Growing pains for the cloud's open source platform

OpenStack vendors are increasingly claiming that the open source cloud platform has reached maturity - but is it truly ready for the enterprise market?

For the uninitiated, OpenStack is one of a few open source cloud platforms available to organisations that don’t want to rely on proprietary cloud platforms. This could include companies wanting to deploy their own private clouds or cloud service providers looking to offer public cloud computing services to their customers.

OpenStack was launched in 2010 by Rackspace and NASA in 2010, both of which contributed code to the first iteration. The project is managed by the OpenStack Foundation, an organisation formed in September 2012.

The OpenStack code is freely available under the Apache 2.0 licence and is open to contributions from the general public. According to the project’s official website, “an open development model is the only way to foster badly-needed cloud standards, remove the fear of proprietary lock-in for cloud customers, and create a large ecosystem that spans cloud providers”.

From a distance, OpenStack seems quite impressive. More than 100 organisations - including major vendors such as IBM, Cisco, Dell and Hitachi - contribute to the project. It also features some pretty impressive users, such as Harvard University, Sony, PayPal and even the US National Security Agency.

OpenStack has come in for criticism in the past, however. These criticisms included, but were not limited to, supposed community infighting and politics, a lack of enterprise features and low-quality contributions from the community.

Recent times have seen OpenStack’s proponents claiming that the project has emerged from its difficult teething period and is now (or is on the verge of being) ready for the enterprise.

But some commentators believe problems remain.

Kevin McIsaac, analyst at IBRS, says that implementing OpenStack is labour intensive and costly.

“Most people I talk to about this [say that OpenStack] takes an awful lot of people and a lot of money to implement and make work. People are saying that it’s still quite amorphous, there are many distributions of it, and implementing it and making it work actually becomes quite expensive and quite complicated,” McIsaac says.

In a report titled ‘An overview of OpenStack, 2014’, Gartner analyst Lydia Leong writes: “OpenStack requires significant expertise to implement and operate for anything other than a trivial nonproduction deployment. Do not expect that you can simply send an engineer on an OpenStack training course in order to acquire the necessary skills. When considering the cost of an OpenStack solution, ensure that you factor in the labour costs.”

When asked about reports of burdensome implementation, Mark Potts, CTO of HP’s Cloud unit, says his own company’s public cloud efforts provide a case study for OpenStack’s progression.

“We’ve been running a public cloud at scale based on OpenStack, going all the way back to Diablo [an OpenStack release from late 2011]. The maturity around some of that in Diablo and [subsequent version] Essex was problematic. It was a very new project, that’s for sure. If you compare [the even more recent] Icehouse now to Diablo, it’s markedly different in terms of the experience, robustness and scalability,” Potts says, adding, “Now that doesn’t mean there is not work to be done.”

Potts points to HP’s hand in advancing the project’s TripleO program as an example of how OpenStack has developed over time.

“The TripleO project is really about how you stand up an OpenStack-based cloud environment, and then how you manage the life cycle of that. Because one of the things customers want is that innovation and that fast cycle, but that can be very, very disruptive if you’re upgrading every three to six months. So we’ve been spending a lot of time on TripleO and the installation to make it a lot easier,” Potts says.

HP runs all of the infrastructure for OpenStack itself, he says, including contributions, testing and CICD (continuous integration/continuous deployment).

“We’ve learnt an awful lot. What we’re doing with our distributions is bringing a lot of that knowledge and understanding from the public cloud, from running the OpenStack project and distribution in a CICD process, in to TripleO,” he says. “[And also] supporting that for customers as they start to think about adopting and building out and having to deal with releases and upgrades at that type of frequency.”

Potts does concede that OpenStack “still has room to improve”, but he maintains that OpenStack is now “very mature”.

“HP runs a public cloud at scale. Rackspace runs a public cloud at scale, on top of OpenStack. We both run on the newest versions, and we would not have the service level agreements that we can put in front of customers if OpenStack was still as immature as it was a year ago.”

Locked in?

Gartner’s Leong is not so sure of OpenStack’s maturity. She notes that OpenStack is a framework divided into multiple components.

“Because components can, to some degree, be adopted independently of one another and each project has its own technical road map and leadership, OpenStack projects vary in their maturity, speed of development, development community participation and level of customer adoption.

“Although significant efforts have been made to stabilise the most widely adopted OpenStack components, OpenStack as a whole should be considered relatively immature software and new projects continue to be added to the integrated releases,” Leong says.

There’s also doubt about whether the open source nature of OpenStack truly does away with vendor lock-in.

McIsaac says: “You’ve heard these stories [about interoperability] from vendors before. ‘You take my magic pixie dust, rub it on some other stuff, and now it’s good.’ And to a degree those claims really are magic pixie dust.”

According to the analyst, we’re likely to see increasing vendor lock-in as time goes on.

“That was kind of the Unix [promise] as well: open systems and no vendor lock-in. Well, no. The lock-in is less severe in the case of Unix, and even less with the case of Linux. But you go down a Red Hat Linux route, there’s a degree of lock-in. There’s always lock-in,” he says.

Gartner’s Leong feels similarly, writing: “OpenStack is open source, but that does not equate to open standards, broad interoperability or freedom from commercial interests.

“Interoperability across OpenStack-based clouds, whether public or private, is currently challenging and, on a practical basis, limits portability between different OpenStack-based clouds as well as the ability to use OpenStack for hybrid cloud solutions. At present, there is no less lock-in for customers adopting OpenStack than for those adopting proprietary CMPs,” she writes.

But lock-in isn’t necessarily a bad thing, according to McIsaac.

“You don’t want to avoid lock-in. What you actually want is to strike the right balance between lock-in and value. Because if you’re not committed to anything, you probably won’t get any value out of anything either. And there’s huge value to be had from being locked into the right stuff.

“Microsoft, for example. You can get a lot of value out of the Microsoft stack, and you are firmly locked into it, but if you really exploit the features and the benefits, you can drive enormous value. What’s the problem with the lock-in? Well there are problems - contract negotiations, for example - but if you’ve got such great value out of it, who cares?” he says.

But don’t be cavalier - lock-in may be the price for getting value, but that compromise must be made with full awareness of what you’re getting out of it.

“You need to be very careful where you do get locked in. There’ll always be some layers you get locked in to. Be clear what they are, and be very clear about how you drive value when you’re locked in, and how you might get away from it at some other stage,” McIsaac says.

Parallels

On top of talking up the maturity of OpenStack, vendors are also employing some interesting rhetoric while plugging their OpenStack-based products. HP CEO Meg Whitman, for one, was widely reported as saying in May, “Just as the community spread the adoption of Linux in the enterprise, we believe OpenStack will do the same for the cloud.”

IBRS’s McIsaac is dubious of the parallels between OpenStack and Linux. OpenStack lacks a leader figure to keep the core of the project together, he says.

“Unlike Linux, I don’t see the equivalent of a Linus Torvalds, who describes himself as ‘a benevolent dictator for life’,” McIsaac says. Instead, “Each one of these corporations - like Red Hat and HP - is doing their own thing and they’re adding their own flavour to it, and then they’re building their own distributions, and then they’re trying to commercialise it.”

Rather than mirroring Linux, McIsaac believes OpenStack is instead a repeat of the Unix story from the ’80s, “where the big vendors all went off and took a great idea - Unix - and then added all the things they thought were necessary, and then made it commercial and proprietary to themselves. So Unix was still Unix, but everybody’s flavour was different and moving between them was quite hard.

“Whereas Linux, even though there is a bit of a split - Red Hat versus Debian, for example - that level of split is much, much smaller, because you’ve got the benevolent dictator for life who controls the core of Linux,” McIsaac says. OpenStack is “likely to be more fragmented like Unix, which was standard but was somewhat fragmented in implementation.”

Potts, however, believes that customers will be able to dictate how far vendors’ offerings stray from vanilla OpenStack, and therefore avoid the fragmentation of Unix.

“I think the customer base will demand that it stays close to trunk, and doesn’t become proprietary and closed, because they’ve been bitten by that before,” Potts says.

Hardball

Some vendors involved in OpenStack are copping flak for their particular approach to making a dollar from the project.

In May this year, in a story called ‘Red Hat Plays Hardball on OpenStack Software’, the Wall Street Journal reported that Red Hat’s rivals, partners and customers feared the company was using its dominance of the Linux market to block organisations from using OpenStack products from other vendors.

Specifically, the WSJ claimed that Red Hat had chosen not to provide support to its commercial Linux customers if they use rival versions of OpenStack. This denial of support could discourage customers from employing cloud products from other vendors, the line of argument went.

Red Hat reportedly said it would cost too much to support different versions of OpenStack, and that its cloud software needed to be tied tightly to its Linux distribution.

“Red Hat has taken the art form of closed open-source to a new level,” the WSJ quoted Martin Fink, the GM of HP’s cloud business unit. As the WSJ notes, HP sells and supports Red Hat’s brand of Linux, but competes with Red Hat in the OpenStack market.

The article also outlined Red Hat’s relationship with a company called Mirantis. The two companies partnered up, with Red Hat using Mirantis consultants and tools to help its own customers build clouds using its version of OpenStack.

But the WSJ said that after Mirantis began selling its own OpenStack software, Red Hat dissolved the services partnership, and the company’s senior leadership encouraged at least one large Red Hat reseller not to partner with Mirantis.

“The problem is that … the majority of [Red Hat’s] business is based on Red Hat Enterprise Linux, and of course what they’ve got to do is continue to push Enterprise Linux,” says HP’s Potts.

So, with OpenStack, Red Hat is “doing a very closed stack, where [host OS, virtualisation and OpenStack distro] are very tightly integrated, and you buy a package. Obviously they don’t want to support other Linux, because that’s not in their interests.”

Not all vendors are following that path, however. Potts naturally mentions HP’s own model as a comparison.

Red Hat has gone down a path with “a different business model, and it’s a different approach - a fully, tightly integrated stack that doesn’t have different layers and options”, says Potts. “We’ve gone for a different approach - we come with a fully integrated stack, but we actually have support for different Linux, and we have support for different hypervisors.”

Image: ©alphaspirit/Dollar Photo Club