Citrix Defends CloudStack’s Value Vs. OpenStack

Grazed from InformationWeek.  Author: Charles Babcock.
 

In an April 18 column, I questioned whether we really needed advocates of one open source cloud project, CloudStack, saying it was superior to another project, OpenStack. My problem was in CloudStack’s proponents suggesting it’s the best because it’s the most compatible with Amazon’s cloud computing infrastructure.

"So Amazon-ness is the litmus test for cloud success, and CloudStack has got it," I wrote, citing comments by Sameer Dholakia, VP and general manager of cloud platforms at Citrix Systems. The piece, CloudStack vs. OpenStack Debate Rages On then countered with arguments by Randy Bias, CTO of CloudScaling, that an equal case on that score could be made for OpenStack. Other open source code, such as Eucalyptus Systems, might have a better case for Amazon compatibility on the basis of APIs, he said, and I agreed…

Since then, I’ve been hailed for a chat by a longtime acquaintance, Peder Ulander, a former senior VP of marketing at Sun Microsystems. Ulander was an effective spokesman for CloudStack when he was chief marketing officer of Cloud.com, which Citrix acquired last July. He and members of the CloudStack team were in Asia when this phase of the debate broke open. After he got back, he wanted to get in a few words.

Ulander started out by noting that Citrix exec Dholakia, on the stage at the CloudConnect show in Santa Clara last February, offered to turn over CloudStack code to the OpenStack project–and received an indifferent response from OpenStack project leaders. John Engates, Rackspace CTO and one of the organizers behind the OpenStack project, was sitting on the same panel as Dholakia at the time and shrugged off the move; however you may characterize it, it wasn’t exactly a warm embrace.

There was no doubt a bit of grandstanding in Dholakia’s move. He knew CloudStack had diverged from OpenStack, and OpenStack developers did not wish to proceed down the exact same path that CloudStack had already trod. OpenStack developers had their own initiatives underway. Nevertheless, Dholakia’s gesture was an expression of a willingness to cooperate, one made well before Citrix turned over the code to the Apache Software Foundation to create a second, and some say rival, open source project.

Ulander says CloudStack "was conceived by Cloud.com from the ground up to be like Amazon" in the way it functions as a cloud service, even though that doesn’t mean strict compatibility. The goal was to give enterprises and service providers the means of building a cloud service that could be easily made to work with Amazon’s, with a similar architecture.

CloudStack is serving as the basis for services used by GoDaddy, BT, NTT, Tata, Korea Telecom, IDC Frontier, LogicWorks, DataPipe, and 22 other customers, most of whom have shown in the field how a CloudStack-based service can work with Amazon’s EC2. For example, the CloudBridge translater in CloudStack understands AWS’s APIs and converts them to be recognizable to a CloudStack architecture. If you’ve used up the storage in your CloudStack-based private cloud, you can turn to Amazon’s S3 cloud storage.

"As we went down the customer priority list, two [items] stood out: build for scalability and Amazon compatibility," said Ulander. That’s what developers concentrated on at VMOps, as Cloud.com was known before it emerged from stealth mode in May, 2010.

Dholakia, Ulander and other CloudStack backers prefer not to be viewed as disruptive of a unified front to create an open source alternative to Amazon, VMware, Microsoft, and other proprietary cloud service vendors. They’re not, when they’re just supporting their own initiative.

But what’s dubious about the Apache CloudStack project launch is the claim that it is somehow the one viable code base on which to base future private cloud/public cloud combined operations. They potential of hybrid operations is key to private cloud builders’ decisions. Neither project has earned an exclusive right to that mantle.

The potential of cloud computing is huge and in theory, there is room for two open source projects in the field. But the fact is, two open source projects with similar goals tend not to persist side by side indefinitely. At some point, the bulk of open source developers decide that one of the projects isn’t accomplishing something new and distinct, and shift their energies elsewhere. As trouble brewed inside the JBoss application server project, for example, dissidents took their talents elsewhere, and eventually founded the Geronimo project inside Apache. But JBoss held together and a shortage of developer talent kept Geronimo from pushing to fruition in the marketplace. Likewise, the Java-like project, Harmony, attracted good talent but lost impetus when Sun reluctantly made Java open source code.

Ulander said not to worry, Citrix is willing to increase its investment in CloudStack development and it already possesses a more mature management interface between CloudStack users and services than what’s available in OpenStack.

Basically, Ulander makes the case that CloudStack and OpenStack are different enough that there’s room for more than one open source project. Making a core product, such as CloudStack open source code through the Apache Foundation is how Citrix wishes to compete. In the end, Ulander said, it’s not a case of CloudStack versus OpenStack. It’s a case of Citrix supplying an open source alternative "to the big, proprietary stuff," meaning primarily VMware with its vCloud product line. One way for Citrix to compete as a virtualization vendor is to undercut the growing list of VMware-compatible public clouds, including SoftBank, Singtel, Colt, Terremark, and BlueLock.

It’s hard to disagree, but for better or worse, CloudStack has not attracted as broad a support community as OpenStack. Could it be the Citrix close association with Microsoft, which has its own cloud priorities, that prompted first, HP, then IBM and Red Hat to make up their minds and join OpenStack?

Ulander points out that Citrix brings its own community of 50 partners to the CloudStack project at Apache, including Intel, Alcatel-Lucent, Juniper, and Brocade.

That may be, but in its proposal to the Apache Software Foundation, Citrix acknowledged that most of the CloudStack code came from Cloud.com developers who are now Citrix developers. Two of those key developers have departed. Citrix cited 30 outside contributors, but that’s a paltry figure compared to major, thriving open source projects, which will have several hundred contributors per release.

"Most of the current code contributors are paid contributors, employed by Citrix. Over the past six months CloudStack has received several contributions from non-Citrix employees…" the proposal stated. CloudStack was open source in name prior to the Apache submission but was essentially a project directed by and paid for inside Citrix. The Apache Good Housekeeping seal of approval may change that. But the proof will be in the diversity of companies backing a project with active code reviews, bug fixes, and code contributions.

For whatever reason, OpenStack has rapidly gained a critical mass on that backing. Its latest members include veteran open source project backers Red Hat and IBM. It already had the backing of Cisco Systems, Intel, AMD, NASA, and Dell, as well as cloud startups Nebula, CloudScaling, and Piston; all in all 165 companies are community members. On the crucial metric of contributors, OpenStack outstrips CloudStack.

An analysis of committer server log files shows 226 contributors to the latest release of OpenStack. Rackspace, Piston, and Nebula, all companies with OpenStack products, figure prominently, but so does HP, Red Hat, and Citrix itself. Also contributing, developers from Cisco Systems, SUSE, Canonical, and Dreamhost.

So we have two open source projects in a broad field. There’s no harm in that. I would just caution that in previous cases of competing projects, the early dominant project has tended to become thoroughly dominant in the long run, as long as the community around it was sustained. JBoss overwhelmed Geronimo; Sun, releasing Java as open source, took away much of Harmony’s reason to exist.

I hope OpenStack and CloudStack will evolve with distinct goals, gain equal but separate reputations, and maintain themselves in independent communities to provide choice for cloud builders. I hope so, but I’m not betting on it.