Since getting deeply involved in OpenDaylight, I find myself talking to many different types of individuals—from vendors and customers to developers—about what OpenDaylight is doing and why it matters. Perhaps most importantly, I talk to users about what OpenDaylight can do for them. For those who aren't familiar with it, OpenDaylight is a multi-company open source software project building a platform to further adoption and innovation in Software-Defined Networking (SDN).
In preparation for Brocade's Federal Forum I have distilled what I am seeing change in networking and why it is that I am excited to be working on it here—both in OpenDaylight and at Brocade. We are witnessing the rise of open source software networking as the way to deliver innovation and features that have eluded us in networking for so long.
The underlying change can be charted along two axes. The first axis addresses how networks are developed—from exclusively buying from vendors to building your own. The second looks at how organizations approach management ranging from device-by-device management to network-wide management.
Opening up the spectrum between buy-from-vendors and build-your-own
Traditionally, there was only one option to acquire networking—whether delivered internally or from a vendor through hardware or software—and that was directly from a networking vendor. Recently, Google, Facebook and others have shown how organizations can build their own networking—including hardware, firmware and software—from scratch. This approach offers different trade-offs than buying from vendors, but for most organizations it is not cost effective.
As open source comes to networking, the set of points along the network development spectrum is expanding, ranging from taking the projects that Google, Facebook and others use to building networks and investing in the local time, resources and talent to deploy the network to buying vendor products based on open source software like OpenDaylight.
As we move from buying to building, an organization's ability to innovate and avoid vendor lock-in increases as they gain more leverage over what the can acquire and deploy. At the same time, the internal cost to an organization tends to rise as an increasing amount of the integration, testing, and development moves from vendors to customers. The good news is that the cost curve is shallow at first giving some points along the curve that are likely to be attractive to nearly all organizations, including buying products that are based on open source from vendors, which reduces vendor lock-in and increases interoperability.
Moving from device-by-device to network-wide
Currently, most organizations run their networks by configuring them using the CLI in a device-by-device fashion. This is largely because the management and control plane interfaces are tightly coupled to and run directly on each device. Device-by-device management is harmful in three ways—it hinders innovation and cost reduction, it makes devices and the siloes they create susceptible to vendor lock-in and it requires organizations wait for monolithic complex upgrades before they can innovate.
The good news is that the tools to break this apart were developed during the beginnings of SDN, when SDN was only about OpenFlow and the separation of the control plane from the data plane. Now, we have standard, open protocols, APIs and interfaces to devices that allow for reliable, cross-vendor orchestration through efforts including OpenFlow, NETCONF and PCEP.
To varying degrees, these allow us to disaggregate the layers of the stack providing control plane and management plane features in software that runs on commodity servers in a logically centralized manner. This allows for both independent innovation as well as allowing for network-wide visibility and policy enforcement.
Putting it All Together
We are witnessing a significant shift in networking to open source, software-based, network-wide networking. The rise of open source software networking including not only OpenDaylight, but also more other open source projects than I can name, are driving this change.
These projects have turned SDN and the future of networking from a pipe dream into something that can be deployed today. I strongly encourage you to figure out what gear you have that already supports the key protocols like OpenFlow, NETCONF and OVSDB as well as adding support requirements to your future RFPs. Further, download the code and start playing with it. You can get OpenDaylight here and a free trial of the Brocade's commercially supported OpenDaylight distribution here.