The Great OpenStack Debate: A Powerful Platform That Will Revolutionize The Data Center? Or A Dud?

A month ago, the OpenStack community was jolted by the news that Nebula, perhaps the hottest of vendors in the early days of one of the most important IT technologies on the market, was closing its doors.

Nebula's demise immediately sparked a larger debate. As with any young and emerging technology, OpenStack boosters and critics rushed to explain (or explain away) an event that could be interpreted as a harbinger of greater problems.

Opinions as to what went wrong at Nebula are diverse. But most pundits -- be they vendors, partners, developers or analysts -- seem to favor one of three explanations:

1. There was a problem with Nebula's specific business model and nothing more.

Sponsored post

2. The stampede of large software vendors into the open-source project is threatening the viability of independent vendors and pure-play startups.

3. OpenStack's organizational structure and technology have systemic flaws that will inhibit wide-scale enterprise adoption.

As insiders debate these hypotheses, the cloud-building framework that was launched in 2010 at the OSCON open-source convention is rapidly gaining influence across the industry. Sales of both hosted OpenStack services and licensed products are expected to grow this year by roughly 70 percent, according to a recent 451 Research market survey.

Software giants such as Hewlett-Packard, IBM, Red Hat and VMware that have jumped deep into the project over the past couple of years are, at the same time, irrevocably changing the distribution landscape and the private cloud marketplace.

While some purists still bemoan their involvement, the imprimatur of those name-brand vendors has signaled to the business community, and to the channel, a mainstream acceptance -- even a sense of inevitability -- that OpenStack is the open-source solution that will knock proprietary cloud vendors, public and private, off their perches.

Impressed with that kind of revolutionary fervor, many solution providers have hitched their wagons to the technology, and to particular vendors.

Which makes recognizing the explanation that best accounts for Nebula's demise critical to the channel.


Few disagree with this assessment.

Nebula was the first of the red-hot OpenStack startups. The company out of the gate raised nearly $40 million in venture capital from prominent investors. Its OpenStack pedigree was impeccable, with former NASA CTO Chris Kemp at the helm. And Nebula promised what everyone wanted -- an easier way to install and manage the complex cloud operating system.

But the company insisted on introducing OpenStack into the enterprise data center with a dedicated appliance, and that approach was doomed, some industry insiders tell CRN.

Jesse Proudman, founder and CTO of Blue Box Group, a hosted private OpenStack cloud, told CRN that Nebula had a few key challenges that aren't endemic to OpenStack at large.

"As one of the first companies in OpenStack and arguably one of the greatest contributors during the early years, Nebula paid a high price when the open-source community and market demand blossomed around them," Proudman told CRN.

That success detracted from the "walled-garden value proposition" of the startup's appliance, said Proudman.

Nebula also chose a business model demanding not only distinct expertise in OpenStack, but also in custom hardware engineering and supply chain -- a huge scope of responsibility for a startup, Proudman said.

It's a sentiment echoed by Boris Renski, CMO of Mirantis, the largest of pure-play OpenStack vendors in terms of both market share and contribution to the project.

"They did an enormous amount of good things for OpenStack," Renski said in praise of Nebula's contributions. "They just took a bunch of wrong turns. It's unfortunate."

The reasons Nebula failed don't bear on the viability of OpenStack, or on Mirantis, he said.

Instead, Nebula's business model failed to capitalize on the most important drivers for OpenStack adoption: flexibility and pluggability. The appliance architecture lost sight of those goals, Renski told CRN.

"If you look at capabilities that OpenStack offers, it's not that it's better than VMware or so different from CloudStack," Renski told CRN. "It's that it's a pluggable and flexible model with hundreds of drivers that are being made by vendors. So if you have heterogeneous data center infrastructure, overlay OpenStack on top of it, or if you are a greenfield, you still have control over what you're buying."

Open-source code alone doesn't deliver on those marketable benefits, Renski said. Potential customers saw Nebula as an appliance that was effectively competing with their own infrastructure.

Renski noted that some vanguard companies invariably fail when markets mature and industries shake up. That doesn't mean there are systemic problems within the industry, he said.


Nebula closed shop not long after two of its competitors were snapped up by behemoths looking to make their own cloud plays.

Cisco acquired Metacloud in September, and the following month EMC bought Cloudscaling. With three of the early OpenStack startups now out of the pack, some question whether their peers can survive in a market increasingly dominated by large, diversified vendors. And if not, do pure-play vendors need to start looking for safe harbors?

Jim Morrisroe, CEO of Piston Cloud, another early OpenStack startup, told CRN that small shops can't ignore the seismic shifts caused by some of the world's largest software companies dedicating unprecedented resources and tapping built-in channels and customers.

"If you look at what OpenStack was four years ago, when pioneers like us and Mirantis and Nebula got started, the goal was to have a cloud platform for companies to build modern applications and have modern workloads and use cases," Morrisroe told CRN.

"What it has become for the most part today, with the joining of the community in the last two years of big incumbent IT vendors, is something different," he said.

The focus of the project, he said, has shifted to delivering operational efficiencies for existing data center infrastructure, much of it not geared for the cloud. It's a new dynamic that does challenge those pioneering vendors, and they must respond by offering a more complete set of solutions, according to the CEO.

"Folks who have been focused in cloud-native are to some extent swimming against the current within their community," Morrisroe said.

It's a process Piston Cloud embarked on with its latest release, which adds new tools unrelated to OpenStack for deploying workloads on bare-metal or in Linux containers. With that release, the company is no longer a pure-play OpenStack vendor, Morrisroe said.

Nebula, with its dedicated OpenStack appliance, didn't appreciate that the community had dramatically shifted, he told CRN.

Al Gillen, a server and system software specialist at research firm IDC, told CRN his sense is that, in the long term, it's going to be difficult for a niche player to only sell OpenStack. The technology is already becoming a feature of larger offerings from name-brand vendors such as Red Hat and VMware.

"When it gets integrated into a larger, more comprehensive set of software offerings, how do you compete when you offer only one segment of the software stack?" Gillen asked.


OpenStack's meteoric market penetration was partly due to sharp marketing at a time when the cloud operating system was vying for recognition as the next big thing with other open-source solutions like CloudStack and Eucalyptus.

Rackspace, the first hosted service to adopt and help develop OpenStack, initially carried the promotional baton. With the project's connections to NASA, OpenStack's boosters were successful in hyping their software as the most-promising alternative to Amazon Web Services and VMware.

It's made for a uniquely interesting market, IDC's Gillen told CRN.

"From the moment it started, it has had more excitement and more backing than almost anything," Gillen said. "It went from zero to being not quite viable, but close to being viable, in just about a year or two. And we've been hovering around this almost-viable status ever since."

The impediment to achieving that last step of technical aptitude is that OpenStack is not really one project, but a collection of projects -- more than a dozen, all maturing at different rates, Gillen said.

"Some are more ready for prime time than others. If one revs, another doesn't, we have a potential for conflicts," he told CRN.

The OpenStack Foundation tries to address that issue by doing a full-scale release every six months to force all collaborators into a compatible state. But "everything is moving so quickly it's hard to get your arms around a relatively stable configuration you can actually stand up and use for a long period of time," Gillen told CRN.

That makes life tougher for vendors that have committed to OpenStack as their flagship product. Enterprises, when they buy software, typically need it to chug along three to five years to see their return on investment. OpenStack is churning too quickly to be consumed at that pace, Gillen said.

Andrew Shafer, senior director of technology at Pivotal, told CRN the problems with OpenStack run even deeper.

An early OpenStack engineer, Shafer was vice president of engineering at Cloudscaling, the OpenStack startup bought by EMC. He's a vocal critic of the project's management structure and the technology it has produced. OpenStack was open-sourced too soon and it's inherently unsteady, he told CRN, only operating efficiently in best-case scenarios.

Efforts to create an inclusive community have compromised the quality of the technology, Shafer believes.

"There is no stewardship. There was no one that protects the quality of the project or protects the experience of the user," he told CRN.

Shafer agrees with the analysis that Nebula was delivering a product the market didn't really want. He also believes the challenges are immense for companies like Nebula.

"To get big enterprises to bet their whole IT future on your startup, that's just a hard sell," he said.

But ultimately, "the inherent quality of OpenStack as a product made it hard to build businesses. It's a shaky foundation," Shafer told CRN.

OpenStack's technology got consumed by marketing and politics far too early in its development process, according to Shafer.

The fundamentally difficult distributed-state problems encountered when attempting to scale across large infrastructure -- thousands of server nodes -- were never quite ironed out. And no one is working on them anymore, he said.

As for the software giants that have become some of the largest individual contributors, they're mostly just making OpenStack compatible with their other products, Shafer said, adding, "it's not the same thing as really developing and stabilizing the core product."

It's not impossible for businesses to manage their own OpenStack-powered infrastructure, he said, but they need to be committed to overcoming a steep learning curve.

"OpenStack can be made to work. You just have to constantly feed it with human blood and never turn your back on it," Shafer said. "It's always by force of will."


Solution providers looking to ally themselves with promising technologies are used to hearing wildly diverse opinions about products, which makes sense when considering the broad range of interests and perspectives of the pundits.

Depending on whom you ask, OpenStack is either the powerful open platform that will totally revolutionize the data center, or a dud.

What all agree on is that OpenStack is a handful -- a uniquely complex piece of software that's evolving rapidly and requires great tools and lots of talent to stand up and manage.

Blue Box's Proudman sees the OpenStack market maturing to one in which buyers want either a highly curated distribution delivered by a trusted enterprise vendor, or a service provider that can insulate them from ever having to interact directly with the platform.

"Both are positive signs of maturation and growth, and 2015 looks bright," Proudman told CRN.

But even if that prediction is true, there are still hurdles to clear for the pure-play vendors that have been the lifeblood of the industry.

"The bigger challenge is, where do they go longer term as the technology goes mainstream and starts to be subsumed by the technology around it?" asked IDC's Gillen.

For software companies to survive in the coming market -- which may still be a few years off -- they will need to think of themselves as developers of top-to-bottom cloud solutions and not just OpenStack. Otherwise, they won't be able to compete with vendors that control complex portfolios and have integrated OpenStack across multiple layers of the stack, he said.

"Basically, they need to stand up a whole platform as a service environment," Gillen told CRN. "And right now that's not what's happening."

But as more OpenStack startups diversify their offerings, while others inevitably exit the market as Nebula did, will the technology itself become more advanced, stable and friendly? Or will its flaws just become more pronounced?

"Everyone's trying to add value and differentiate their products," Shafer told CRN. "But the core functionality has sort of been neglected."

This article originally appeared as an exclusive on the CRN Tech News App for iOS and Windows 8.