OpenStack approaching a critical phase in its development

With OpenStack now 'feature complete', V3 technical editor Dan Robinson ponders which direction the project will take in the future

The OpenStack Foundation held its first summit event in Europe earlier this month, where the developers and vendors involved with the open source cloud computing framework got a chance to get together with partners and customers, as well as industry analysts and the press, to share the latest updates on what is happening with the project.

As one of those attendees, I was struck by the almost palpable buzz of excitement around the conference hall during the event. Perhaps this is because it was more than just a technical conference; it was also a chance for those involved in the OpenStack ecosystem to show off their wares in the OpenStack Marketplace and draw in potential customers and partners.

Whatever the reason, OpenStack feels like the place where a lot of the action in the IT industry is happening right now, and that's perhaps not surprising considering the number of vendors and organisations that have lined up behind the project, including industry giants such as HP, Intel and IBM.

OpenStack also seems to have become the focus for much of the open source community's efforts, not least because it has allowed many of the Linux vendors such as Canonical and Red Hat to get into the cloud business by integrating it with their own operating system distributions, but also because it involves contributions from many other open source projects.

However, the real reason for OpenStack's success is probably more prosaic: it is open and non-proprietary, which means it is not controlled by any single vendor, and can readily be adapted as required. Because it is essentially a management framework, it also allows users and vendors to plug in components that they want in a modular fashion.

This has enabled VMware to offer its own OpenStack distribution, for example, integrating it with its vSphere hypervisor, NSX network virtualisation layer and VSAN virtual storage platform, ostensibly so that existing VMware customers can take advantage of the OpenStack APIs for developing new applications and services.

In fact, it was mooted at the summit that OpenStack's APIs could turn into an industry-wide standard for accessing IT infrastructure, which could deliver a welcome measure of harmonisation for developers and customers alike.

The OpenStack framework itself is also finding potential applications beyond cloud computing. Many telecommunications firms are looking to use it as a platform for implementing network function virtualisation, for example.

This approach would see costly and inflexible network hardware replaced with software functions running on industry standard x86 server boxes, a move to which Intel is understandably keen to give its backing.

With all this in mind, the future for OpenStack looks decidedly rosy, but there are a few potential clouds on the horizon, if you will forgive the pun.

OpenStack's development pace has been rapid until now, the platform having notched up 10 releases already since its first appearance as a joint project between Rackspace and US space agency Nasa.

This blistering pace was necessary in order for the fledgling project to gain all the features and capabilities required to deliver a fully functioning infrastructure-as-a-service cloud without being left behind by rival platforms that had a head start.

This has arguably come at the cost of platform stability, and there have also been accusations from some quarters that a focus on features rather than usability has led to OpenStack being complex and tricky to deploy compared with other cloud platforms.

With the latest Juno release, the platform is regarded as more or less complete from the standpoint of features, and attention is now turning to ease of use, stability and manageability.

However, the project appears to have reached a crucial point in its development, and there are already competing views over how to proceed. Some have advocated a change from the current six monthly release cycle, for example.

Meanwhile, Ubuntu founder Mark Shuttleworth has recently advocated splitting the modules that make up OpenStack into those that are 'core' and those that are 'common'. The 'core' module label would be restricted to those that are essential for a functioning OpenStack cloud.

These kind of discussions are almost inevitable when dealing with a collaborative project that has a large number of contributors, but they show that the platform has now matured to the point where it has met most of its original goals, and the community needs to decide what the future goals should be.

Hopefully, these problems can be ironed out without the kind of acrimony that has seen other open source efforts 'fork' into separate projects that take diverging paths, or even come to grief.

As things stand at the moment, OpenStack looks like a fine example of what the open source development model can achieve.