Mark Shuttleworth on OpenStack: HPE layoffs prove 'big tent' approach has failed

Ubuntu founder and product lead at Canonical Mark Shuttleworth says he feels validated by his earlier claims that the expansion of OpenStack projects – known as the ‘big tent’ approach - would collapse and that the community needs to focus on its core services.

Speaking with Computerworld UK at the OpenStack Summit in Barcelona, Shuttleworth claims that the layoffs from the OpenStack teams at both HPE and Mirantis lend further credibility to his theory that all the surrounding bolt-ons, which he previously described as ‘bullshit as a service' in an interview earlier this year, were detracting from OpenStack's reputation.

"I think it's reality check time for OpenStack and that's long overdue," he says. "Somewhat controversially, we have long said OpenStack needs to focus on the fundamentals – we have chosen to focus on the fundamentals, which is virtual networks, virtual compute, virtual disks, virtual machines. That aspect of OpenStack and the people who focused on that are doing very well.

"But there's another element of OpenStack, which was essentially a lot of projects really driven by a single vendor, which was hoping to save themselves by getting a cloud story," he continued. "My rule of thumb is if you're not [creating] virtual networks, compute or disks, and you can't survive on AWS, you are never going to survive on OpenStack. That's the bullshit as a service story, and look what's happened since.

"HPE has laid off their entire OpenStack team, those were the guys who were pushing Ironic, Trove, Heat. That was never going to work. Mirantis laid off 300 OpenStack developers – those were the guys pushing Savanna, pushing Solum, pushing bullshit as a service.

"If I had advice for OpenStack at large, it's get rid of BS as a service," he continues. "The other stuff taints the reputation of OpenStack. When Heat fails, everybody can easily think OpenStack is failing- it's not, it's just bullshit as a service that's failing.

"When Trove fails, everyone thinks OpenStack isn't going to work. No, it's just Trove that isn't working. Those things don't matter. So if I could counsel the community and the Foundation it would be finally, for heaven's sakes, just be clear that this stuff in the middle, that's OpenStack, everything else doesn't matter."

Shuttleworth goes on to say that OpenStack as a whole is doing "great" and markedly becoming more scalable – with better quality, reliability and performance as it continues to mature in the IaaS space.

"But there's the sort of, oh my god, reality check happening all around the periphery as people realise that what customers want is the basics done really well at a really great cost," he says. Shuttleworth adds that a lot of the intrinsically valuable problems to solve are in reducing complexity towards everyday work – backing up data to the cloud, fixing broken disks, safely rebooting a hypervisor, realigning IP addresses in a virtual network, and so on.

"If you do these things the old fashioned way with Puppet, Chef and Ansible, they can be incredibly expensive because now you need the experts for everything," he says. "If you do them with Juju and Charms, you're sharing the cost of operational code with everybody else using those Charms."

According to Shuttleworth, there are parallels that can be drawn between the rise of software like OpenStack and the evolution of ever-increasing amounts of big data. Over the years, data itself slowly got bigger and bigger, and more rich and diverse, he says, and the complexity of it all grew until businesses understood they needed new ways to interpret, understand and use it.

"In the same way, the complexity of software has grown steadily and smoothly – microservices, cloud-scale, cloud-native architectures, none of these things happened overnight," he says. "There has always been one extra service – Hadoop went from two pieces of software to four pieces of software to seven pieces of software. OpenStack went from five pieces of software to nine pieces of software, all one step at a time.

"But there comes a point when the old way doesn't work anymore, and that's big software," he explains. "OpenStack was just the first. Now we have Kupernetes, we have IBM Watson, all the machine learning stuff – all the interesting stuff on the cloud is big software. You can't be successful with it by just understanding one part and automating one part.

"You have to understand 20 parts, and you have to understand them now because in six months they will have updated – it is the integration burden," he says. "If you do it the old way, that means hiring, training, writing code. That takes time. If the software has changed by the time you get through that process you are by definition failing."

While he remarks that, with companies like HPE you "reap what you sow" Shuttleworth is overwhelmingly positive about the OpenStack landscape and the community. It is clear, he says, that OpenStack has won the battle in terms of the open platform – and it's also evident that the big players in the entire technology industry are rallying behind it as a platform.

"The experts are here," he says. "If you think about a VM on-demand – I think Intel, IBM Power, ARM, those are the experts for how you should get a VM on demand. And they are here. If you think of disks on demand – EMC, Seagate, they are here. And if you think of networks on demand, Cisco, Brocade, Broadcom, they are here."

What is next for OpenStack in an environment that is trending towards multi-cloud, and where does OpenStack sit in the wider cloud landscape – next to public cloud, and so on?

"When you look at a public cloud vendor, it's difficult to disentangle the infrastructure as a service piece from the software as a service piece," Shuttleworth says. "OpenStack will never beat the public clouds at the software as a service piece, because you're falling further and further behind every year.

"Whereas the infrastructure as a service piece – once you get a VM of any operating system, on demand, attached to any network with any kind of disk, you are done," he explains. "So even though it's kind of late, OpenStack will be the internal infrastructure as a service of choice.

"I think it will sit alongside Azure Stack," he says. "I think Microsoft is very credibly focused on delivering a private version of its public cloud – so I think OpenStack and Azure Stack will sit next to each other very credibly inside the organisation. I'm pretty confident that for infrastructure as a service, OpenStack will be fine."


Copyright © 2016 IDG Communications, Inc.

Where does this document go — OneDrive for Business or SharePoint?