Interop 2017

This year’s annual Interop ITX conference was the second year I attended this event in Las Vegas. And while I am preparing a presentation for an upcoming debriefing to my colleagues I am trying to summarize the content and try to come to some sort of conclusion. What do I take back to my colleagues and clients? What are the relevant trends I see we need to take into account? Many thanks to the PacketPushers team for the two day Future of Networking summit. A lot of valuable insights were gained during these two days.

Takeaways

  • Automation and programmability are becoming an important skillset for the networking engineer in the years to come. I am reluctant to say they are “essential”. It’s still too early how the market and products will evolve. Will everything really be abstracted behind a shiny intent or intuitive based GUI or are hardcore coding skills necessary to glue it all together? I think we still have a long way to go. Especially in Enterprise IT where networks are far from standardized. Let’s put it differently: If scripting skills are needed in Enterprise IT infrastructure to program anything, they will be used to script deployments to the public cloud. Where networking is abstracted away to the point it just works and your typical developer can’t be bothered.
  • Closely related to automation and standardization are the actual standards. Not only standards like YANG or RESTCONF, but also standardization in design and of interfaces.
  • SDN is becoming a race between an programmable underlay (like Cisco ACI) or overlay (like VMWare NSX). And apart from the fact the choice between the two is never definitive or a winner takes all situation as it very much depends on your own IT infrastructure, I do think that in the end the underlay becomes less relevant (hence the advent of white box switches becoming cheaper). In the future one will have a simple programmable underlay (but with a simple design programmed) and a complex overlay. Just as the cloudscale companies are presumably doing it now. I think it will take a few lifecycles (of three to five years) for enterprise IT to reach this level. But then again, in enterprise IT a lot of factors come into play which are mostly non-technical. I will create a blog post “Control of your network in an outsourced environment” somewhere in the near future around this subject.
  • SD-WAN is interesting and can solve a lot of headaches large hub-spoke networks introduce with the networking community. Although cost is far less a driver in Western Europe and Holland specifically, ease of management and application awareness is an interesting factor. Currently, I don’t know of any large scale deployments here in Holland. Correct me if I’m wrong in the comments.
  • Network security. Short story: Firewalls are useless 😉 Long story: Firewalls are somewhat useful, but in many scenarios the risk you want to mitigate does not involve a firewall as we know it. Protection from lateral attacks are becoming more and more important and so micro-segmentation and host based solutions are becoming increasingly important. But how do you manage this in a scalable way. RFC1925 rule 6 can rear its ugly head.

Odds and ends

  • Cloud; interesting to hear some reallife use cases. Different drivers to move to cloud and it’s not always cost to start with (although companies might move back due to rising cost, but that’s another discussion). Interesting to hear that moving to public cloud to really leverage the innovative capabilities is a main driver. At the same time this is the reason why shadow IT and public cloud are so intertwined sometimes.
  • Orchestration of it all.. This is for me still uncharted territory. I have looked into Cisco’s Cloudcenter (used to be Cliqr) and that certainly has some potential. I see also ServiceNow as an interesting product used and off course Openstack’s Heat. But as said, I am too far removed from this specific subject to know the dirty little details. Definitely something to know more about!

Interop ITX itself is positioned as a vendor independent conference. And although the sponsors (VMWare among others) are far from independent, they were able to attract a wide variety of speakers. In that lies an important value. But if I have to believe my Twitterfeed and compare the Cisco Live US stream with that of Interop I wonder if it is able to live up to the vendor conferences. I don’t think they have to have that goal, but getting behind the curve can create a negative spiral in less attendees, which means a smaller event and maybe less speakers. Let’s hope it doesn’t go that way. We need vendor independence, it is the raison-de-etre of the company I work for and I think my customers need this objective view to make the right business and IT decisions.