The topics of debate at the OpenStack Silicon Valley conference at the end of August were not restricted to new features and capabilities under development for the cloud computing framework, but extended to discussions about what direction the project should take in the future, often with diametrically opposing views.
Does this mean that the cracks are starting to show in the OpenStack community? Or is this all simply a sign that OpenStack has now become such a key technology to many companies that its future direction is a matter of great import?
OpenStack may have started as a collaborative project between Nasa and Rackspace to emulate Amazon's Elastic Compute Cloud, but as an open platform for building cloud infrastructure. However, it has since mushroomed into a kind of overarching umbrella project for all kinds of open source solutions covering software-defined storage, big data analytics, software-defined networking and more.
OpenStack is basically a management framework into which services must be plugged to deliver a working system, and this has become its strength and weakness. It can be configured in numerous ways to suit almost any particular application you have in mind, but this has led to complexity and stability problems for some users.
Randy Bias, EMC's vice president of technology, who is also a director of the OpenStack Foundation, addressed this at the conference, calling for the creation of an OpenStack "standard" in the form of a base reference architecture against which specific implementations could be tested and verified.
Bias also expressed his opinion that OpenStack should follow its own course, and that developers should focus efforts on emerging applications and use cases, such as those operating at massive scale, instead of trying to compete with established cloud platforms such as VMware.
"OpenStack is not a cheaper VMware. It doesn't have all those capabilities for the enterprise like high availability or disaster recovery," he said, adding that the community should not waste time and effort trying to deliver these when it could be working on unique capabilities for the platform.
However, Bias was followed on stage by Intel's Diane Bryant, who said that her firm is investing in exactly that, adding in all of the enterprise-grade features that will enable OpenStack to be widely used for running mission-critical enterprise workloads in the cloud.
Later on, a panel session on microservices exposed the opposing viewpoints on technologies such as containers, which some in the community regard as the way forward, while others point to the problems in keeping workloads isolated from each other in containers and insist that virtual machines are a better approach.
Everyone present agreed that building applications and services from smaller components - which is what microservices is about - is desirable for various reasons, but the debate soon heated up over how best to approach this.
Proponents of containers pointed to the fact that these are lightweight and enable many more of them to be run on a single server node and spun up faster, while those backing virtual machines said that the technology is more mature and widely supported, and the density problem can be addressed by new techniques such as micro-virtualisation.
Even in the containers camp it is felt that there is no single way of approaching this technology because Docker and Kubernetes are as different as chalk and cheese in the way they implement provisioning and orchestration, according to one commentator. This is one reason why OpenStack contributors created the Magnum containers project as a higher-level service that Docker and Kubernetes can plug into.
These differences could be seen as troubling for potential users, but instead they show that OpenStack is becoming a key foundation piece for IT infrastructure in many organisations, and that users and developers (the two are the same in many cases) care about which features are implemented and how.
And while there may be a danger that OpenStack is being pulled in different directions, there is no reason to believe that this will be its undoing. The very flexibility of the platform means that those wanting to build an enterprise-grade cloud stack can plug in modules that will do that job, while those wanting a scale-out infrastructure for telecoms providers can plug in modules that will do that job.
A base reference architecture could encompass only those core modules that are common across all OpenStack implementations and still deliver on Bias' vision of a testable and verifiable build of the platform.
So while the debates at OpenStack conference could be seen as divisive, I see the enthusiasm of a community determined to build a platform for all. The same arguments are probably raging among developers of other platforms, just taking place behind closed doors, while OpenStack's debates are out in the open for everyone to see and join in. And that seems like a better approach to me.
Q3 losses reverse Q2 gains
FBI briefing US companies to dump Kaspersky, claiming intelligence prove it a 'threat to national security'
Kaspersky rejects FBI accusations that its products are a 'threat to national security'
But breached contractor says that it simply didn't have that much data
EE follows Three in threatening legal action against Ofcom - but for entirely different reasons