VANCOUVER - OpenStack is enterprise-class, production-ready – sort of.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Several enterprises were prominently featured at the OpenStack Summit here this week. They heaped praises on the technology for large-scale deployments and its ability to do what other cloud technologies cannot, all while adding the important caveat that not all of its parts are equal.
For DigitalFilm Tree, a post-production, software and process consulting company in Los Angeles, OpenStack is used in production with Rackspace and others because it's the only truly mature technology for hybridized environments, according to CTO Guillaume Aubuchon.
"OpenStack is a conglomerate of projects and some of those are more mature and other aspects are less mature," Aubuchon said. "As far as I'm concerned, as of two years ago the balance of the projects being mature and stable and production-ready shifted in the right direction."
Companies such as Walmart, Comcast, Yahoo, PayPal and TD Bank were on stage during the keynotes to discuss their use of OpenStack in production at massive scale. The needs of enterprises and how best to utilize OpenStack were featured in breakout sessions scheduled throughout the week.
Enterprises come on board with OpenStack
There was near unanimity from attendees that OpenStack has improved by leaps and bounds in the past two years in nearly every category. The OpenStack Foundation made moves this week that should be attractive to more mainstream enterprises with improved interoperability and clarification of which projects meet its standards on various components.
The fact that Fortune 100 companies are adopting it in production is a clear sign that OpenStack is ready, said Lauren Nelson, an analyst with Forrester Research, Inc., based in Cambridge, Mass.
"The challenge is when people talk about whether it's ready for production, they look at it as a 'yes' or 'no' rather than looking at what are folks really doing when they're running in production," Nelson said. "It is being run in production in many use cases today; the question is for what environments."
One of the biggest mistakes companies make when they move to OpenStack is trying to move legacy applications the cloud, according to industry experts who say OpenStack is really suited to applications that are designed for the cloud and have a lot of resiliency built into them.
One of the biggest challenges facing OpenStack is that people come in with unrealistic expectations, said Randy Bias, vice president of technology at EMC and a member of the OpenStack Foundation board of directors, during a panel discussion. It's more than just compute, so there is an exponential rise in complexity of the overall system when you start adding in storage, networking, authentication, and application deployment and management.
"People have an expectation that they're going to take OpenStack, download it, sprinkle it in their datacenter and poof, there's Amazon - except it's just not that easy," Bias said.
OpenStack has moved from Web-scale and tech companies that were the early drivers to more technically aggressive enterprises that rely on IT to gain a competitive edge, according to Donna Scott a distinguished analyst with Gartner, Inc., based in Stamford, Conn.
"It hasn't gone mainstream," Scott said. "It's just too hard for the mainstream, but it's moving in that direction."
It will likely be another two to three years before that happens, and the progression toward mainstream could come in more software as a service-based or appliance approaches or more hosted private cloud, Scott said.
"It'll probably be more packaging of hardware and software together," Scott said. "Look for things that are easier to roll in for the mainstream and easier to consume."
OpenStack Neutron remains a challenge
While OpenStack has progressed, some projects remain problematic. But there's an unwillingness to kill OpenStack projects that cause a lot of grief, Bias said.
"If people hate them and they're hard to operate and they increase the complexity of the system, why aren't they just kicked to the curb?" Bias asked. "The reason is we feel like every single piece of code in OpenStack is precious, which is not how well-run companies do things."
The most common complaints about OpenStack were around Ceph, Neutron and updates. Neutron in particular caused some users to wince when mentioned.
FICO, an analytics software company in San Jose, Calif., uses Red Hat to deploy its OpenStack private cloud as opposed to one of the larger public cloud options because it's a more distributed architecture with a higher number of availability zones internationally, said Nick Gerasimatos, director of engineering and cloud services.
The workloads have been in production for 90 days and everything has gone smoothly so far, but there are still issues with some of the projects not maturing fast enough, Gerasimatos said.
"The extensibility of Neutron, it's more for flat networks and it's not great for large scalable infrastructure," Gerasimatos said. "A lot of that is supposedly changing in Kilo but we would have liked to have seen that a few versions ago."
DreamWorks partners with HP on its OpenStack cloud for development and testing and has an early prototype for some outward-facing media projects and internal cloud products that will be in production later this year. So far this year, Neutron has been at the root of all the problems the IT team has run into, said Skottie Miller, senior technologist, during an HP user panel.
"The recommendation from some of the cloud architects is [to] do networking redundancy and physical infrastructure and then maybe use VLANs for network subdivision instead of relying on Neutron today," Miller said. "It will get better and there will be a transition time in the future, but right now I wouldn't trust Neutron."
One way around the scale issues is to build a federated cloud and keep things as simple as possible, said Orlando Bayter, CEO and founder of Ormuco Communications, a Montreal-based telco that has gone the federated route.
"If you want to have more than 100 nodes you don't want to modify the OpenStack code, you don't want to go outside this pattern," Bayter said during the same HP panel. "You want to make sure you can put 10 OpenStack installations of 100 nodes to be able to go to 1,000 nodes in order to manage it."
Trevor Jones is the news writer for SearchCloudComputing. You can reach him at firstname.lastname@example.org.