OpenStack (Photo credit: Wikipedia)
Most open source projects are about software and only software. Developers create a software tool, language, specialized server, or an application, and people take it and run it on whatever computers they have handy, or in the cloud. In other words, the collaboration about the software being created is the main event.
But now because of the way OpenStack is applying the open source model to creating cloud infrastructure, the open source model is on the verge of being extended so that the collaboration and design process will include software, hardware, and networking in the data center as well as operational processes.
In modern development, the idea of designing software, data center, and operations using one integrated team is called DevOps.
It seems to me that within the next year, the OpenStack project, an open source community created by Rackspace to support cloud computing, will become the first explicit Open DevOps project, one where collaboration and knowledge sharing extend beyond software to include data center design and operational processes.
While the Open DevOps vision has not been formally adopted by OpenStack or any other related project, my view is that it will be. I suspect such a development will provide a stimulus that will force Amazon, a secretive company, to adopt a much more open way of guiding the massive ecosystem that is developing around its cloud computing services. Last week I argued that it was in Amazon’s interest to do this anyway. (See “Questions Amazon Should about its Cloud Strategy”) If OpenStack shows that the Open DevOps accelerates innovation, perhaps Amazon will open up even sooner.
Here is how I suspect this will play out.
OpenStack Becomes an Independent Community
The first step toward Open DevOps is that OpenStack must become an independent community. Up until now, OpenStack has been directed by Rackspace, which started the project.
On Friday of last week, Rackspace announced that formation of the OpenStack Foundation was moving off the drawing board into the implementation phase and a new set of platinum and gold partners had joined. John Engates, CTO of Rackspace, said the conceptual design of the community is complete and now the members are all working to hammer out the processes that will be used the run the community and the legal agreements that will provide the governing framework. This step is part of an effort that had been announced in concept in October to move control of OpenStack from Rackspace to a foundation run by a community of developers and participating organizations.
The new platinum partners for the OpenStack Foundation include IBM and RedHat, who joined AT&T, Canonical, HP, Nebula, and SUSE at this level, which requires that each company provide at least two full-time developers to the OpenStack project. Gold members include Cisco, ClearPath, Cloudscaling, Dell, DreamHost, ITRI, Mirantis, Morphlabs, NetApp, and Piston Cloud Computing.
In addition, on Monday 4/16, Rackspace announced that starting in May, it would be offering cloud services based on the OpenStack software for computing, databases, block storage, and networking, along with a control panel to make running and managing a cloud easier. Rackspace is also adding a cloud monitoring service that is not part of OpenStack.
Rackspace already runs the second largest public cloud after Amazon, but its current cloud services are not powered by OpenStack. The significance of this announcement is that, Rackspace is now offering a cloud in production on the OpenStack software. Engates said that while some of these capabilities such as computing and databases are on parity with Amazon and other cloud offerings, some of them are even better.
“The OpenStack Cloud Networks capability, which we will launch soon, has the benefit of a second-mover advantage,” said Engates. “We have seen the flexibility, control, and security our customers need and have created an API that meets those needs by allowing the creation of software defined virtual networks.” In addition, the Cloud Block Storage service offers an option to deploy using SSD storage for faster performance.
Some of these services are available in early access form so that Rackspace can control the volume of customers that are using the services. The Cloud Control Panel and Cloud Monitoring systems are aimed at making its cloud much easier to use, which Engates sees as one way to differentiate Rackspace from Amazon.
What this means is that OpenStack has become stronger as an ecosystem and that it is on the way to being a fully functioning open community like the Apache Foundation, the Linux Foundation, and the Eclipse Foundation.
What Open DevOps Means
So, let’s think for a minute about what the community members of OpenStack will be doing this week at their design summit, which happens twice a year. The members will be in San Francisco talking about how to create software that runs various parts of a cloud computing infrastructure. But you cannot really talk about the software in isolation. The software for cloud computing must be designed with a specific type of computer and network infrastructure in mind. In addition, to run a cloud computing server farm, you must have automated operations so that you can manage thousands of computers with a handful of staff.
To me what Open DevOps means is that the design of the data center and the operational processes must formally become part of the project. Jonathan Bryce, member of the OpenStack Project Policy Board and co-founder Rackspace Cloud, said that OpenStack already includes a huge amount of collaboration about these topics already, but most of it is informal and not recorded in a form that becomes part of the project.
Rackspace had already seen the need to write down the reference architecture it had in mind for the OpenStack software at www.referencearchitecture.org. Right now, this architecture is intended to explain how to create an OpenStack cloud in a Rackspace data center. In my view this reference architecture must become part of the OpenStack project.
The next thing that should happen is that some level of automated operational processes must be designed and included as well. OpenStack is developing various tools to support automated operation of cloud data centers. The processes are vital to the success of a cloud computing infrastructure and must also be included.
With software, data center architecture, and process architecture all being designed at once, OpenStack will truly become an Open DevOps project.
I should point out that for many open source projects such as Hadoop, a hardware architecture and certain operational processes are implied. In addition, the Open Compute project focuses directly on collaborating about the design of data center architecture.
But both of these projects fall short of I what I think OpenStack is likely to achieve if the project evolves according to the steps I have laid out.
Challenges in the Open DevOps Model
If OpenStack becomes an Open DevOps project, it will have new collaboration problems to work out. While the OpenStack software can be seen as a shared commodity, the design of the data center and the operational processes are going to be a competitive advantage.
Rackspace has built its business by treating technology as a commodity and focusing on customer service as a differentiator. In the world of the cloud, it will be interesting to see if there are limits as to what type of sharing will take place on the design of the data center and related operational processes.
For example, the Cloud Monitoring service announced by Rackspace is not part of the OpenStack project. This could be an example of the type of operational support that will always be unique to a cloud implementation and not shared in an Open DevOps project.
Even if everything about operations is not sharable, It seems to me there is plenty of room to collaborate and share information.
Will CloudStack or Amazon Follow Suit?
To me, the power of the open source model is obvious. Its extension to DevOps is a natural next step.
But for Amazon, such open collaboration would not be a natural next step; it would be a major departure. It is in Amazon’s interests to nurture a complementary ecosystem, not a competing one. (Although as I argued in the article about questions Amazon should answer, I think that Amazon would help its complementary ecosystem by communicating and collaborating more openly.)
Amazon’s closed stance means that if anyone is going to implement an Open DevOps model based on Amazon cloud standards, it will be centered around the CloudStack project, which was recently donated to the Apache Foundation by Citrix (see “Why Citrix Left OpenStack”). CloudStack announced its intention to create an open source software stack that is based on Amazon cloud computing standards.
If it makes sense for companies implementing OpenStack to use the Open DevOps model, the the CloudStack community should do the same thing. I think Citrix should take the lead in turning CloudStack into an Open DevOps project. It will be interesting to see how fast the CloudStack community moves in this direction and if Amazon plays a helpful role.
Is Open DevOps really something new? I think so, and I think that in the next year the natural forces that lead people to work together will make the idea a reality to the benefit of everyone involved.
Dan Woods is CTO and editor of CITO Research, a firm focused on advancing the craft of technology leadership. He consults for many of the companies he writes about and has written content in the past for Rackspace, HP, and Cisco. For more stories about how CIOs and CTOs can grow visit www.CITOResearch.com.
