You’ll be forgiven if you’re not familiar with IP Infusion. I know I wasn’t familiar with the company before I saw their presentation at Networking Field Day last year. But while their name might not ring any bells, you probably know something of their work. This isn’t surprising given that the company has been around since 1999. So who exactly are they?
The company was founded by Kunihiro Ishiguro and Yoshinari Yoshikawa, who developed GNU Zebra in the late 1990s. When this open source routing platform was decommisioned in 2005, many switched to the Quagga fork of GNU Zebra as the logical successor. But IP Infusion was founded to commercialize GNU Zebra in the form of ZebOS. While Quagga found adoption with Cumulus Networks, Vyatta, and other organization, IP Infusion has stayed under the radar. That’s because they’ve mostly targeted OEMs to date. While they may lack name recognition in the IT space, they have found adoption for ZebOS among companies that do: Citrix, F5, Huawei, and Plexxi to name a few.
Outside of the ZebOS source code that’s been their bread and butter for a while, the company is shifting strategies to continue their relevance in the modern networking world. The company provides both a binary network OS for white box vendors,and a virtual network OS, OcNOS and VirNOS, respectively. These were developed as a direct response to customer requests for greater network disaggregation.
I really think OcNOS is the more interesting of the two, simply because of the emerging market around white box switching. IP Infusion is able to provide a fully baked, carrier-grade network OS directly to this market. To do this, IP Infusion built their own Data, Control and Management Planes. They abstract the OS and hardware via a platform abstraction layer, transforming a physical entity into a logical one. At the end of the day, OcNOS has no visibility into the physical hardware. All this means it’s compatible with a range of pizza box and multi-slot chassis hardware.
Without getting too much into the weeds, the key words to remember for OcNOS are modular and scalable. On the control plane, the modularity comes in with a fully programmable pipeline from the Linux TCP/IP stack on up. If the underlying hardware doesn’t have a robust SDK, IP Infusion will write its own forwarding plane to make sure you’re able to address the needed functionality. I would think that if you’re invested in white box switching this wouldn’t be too much of an issue, but it’s nice to see IP Infusion realize that existing hardware might not be ready right out of the gate. There may be some minor inconsistencies if underlying hardware has locked down proprietary features, but from a management perspective this shouldn’t cause issues.
I have no doubt that given their legacy of building network operating systems, both OcNOS and VirNOS are full featured and robust solutions for a given organizations needs. In his look at the company, Brandon Carroll does raise the interesting question of support. How well does a company that’s previously only had OEM relationships, and didn’t have to worry about direct customer support, react when issues inevitably crop up? That’s the one point where IP Infusion can’t lean on their long legacy in the networking space. I don’t have any reason to doubt they’re prepared for this, rather to point out this is one areas that’s not an established strength.
Still, white box networking is one of the most exciting areas in enterprise IT today. While IP Infusion might not be the most prominent player in the space, they have the technical and intellectual property experience to be extremely relevant players in it.