In this conversation, sFlow creator and InMon president Peter Phaal catches up with Art Fewell on Open Networking TV to discuss new networking technologies including SDN, network programmability and network virtualization, and the new innovations in monitoring needed to tame the wilds and put these new technologies to work for you.

For the full experience, watch the video below, or read the transcript. For our TL;DR readers, here’s a couple of quick snippets of the conversation, especially as it pertains to networking:

• Networking is in the midst of a renaissance of innovation with numerous new architectures including SDN, programmable networks and network virtualization. New innovations in monitoring and management are needed to make the most of these new technologies. sFlow has risen to the occasion with several options to help understand and optimize new networking solutions.
• “Software is just a lot more flexible. There's always fixed capabilities in your hardware platform, whereas software life cycles are just much faster and innovation happens much more rapidly. I think the pace of innovation that you can expect to see at the [virtual] edge is just going to rapidly exceed what would be possible in hardware.”
• “I think pretty soon CLIs will be dead. You're just basically getting into the install firmware onto the switches that tightly couple with the orchestration tools that you're using to manage all the rest of your IT infrastructure, probably under something like OpenStack.”
• sFlow and related monitoring technologies now have standard, built-in support in most newer networking chipsets.
• “We had this rather quixotic goal of embedding our measurement technology into the hardware. That's really what took 15 years to get the instrumentation ubiquitously implemented in networking hardware. Today that's really happened. Pretty much every switch you buy has the measurements built in.”
• Open Networking platforms are creating never-before-possible opportunities for innovation in networking, creating fertile grounds for rapid innovation by independent software vendors
• “Picking up on the move to bare metal switches, as a software vendor, is really exciting for us because right now people think bare metal, and they're thinking they're going to save money. Our idea that the reason why you're going to move to bare metal is not because of the low cost, although that's a good plus, it's really that software vendors like us and many others can really deliver value on top of those software platforms. We're working with Cumulus Linux and it gives you a tremendous environment in which rather than being hamstrung by whatever CLI commands or APIs a vendor chooses to expose, you can just go right in there, compile your own software, install your own applications and make that hardware do what you want and deliver tremendous value on top of these open platforms. In the bundled model, if your switch vendor isn't innovating in the areas where you need them to, you're sort of stuck with something like Cumulus Linux. If it's competitive advantage to you, you can write your own applications in-house.”
• “Having the same measurement technology in the physical and the soft switches helps a lot to collapse the silos and give you that end to end view that you need. With sFlow you can enable it on all the ports, physical and virtual, and get that end to end view of how encapsulated traffic is flowing over the physical infrastructure.”
• A key wave of innovation that is now becoming possible is the use of sFlow to provide real-time responses in infrastructure for changes in traffic conditions and application requirements.
• “I think continuous feedback and monitoring of everything is just a radical transformation in the way you orchestrate infrastructure so you don't need to guess anymore. You just start directly measuring everything and that to me is the real promise of cloud infrastructure. In spite of the fact that people have virtualized all their workloads, CPUs still run with very low utilizations. There's this tremendous cost savings, or increased revenue to service providers that are available if you can right size resources. That really requires an accurate measure of actual demand. Then you can allocate the right bandwidth to that service. You can allocate the right amount of storage, the right amount of CPU, the right amount of memory. Then you can very rapidly flex those resources between different applications as things change by time of day or as new services are set up or spun down. I think that's going to be the really interesting thing. Having the whole data center really adapt dynamically to changing workloads rather than statically provisioned, very over-provisioned resources.”