounding the project.
This is precisely the mission of “Work Group Product” in charge of deciding which major developments will be followed for the next releases collecting feedback from users, especially the “Ops” such as OVH.com – which claims to have launched, in combination with its Public Cloud, VPS, RunAbove and hubiC offers, several hundreds of thousands of VMs under OpenStack and tens of petabytes of object storage under Swift. End users and Ops will be best informed on the state of project advancement thanks to a new tool: Project Navigator
The network will be at the heart of the next development cycle
Neutron (component that manages network) is becoming the most active project in terms of growth in the number of lines of code, for the first time, dethroning Nova (component that manages instances). This is a sign that Neutron is still in an evolution phase, while Nova is concentrating on reliability and robustness of existing features. The influx of ‘Telco’s’ (telecommunications business sector) into the project is intensifying the work at the level of NFV (network function virtualization). Despite its obvious lack of features, the DVR (distributed router), also under development, is beginning to be adopted. As a reminder, this new network architecture has been developed to resolve bottle-necking issues at the level of the network nodes. The BGP protocol is at the heart of the solutions under consideration and should be prominent in the next development cycle.
Containerisation and consideration of Public Cloud use case
Six months ago, at the OpenStack Summit Vancouver, we saw a favorable attitude towards container management, and in particular of Docker within OpenStack. Magnum which is the main container management project, continues its evolution and encompasses a number of side projects necessary to put it in orbit. For example, Kuryr, offers to take charge over all the network elements of containers related to Neutron. There is no doubt about the future of containerization, it is an ideal complement to the large virtualization family of cloud computing. OpenStack has incorporated this strong trend.
Among the other evolutions, of particular interest to OVH, we have seen a turn in OpenStack from Private Cloud towards the Public Cloud. OpenStack isn’t neglecting Private Cloud projects, but is diversifying and taking better consideration of issues that occur during deployment on large scale infrastructures, exploited by a multitude of users. Public Cloud operators have more and more weight within the community and that is a good thing. At OVH.com, we have made several adaptations of OpenStack code to satisfy our large scale operator needs. We had to revise all the north-south traffic of Neutron, to move it on the hypervisors and we have worked on parallelization of the Nova schedulers to respond to high stresses caused by simultaneous starting several hundred instances. The issues, addressed thus far within the scope of the project, will be a topic of future work within the entire community.
Beyond interoperability: a total mix of infrastructure and services?
Interoperability remains a central topic. Six months ago, there was an effort to form identity federation (Federated Keystone), a mechanism making it possible for uses to use the same OpenStack identity between several providers. This effort continues and extends to the other layers of the project.
Unification at the L2 and L3 levels of the network on several sites was discussed, as well as the sharing of resources between several providers (for example the instances at a provider, or at home, and Object Storage at another place) or even the unification of several OpenStack infrastructures. Everything remains to be done, as technical solutions are far from being accomplished. But discussions have indeed been initiated and the global decision has been taken to go in that direction. The ultimate goal is to achieve a universal system such as DNS. Inspiring!