Open source is playing an increasingly important role in IT infrastructure generally. Certainly, the role of open source in the compute space is well understood, and networking has been making its own migration towards open source with the OpenDaylight movement. But is open source a natural evolutionary path for all IT disciplines, or are there certain characteristics that make some areas more ripe for open source than others?
When we think about networking as an industry, we tend to compare its progress to the evolutionary track taken by the compute world. The base assumption here is that the industry will unfold in much the same way that the compute industry did, marching past some set of ubiquitous mile markers. But this view of the world sort of assumes that evolution is a two-dimensional track, and that industries are either parked somewhere along the continuum or they are moving forward towards a predetermined end.
But what if evolution doesn’t follow some set schedule or even a singular path? If we assume that technological evolution is not predetermined, then what conditions drive an industry towards open source?
There are at least three major drivers for broad open source adoption:
- Single platform – When there are lots of applications that run on a single platform, that platform is particularly well-suited for open source. For most platform plays, the value and differentiation is not in the platform but rather in what runs on top of the platform. It makes sense that, to the extent possible, the vendors developing on top of the platform should leverage a common body of work. Re-creating foundational elements that are largely not differentiating is duplicative work that doesn’t ultimately help the end user. Additionally, a common platform helps to ensure that all the applications on top of the platform can run in what ends up looking like a fairly ubiquitous execution environment. This is largely what drove the migration of compute towards Linux. That a platform is open source and ubiquitous does not mean that it cannot also be lucrative. Companies like Red Hat have been successful at leveraging the broad install base to generate a solid revenue stream. That the platform they support is common helps to ensure that their customer base is as large as possible. Even small deviations in the underlying platform would fracture their customer base into a smaller set.
- Single point of control – Not unlike the platform driver, when there is a single point of control for a large number of infrastructure elements, that point of control lends itself particularly well to open source. The value in a point of control lies either in managing very specific workflows (as with most single-vendor management platforms), or in broadly orchestrating workflows across disparate elements in a heterogeneous environments (as with SDN controllers). The former tends toward tightly integrated management/execution solutions; the latter provides a fertile breeding ground for open source. By adopting an open source framework for points of control, the community helps ensure that individual players do not end up with monopolistic control that can then be used to unduly influence decisions further down the technology stack. In essence, open source creates a very natural counterbalance to what would normally be competitive efforts to create “sticky” solutions.
- Nascent technology – Innovation is always important, but in a technology’s formative stages, that innovation is necessarily less focused in a particular direction. When the outcome is uncertain, the number of potential paths approaches infinity. During these times, the best thing for the technology is unbridled support. Open source allows the widest aperture for new ideas to come into the space, which makes it ideally suited for nascent technology spaces where iterative experimentation is necessary. Open source does not preclude companies from creating protected innovations. Certainly, open source projects can be extended in commercial and even proprietary ways. But open source does ensure that access to the most important base concepts and foundational elements is uniform and open.
With these drivers in mind, it is relatively straightforward to see why open source plays a large role in certain areas of IT. On the server side, the proliferation of applications and the desire for those applications to be portable was enough to ensure the emergence of an open source compute platform like Linux. Once performance was good enough, differentiation was always going to move to the applications, which made unique platform capabilities unnecessary for the lion’s share of apps. Where performance or specialty capabilities remain important, there is still a small market for special operating environments.
As we look to networking, open source seems like a foregone conclusion as well. The push towards SDN makes the controller space particularly well-suited to open source. The desire to have a common control platform capable of near-ubiquitous deployment and with control hooks into a large number of heterogeneous elements is likely enough to guarantee a significant role for open source in networking. This is a large part of why projects like OpenDaylight hold such promise. The viability of proprietary, standalone control platforms in the face of a push towards orchestration and automation is questionable at best, except in the case of very specific (read: niche) workflows.
Storage would seem to be the next logical vertical to be impacted by open source. With a trend towards federated storage clusters, there will be a need for a central control mechanism, not unlike SDN. A single point of control spanning heterogeneous architectures is ideally suited for open source efforts.
The point here is not that open source is a necessary evolutionary step but rather that open source becomes a key ingredient as conditions in a technology space favor the value that open source brings. Where there is commonality in platform or control, open source will thrive. Where specialization is critical, open source is less relevant. As companies look at their own open source participation, this suggests that they ought to be examining natural points of convergence. The conditions more than the projects are likely to determine the success of open source across the various pockets within IT.[Today’s fun fact: Actor Tommy Lee Jones and Vice President Al Gore were freshman roommates at Harvard. Imagine the parties with one roommate a fun-loving guy and the other one named Al.]