As an integral part of our ongoing commitment to composable networking, Metaswitch has joined the board of OpenSwitch, a networking operating system (NOS) for whitebox switches that the Linux Foundation runs as an open source project.

So, why are we getting involved? Well, first and foremost, OpenSwitch aligns firmly with our vision for composable networking, which we believe will yield much-needed innovation, feature velocity and cost savings to the network. OpenSwitch facilitates the same business and cost model for whitebox hardware and disaggregated software that has been so successful in the compute environments: “standard” hardware provided and ported by specialists, an open source OS (Linux) that abstracts those hardware resources, and an ecosystem of applications on top.   OpenSwitch does the same for whitebox switches: it represents the open source OS that is required to demarcate and abstract hardware from applications, encouraging the same type of vibrant ecosystem of open source and vendor applications. Users can then pick and choose just the applications they need. This sets it apart from other forms of disaggregation that some vendors are pushing today, where the OS is vendor-specific and comes tightly coupled with a constrained set of networking protocols and applications. Also, and very importantly, it isn’t just a hobby project; Dell has backed the ecosystem by offering to provide sales and support for the OpenSwitch distribution as well as the solution, giving operators the single throat to choke.  

But, as an engineer, why am I excited? Well, there’s a lot to like. Firstly, it incorporates a standard Linux distro -- Debian, which means that, for application ecosystem players, there is a known base to work with, rather than some proprietary environment. And there is the VM support. We use this extensively in our testing, where our test framework can create any particular topology out of a mixture of physical switches and VMs -- which we integrate into our own CI/CD environment. That gives us a lot of confidence that what we are delivering will work in each use case. The VM also provides a handy format to allow the community to evaluate our application easily too.

Looking forward (and perhaps most importantly), it seems to me that the OpenSwitch architecture gives us the potential to deliver on some really interesting use cases that go far beyond the enterprise/DC use cases people typically associate with disaggregated systems.   For example, although it supports the Linux Networking API as a means of programming the ASIC, OpenSwitch also has a flexible YANG-based API (the control plane services API). This gives us the opportunity to innovate and develop new features. If we want to extend the NOS to support features that you can’t straightforwardly get working with the standard Linux networking API (like BFD hardware assist, loop free alternates or BGP Prefix Independent Convergence), we can. As a result, we can use OpenSwitch as a NOS that enables a carrier-grade, Internet-scale composable networking-based solution based on open networking hardware.   And with switches like the Dell S4200 (which, based on the Qumran chip, has the option to scale tables and flows to very large numbers) we have the opportunity to turn that “cool idea” into a real walking, talking -- or at least sending and receiving packets -- network device, not just a PowerPoint slide.   

So, that’s why we are getting involved in OpenSwitch. If you’ll forgive the immodesty, I’d like to think this can be mutually beneficial: over the past 35 years in the networking space, working with customers to get our software deployed in a wide variety of use cases, we’ve certainly learned a lot about what is need to make a platform robust and deployable.  We’re looking forward to bringing that experience to the OpenSwitch community.

Click here to learn more about our portfolio of Composable Network Protocols.