Open source communities are fluid, dynamic places, whose output is liable to change based on how the individual contributors’ commitment and involvement waxes and wanes over time.
In the case of OpenStack, and its pool of contributors and supplier partners, any hint of a company opting to downsize their involvement is often seized upon by industry watchers as a sign the wheels are coming off the open source cloud juggernaut.
It is a conclusion that is easy to jump to in response to reports of Cisco shutting down its OpenStack-based public cloud, HPE selling its OpenStack assets to Linux distribution provider Suse and Intel withdrawing support for a Rackspace-backed push to boost the enterprise appeal of the platform.
But, just because a conclusion is easy to reach, does not automatically make it right, and – in the case of OpenStack – such changes should be viewed as a sign of a community coming of age, rather than dying out, its supporters claim.
“There’s been a lot of fake news about problems in OpenStack,” Scott Crenshaw, senior vice-president and general manager of OpenStack clouds at Rackspace tells Computer Weekly.
“Those problems are driven by suppliers having the wrong model. They’re trying to take a complex cloud infrastructure and sell it like it’s a simple operating system. So when you see supplier after supplier leave, that’s a reflection of their own problems rather than OpenStack’s.”
The OpenStack community of contributors is, effectively, following the same evolutionary model as Linux did 20 years ago, but the pace of change is a lot faster, says Crenshaw. “I remember the days when Linux had over a hundred suppliers involved and now, depending on how you count it, there are one, two or three. That’s not a bad thing, and is reflective of a vibrant ecosystem and community,” he says.
Read more about OpenStack
- The OpenStack Foundation is on a mission to clear up a number of misconceptions about the open source cloud platform, particularly those pertaining to its often predicted demise.
- Enterprises have come to view private cloud as a “toxic” phrase, but new ways and means of building deployments is changing that, according to the OpenStack Foundation and its stakeholders.
Putting OpenStack’s detractors straight emerged as a recurring topic of conversation at the bi-annual OpenStack Summit in Boston, which took place between 8-11 May 2017.
OpenStack Foundation executive director, Jonathan Bryce, kicked off the event’s opening keynote by hitting out at the doomsayers, and the pleasure they seem to take in running down OpenStack, and predicting its demise.
“It’s kind of a hobby. People love to make these comments and make these predictions, and they do them year after year, but those are opinions – what I really like to base my plans and decisions on is facts and data,” he says.
Indeed, the foundation’s most recent user survey highlighted a 44% year-on-year rise in the number of OpenStack deployments, with two-thirds of users reportedly now using it to handle production-grade workloads.
While these figures certainly help counter the view that the wheels are coming off of OpenStack, the survey also uncovered a rise in the number of users expressing dissatisfaction with using the platform, particularly amongst those running deployments created between 2010 and 2014.
OpenStack deployments showing signs of improvement
Meanwhile, users operating OpenStack deployments created since 2016 reported higher levels of satisfaction. Crenshaw admits there may have been a number of enterprises, in the early days of OpenStack, who bought the software to power their private clouds and had their fingers burned because of a lack of supplier support.
“Most of the suppliers in this space historically have been trying to sell OpenStack cloud software as if it were an operating system, and not as a cloud. When you’re selling an operating system, you just give it to the customer and they know how to operate it successfully,” says Crenshaw.
“So all these suppliers just dumped the cloud software on their customers, who took the software, tried to operate it, and they struggled.
“A large part of our private cloud as a service business in the early days was generated by what we termed ‘burn victims’, which are companies who tried and failed to run their own clouds,” he added.
The situation has improved since then, with various suppliers – including Rackspace – now offering managed services around OpenStack to solve the ease of use issues some early adopters ran into, while the availability of IT staff with skills in using the platform has also grown over time.
Getting to this point has taken a while, Crenshaw concedes, in part because of how the OpenStack community operates and conducts itself.
“Because the community is open and perceived to be a neutral place, hundreds of suppliers and thousands of contributors jumped on the platform, and we saw a rate of innovation that allowed us to move to enterprise class production twice as fast as Linux,” he says. “The downside is it has been a chaotic community, meaning things like ease of use have not been dealt with as rapidly as anyone would like.”
This is a view point Canonical founder, Mark Shuttleworth, shares, saying the way the community has attempted to position Openstack as “all things to all men” in the past more than justifies some of the criticism fired its way.
Canonical is responsible for the Ubuntu Openstack distribution, which has been keenly adopted by the telco community, retailers, financial services companies, media organisations, with its list of reference customers extending to Netflix, PayPal, and Sky, to name a few.
“The mission of OpenStack that’s really important is virtual machines, virtual disks and virtual networks, on demand with great economics,” he says.
“We’ve always been contrarian in the OpenStack community because when everyone else was saying we should do everything, I was saying just do these core things and do them well. Get your operations right, get your costs right, get your integration right and make it something people can just consume.”
Attempts by the community to “do everything” have fuelled the development of unnecessary features and standalone OpenStack projects that users struggle to see the value in, he continues.
“The criticism is because all of that complexity is just unnecessary. People are saying, ‘I don’t understand this bit and I don’t understand that bit’, and the truth is those bits need to go. That will happen. It may not be a painless process, but in the end OpenStack emerges stronger.”
The need to trim the fat from OpenStack, from a project and feature perspective, is clearly not news to the OpenStack Foundation, as Thierry Carrez, its vice-president of engineering referenced this as a priority during the Summit’s keynote.
“This is something we are taking to heart. We have started to more actively identify areas where we can address this complexity in projects by removing unused features or pruning extraneous configuration options, or culling projects that are not going anywhere,” says Carrez.
Betting on OpenStack
While efforts to lower the barriers to OpenStack have and will benefit some organisations, others have managed to work round them, including online betting giant Paddy Power Betfair.
The company is in the throes of a multi-year migration project, which – to date – has already seen it move more than 25% of its legacy infrastructure stack to OpenStack over the course of a year.
“We’ve automated everything using ‘everything is code’ methodology and we’ve created a self-service infrastructure pipeline the developers can use to put their applications onto the platform,” Steven Armstrong, principal DevOps automation engineer at Paddy Power Betfair tells Computer Weekly.
“We’re looking to put everything onto the platform, moving all our workloads over to OpenStack and that project will run for around a year and a half.”
As an example use case, Armstrong says the company’s customer platform is running on OpenStack now, which allows its users to access either the Paddy Power or Betfair online sites.
“With OpenStack, we wanted AWS-like agility but in the private cloud with our own datacentres. We wanted to have and be able to programme each of the different tiers; the storage and networking, all through a set of common set of APIs,” says Armstrong.
“This means, if we got with a different networking or storage supplier at a later date, we could just swap them in. That was really important for us.”
Relying on a small team
The company runs a Red Hat-based distribution of OpenStack now, and relies on a core team of just eight engineers to run it, says Armstrong. “They’re dealing with all of the deployment pipelines. For us, it’s quite easy to run, but you do need a highly-skilled engineering team to do it, which we have,” he says.
Initially, the company entertained the idea of getting a company, like Rackspace, to host OpenStack for them, but strict limitations on the number of compute nodes that could be accommodated through this model at the time put paid to that.
“RackSpace now run a Red Hat distribution and offer that as a managed service, which wasn’t an option for us when we started the project a few years ago,” he says.
For some organisations, such challenges may represent something of a stumbling block for their private cloud ambitions, says Armstrong. “It is not their core business and it’s not how they make money, whereas we wanted to use it as an innovation platform.”
Ignoring negative attention
On the topic of detractors, Armstrong says the negative attention OpenStack receives is best ignored, and has had no bearing or influence over his company’s ongoing investment in it.
“There are a lot of detractors because they’re worried about it. They see it as a threat,” he says. “A lot of people are bought into public cloud so they like to bash anything that is private cloud because they want people to move to public cloud and, generally they lock them into the public cloud.”
OpenStack is swiftly emerging as the industry’s preferred means of consuming private cloud, which may be a difficult pill for some to swallow.
“AWS provides great agility, but it is also a lock-in, whereas OpenStack isn’t because it’s open source, you can see the code, contribute to the code and control the destiny of the platform,” says Armstrong. “That’s where the industry is going. If something isn’t open, people are suspicious about it.”
Read more on Infrastructure-as-a-Service (IaaS)
Canonical founder calls out OpenStack suppliers for ‘lack of focus’ on datacentre cost savings
Redefining OpenStack: Addressing identity and integration for enterprise-readiness
How OpenStack is coming of age and tackling its growing pains head-on
Managed OpenStack private cloud gains support vs. roll-your-own