Over the past couple of years, OpenStack has turned from a small skunkworks effort to build Amazon Web Services-like clouds but open source into a movement backed by some of the IT industry’s biggest legacy vendors. It’s not uncommon nowadays to hear that OpenStack has become the de facto standard for building cloud infrastructure.
Such mainstream support, however, comes at a cost, threatening to detract from the project’s original goal. That’s according to Jim Morrisroe, CEO of Piston Cloud Computing, a San Francisco-based startup co-founded by Joshua McKenty (one of OpenStack’s founding fathers) that helps customers stand up OpenStack clouds of their own.
Morrisroe believes the open source project’s original mission was to help companies build clouds like the ones Amazon has built: web-scale, hyper-converged, running on cheap commodity hardware. But now that the likes of Cisco, HP, and IBM have gotten into it, they’ve been busy building OpenStack plug-ins for their proprietary hardware, diluting the focus on that homogeneous web-scale future.
“Amazon doesn’t buy Cisco, and EMC, and Hitachi,” Morrisroe said. “Anybody that has a box business – proprietary hardware – has jumped into OpenStack to try to get some new life into their proprietary boxes. And so OpenStack, the community, has gotten a little too focused on all of these plug-ins and drivers.”
Juno, the latest OpenStack release that came out in October, was a case in point. It included 10 new drivers for block storage systems by the likes of Fujitsu, FusionIO, Hitachi, Huawei, and EMC. IBM, Mellanox, Juniper, and Brocade all contributed plug-ins for Neutron, the software-defined networking portion of OpenStack, in the release. Cisco added a Neutron plug-in for its Application Policy Infrastructure Controller technology.
Diverse Goals May Be a Good Thing
But the APIC plug-in isn’t the only thing Cisco added in Juno, which illustrates that the issue isn’t quite as black-and-white as it may at first seem. Some of the big vendors do make a lot of contributions that are not exclusively self-serving.
HP, for example, is one of the top contributors to the project. The vendor has a huge vested interest in making sure core components of OpenStack work, since its entire cloud services strategy revolves around the open source technology.
“Of course we want to sell HP hardware,” Bill Hilf, senior vice president of products for HP Helion (the name of HP’s cloud business), said. But, the Helion OpenStack distribution supports other vendor’s hardware, not just HP’s.
There are many vendors in the community, each trying to differentiate based on its line of business, and that’s a good thing, Hilf said. “You actually want multiple types of business models to be in the ecosystem.”
To Each His Own
There isn’t a one-size-fits-all OpenStack cloud, obviously. Whether a company buys commodity x86 boxes to deploy its cloud or tries to bolt it onto existing hardware depends on its individual circumstances.
Mark Baker, Ubuntu server and cloud product manager at Canonical, the open source software powerhouse, said he saw big players gravitate more toward commodity infrastructure for their OpenStack clouds, especially companies in the Far East. Financial services and media companies, on the other hand, tend to favor their trusted providers like Dell or HP. But they buy hardware from those trusted vendors’ commodity lines.
Enterprises and service providers typically buy new hardware to build their clouds, while telcos have huge amounts of infrastructure already in place, a lot of which is telco-specific, non-x86 hardware. They are keen to use what they already have.
“All customers have complex heterogeneous environments,” Baker said. “If they have EMC storage, they want to be able to connect [cloud] with EMC storage. If they have a big investment in Cisco or Juniper, than they’ll want to be able to use that technology.”
A Vision for the Future
Piston is proposing an infrastructure vision for the future, but many companies are simply not ready to make the leap just yet. Morrisroe said vendors realize that the future data center is one where hardware is cheap and homogeneous, its functions defined by software, but getting to that future without going out of business is a challenge for them.
CIOs want to put OpenStack in front of their legacy gear because their developers have gotten too used to going around the IT department, renting cloud VMs from Amazon with their credit cards, he said. If they make their own platform more agile, they can stop the “bleed” to the public cloud.
But agility alone won’t cut it. “No matter how agile you make this platform, it still has a cost of goods that is radically different than the construct of the cost of goods on the public cloud side,” Morrisroe said. An agile platform on legacy systems doesn’t address the ultimate problem, which is that the web-scale cloud is not only fast and fungible, it’s also infinitely scalable and more cost-effective, he explained.