Two more open-source networking projects have been accepted into the Linux Foundation. Those projects are OpenContrail, originally developed by Juniper Networks; and dNOS, an AT&T initiative.
OpenContrail, which has been renamed Tungsten Fabric, is an SDN platform aimed at organizations building hybrid clouds or multi-cloud infrastructure. It’s built around a virtual router that provides connectivity for containers, VMs, or bare metal servers. It also includes a controller to orchestrate networking services and serve as a single point to set and manage policies.
OpenContrail is the open version of Juniper’s proprietary Contrail software. Juniper acquired Contrail back in 2012, but as Randy Bias, VP of Technology and Strategy, Cloud Software at Juniper, notes in a recent Network Field Day presentation, it was open source “in license only.”
In other words, the software was freely available, but Juniper kept a tight rein on the community and governance. Customers and partners complained, prompting Juniper to revise its approach.
By moving the project to the Linux Foundation, the company hopes to revamp community interest, attract more developers, and gain more users.
From dNOS To DANOS
AT&T’s dNOS is also being moved into the foundation. The telco announced its dNOS project in 2017 with the goal of a creating an industry-standard, disaggregated network operating system. The dNOS software includes code from the Vyatta router, which AT&T acquired from Brocade.
AT&T wants an OS that can run on multiple hardware types, including CPUs and ASICs. It also wants a set of open APIs to separate data plane and control plane functions, and to be able to integrate with existing and future tools.
This isn’t just a theoretical exercise; AT&T recently announced plans to use dNOS/DANOS on 60,000 whitebox routers the company will deploy as part of its upgrade to 5G.
The traditional network stack has been split open and atomized. Open source projects target everything from hardware design to programmable ASICs to the device OS to controllers and orchestration platforms.
It’s a time of amazing, and often confusing, development. Some of the efforts being undertaken seem duplicative, but some also have powerful patrons such as Facebook, Google, Microsoft, and AT&T. These companies have the in-house developers and market clout to ensure their bespoke projects will attract ecosystem partners and won’t languish in a corner somewhere.
But will these projects matter to the non-hyperscalers and regular old enterprises?
I don’t know that I can point to a specific project and say “That’s the one.” But I do think the cumulative effect of these efforts are breaking the stranglehold of legacy vendors on how networks are built and operated.
We don’t need to do things the way they’ve always been done anymore. That’s exciting, and maybe a little frightening. But it’s an interesting time to be around.