Last week I was lucky enough to attend the 2018 North American Open Networking Summit put on by the Linux Foundation.
Overall, the ONS summit was great. I learned a lot, had some good conversations, and was reminded of how much cool stuff is going on in the realm of networking.
There were a few themes I identified:
The amount of NFV related work being done is amazing. I was a bit overwhelmed the first day as I reminded of all the work that is happening, most, if not all, in the NFV area. It’s not a stretch to say that Linux networking = NFV, IMHO.
Optical networking was also often mentioned as a place for automation, which surprised me.
ONAP - The Open Networking Automation Platform is nearing the top of the hype cycle. I believe ONAP is extremely important. However, it’s quite complicated and has a long way to go in terms of stability. Some telecoms are using pieces of it in production right now. It’s a large project and not everyone will use every single sub project.
Canada has a many companies working in networking and telecom, especially in Ottawa and Montreal. Ottawa has Wind River, Ciena, and Amdocs to name a few.
Speaking of complexity - I complained several times to people about the level of complexity we are entering. People used to think OpenStack is complex. Well now you have to add SDN, ONAP and containers onto that. OpenStack is probably the easy part now (relatively speaking). Sure, Kubernetes is easy to deploy, but how many people who deploy it truly understand how it works.
Dependencies - Still on the complexity front, I see tons of issues with dependencies and managing them. All of these systems, OpenStack, SDN, ONAP, etc, will have to not only be deployed, but talk to one another. We are heading into some serious dependency issues. I took some ONAP training on the last day, and we had several dependency issues just in terms of deployment, and the trainers had deployed this particular set of ONAP modules hundreds of times.
Central Office - There are so many COs spread across NA…tens of thousands of them. This is a real key area, perhaps the most key, for NFV in the next few years. ONF is doing a ton of work in this area around the idea of “Central Office Rearchitected as Data Center” or CORD for short.
A few other small points:
OpenContrail has been renamed “Tungsten Fabric.” Bit of an awkward name, but I don’t mind it. Especially if you drop it down to just TF. Juniper really needs the community around TF to grow, as it’s really a single vendor project which is no good to anyone in terms of open source.
dNOS has been renamed DANOS. There are a few Network Operating Systems (NOS): DANOS, Sonic (from Microsoft), and the recently announced Stratum (thought I am not completely sure Stratum would consider itself a NOS). AT&T plans to deploy 60k whitebox switches with DANOS. I am very excited to see more work being done in the NOS area. All we really need are some good drivers for switch hardware and some more work around common networking protocols (as well as the new advanced technologies) and we should be able to have a thriving Linux NOS.
The P4 programming language was mentioned quite a bit. Will be very interested to see where that goes.
AT&T sees 200PB of mobile traffic per day, up 50% from a year ago.
Wind River is open sourcing some (all?) of their platform.
OpenStack is still trying to figure out what it is going to look like on the edge. Will it be remote hypervisor, multi-region, cell based, or light weight. No one knows. Probably all of them I guess.
Network slicing was often mentioned, but no one seemed to know what it was actually going to look like in terms of implementation. I heard it in several presentations, but no real details.
Fd.io/vpp was mentioned quite often. I need to better understand that technology.
Again, overall, the amount of work being done in the networking area is astounding and it will be hard for people to keep up with the pace. It’s a great area to be working in. I’m going to try to attend the European ONS summit upcoming this year.