Four years of OpenStack: will it dominate cloud in four more?
In just four years OpenStack, the open source cloud orchestration project originally launched by NASA and Rackspace in 2010 has garnered more industry momentum and community involvement than nearly any other cloud-centric open source project. That said, as the project turns four it’s worth asking whether OpenStack, and indeed the paradigm it represents, will come to dominate cloud?
Looking at the stats one gets a good sense of the buy-in OpenStack has garnered in a relatively short period, particularly over the past year.
As of May this year OpenStack boasted 16,266 members from 355 organisations. So far this year there were 2,130 contributors and 466 average monthly contributors. At the same time last year, the project boasted 9,511 members from 209 organizations, 998 total contributors and 230 average monthly contributors.
The past year also saw serious enterprise IT heavyweights like IBM and most recently SAP join the ecosystem, a sign of growing acceptance even among vendors that have traditionally been less engaged with open source.
But despite its growing developer diversity OpenStack is still dominated by a handful of service providers and vendors, which is fairly common among open source projects. Combined, Red Hat, Rackspace, HP and IBM have contributed nearly 80 per cent of the total code for all distributions. And according to Laurent Lachal, a senior analyst at Ovum, there is an even stronger concentration of power when it comes to up-and-coming (not yet officially recognised) OpenStack projects managed in StackForge.
“At the moment OpenStack is looking to find the right balance between bottom-up innovation and top-down management,” he said. “There are so many sub projects out there at the moment, which is healthy to some extent, but many of them have barely scratched the surface in terms of their completion.”
“The Foundation really needs to find that balance if it’s going to deliver on its related goals of interoperability and consistency.”
It’s unlikely that OpenStack will become dominated by any one vendor, however, because there are simply too many vendors and service providers involved in the project.
“They simply wouldn’t let that happen,” he said. “There are so many big vendors involved, each with their own view of what OpenStack will do and needs to do for them and how it can help generate revenue, which counterbalance one another.”
He also explained that as OpenStack grows to include more and more toolsets there is an increasing need to ensure consistency between the various components at all levels of the platform – in APIs, extension and plug-in management, and configuration.
Without commercial interoperability standardisation enterprises will need to continue to fill in the gaps with configuration automation framework tools like Chef or Puppet for instance, along with other cloud orchestration technologies to create an abstraction layer (which can add overhead).
There is, he said, a need to ensure interoperability between commercial distributions as well, which will come when the community defines a core set of capabilities that should be expected to ship stock with all commercial products with an ‘OpenStack’ stamp on them.
It’s an initiative currently being taken up by the Foundation. Beyond Tempest, OpenStack’s current interoperability test, the Foundation is developing its first faithful implementation test specification or FITS, which will help establish a solid benchmark for interoperability.
In addition to making commercial-grade OpenStack more enterprise-friendly the FITS may also go a long way towards helping OpenStack articulate a cohesive, clearer vision of what it offers users and service providers, something the Foundation has also struggled with to some extent.
“Open source movements will always struggle with the question of what projects will make it into the main project and what won’t. In the case of OpenStack, Solum is a great example of that,” said Toby Owen, head of product strategy at Rackspace UK.
“The flips side is, that’s what is so great about the democracy of open source. If a feature or sub-project doesn’t gain enough traction in terms of contributions you know that it probably doesn’t have a use case broadly appealing enough.”
Owen explained that OpenStack will likely evolve distinct distributions to suit two broad camps: enterprises and service providers.
“There are also groups starting to talk about industry-specific OpenStack distributions – so an OpenStack for government for instance, with all of the heightened security and information management capabilities, or an OpenStack for banks.”
There’s also an opportunity to tweak OpenStack specifically for certain workloads, he said, which will likely happen as more members contribute use case experiences back into the community.
“It’s fair to say it’s still early, but we’re already seeing a number of large enterprises use this for their private clouds,” he said.
Nevertheless, while OpenStack is picking up traction among large enterprise early-adopters like Wells Fargo, Disney, AT&T and Sony, it’s clear commercial deployments among mainstream businesses are quite low.
“We’re seeing around 10 per cent compound monthly growth in terms of instances, mostly for on-premise private clouds,” said Sean Lynch, chief executive officer of Metacloud, a vendor of OpenStack distributions and an OpenStack-based cloud service.
The company recently helped stand up its twentieth OpenStack-based private cloud, and Lynch said that while OpenStack is still fairly nascent 10 per cent monthly growth projected out over a year or two still amounts to significant growth.
“While adoption is clearly nascent now it’s not hard to envision a future where OpenStack is the monolith, and where players like Amazon Web Services are smaller – not in and of themselves, but certainly with respect to the overall OpenStack capacity,” he said.
Particularly with identity federation (Keystone), cloud peering, app and VM portability taking such renewed prominence within the OpenStack ecosystem, it’s possible the open source platform could win out against proprietary equivalents, he explained.
“I doubt you’ll see one single OpenStack player get much bigger than AWS on a per-vendor basis. However, I really believe Amazon will look relatively small compared to the global aggregate OpenStack footprint.”
The industry has already seen OpenStack peering, for instance with the CERN and Rackspace linking their OpenStack based clouds, and Lynch said that the next logical step is to create a many-to-many market oriented cloud platform. It’s more challenging than it sounds, and with current approaches just isn’t feasible without significant overhead in terms of latency and application performance, but Metacloud and Rackspace are among a few vendors and service providers keen to develop these capabilities.
Whether OpenStack clouds will dominate the market very much remains to be seen. AWS continues to eat up the lion’s share of IaaS spending, and others with powerful proprietary platforms like CenturyLink and DigitalOcean continue to eat into the market, too. With complex peering and federation capabilities within OpenStack still in development, and with cloud infrastructure spending still accounting for a sliver of overall IT spending, it may be some time until an OpenStack reaches critical mass in terms of the vision Lynch articulated.
But the increasingly robust platform has gained buy-in from startups like Metacloud as well as, critically, IT incumbents like Oracle and IBM in addition to direct competitors like VMware, and the Foundation is mobilising a rapidly growing developer base – a community which is growing far more quickly than those involved with rival projects like Cloud Stack and Eucalyptus. With 16,266 members from 355 organisations and counting after just four years, OpenStack boasts more momentum than even Linux enjoyed this early in its development (though the wider popularity of computing had something to do with that, to some extent). And as long as the industry continues consolidating around the platform, and the challenges it aims to solve remain relevant, there’s seemingly little to prevent it too from dominating the X86 cloud landscape.
If one thing is certain, it’s that it will be interesting to see how the OpenStack ecosystem looks in another four years.
“It won’t happen for some time, but I don’t think it’s an overstatement to say that when more complex peering and federation with OpenStack starts happening, players like AWS will start to look quite small,” Lynch concluded.