Thread by @arielsanchezmor

The event has started and we're having some fun polling the room on networking #futurenet

Lots of great insight being gathered. I'll check if it can be made public

Tom is now introducing Rajiv Ramaswami

Networking have been more exciting these last 10 years! Scale, software and merchant silicon have changed the game #FutureNet. Today we will see what will come in the next 5 years!

We start with a panel of networking wizards - people who have created this revolution #FutureNet

Nick McKeown is walking us through the future of networking by revising how the networking process has changed.
1 network owners took control of their software
2 " " took control of packer processing
And we're checking the state of programmable chips (using #P4) #futurenet

Apart from capacity increases - packets as important as memory in regard to the CPU processes (death of PCIe?) #FutureNet

As the network behaves much more like software - protocols and standards will be less prevalent, and more custom-tailored networks in software will become the norm (this is already happening today, especially in QoS)

One thing that resonated - lots of people will program networks, but few will operate them - because they will be much easier. #futurenet

We're now listening to Bikash Koley #FutureNet

Differentiation between Software Defined Networking (stating the desired state of the network) and Intent Based Networking (closed loop reinforcing concept that optimized the network continuously) #futurenet

Almost broke my thread, sorry :)

https://twitter.com/arielsanchezmor/status/1167112574690660353?s=19

It's interesting the AI/ML in software is not only popular, but *already* in products in 2019. We didn't talk about it much in 2018 #FutureNet

Very nice concept - let's stop trying to bring the legacy infra to the new world #FutureNet

A statistic just mentioned - 70% of network outages were unintended human error. So, the self driving network is also about making networks more stable, much like self-driving cars #FutureNet

Nick McKeown raised an excellent point - if you don't know what a packet is doing, then don't apply AI/ML, deploy the existing tools so you *know* why. That is more secure than applying techniques.

We'll be now listening to Vijoy Pandey from Cisco

after building complex solutions, and then seeing how this complexity was being globalized - the intent moved to simplifying #FutureNet

"virtualizing" the network for s while was just putting stuff into a VM. But the applications should not have to rely on switches, routers etc. Everything should be re-architected, starting with the application programming/development #FutureNet

(I'm thinking that if my application defined its own global networking capabilities, then the network would have to re-arrange itself to carry out that idea, and there would be a trust process. problems with this - who owns equipment, how it's charged, admin rights, etc)

Super heated debate - let's not talk about open networking when I can't bring my own code, or your competitor's to your box. It started because we compared that when we write software, we don't care if it will run on a dell/hp/smc server #FutureNet

Now listening to Ken Duda - he pointed out that more important than anything else is increasing quality and manageability. Things are easier to build than operate. I love it #FutureNet

Mutli-vendor effort to provide a single plane of control for the network #FutureNet

Very interesting concept - allowing the customer to see the whole network across vendor/cloud boundaries #FutureNet

Super interesting debate - is the right approach from top to down (program the network behavior) versus from the bottom up (I have X and Y and I need to manage it)? It seems both can exist and may be different phases #FutureNet

Now Pere Monclus - opening with a great point. You might be sending your people to cloud conferences, but not computer architecture conferences.

Pere argues that businesses have changed and the network has to adapt to focus on serving the new communication models #FutureNet

Now we'll listen to Albert Greenberg - and he doesn't want to talk about the network, but us, the operators and beneficiaries of it

Al is making some great points - it's harder to teach people to be open minded, creative and welcoming to change. And diversity is important to reach the best solution #FutureNet

Interesting slide - didn't realize Azure had optical design and field engineers

The culture is so important - embrace adaptiveness to change, diverse thinking, skill sets and talent. #FutureNet

Asked by a minority attendee (like myself) how did you achieve a diverse team in your org? One of the answers is "having coffee with people". THIS. Seriously, nothing has helped my career more than having a person senior to me dedicate some time to help guide me. #FutureNet

Love a question from the audience - how will you pay down technical debt? Because we won't be able to do the future of networking if we don't bring down technical debt. #FutureNet

This has been very good. lots of questions from the audience and back and forth. 15 minute break #FutureNet

We're restarting #FutureNet, gotta herd the cats

Tom Gillis is also talking about the "simplify" part of the network industry trends. We don't get rewarded for uptime, but definitely get penalized for downtime #FutureNet

Joe Skroupa will now talk about availability as a whole entity - talking about just signing a contract with a SLA and you don't worry about anything else. #FutureNet

The fact that missing SLA has little relative penalty means that accountability is not very high. in the early cloud days, a CIO asked "when my applications go down, I'll need to hear who got fired" . Harsh but important to really drive home accountability.

Some good concepts in these slides - one very true is that most large companies are not planning on decommissioning their on-prem datacenters #FutureNet

"The only workloads that have come back from the cloud were the ones that shouldn't have been on the cloud in the first place" #FutureNet

Secure Access Service Edge - a Gartner model #FutureNet

There is a fundamental disconnect between the funding of Digital Business (growing fast) and the trends of IT FTE's (decreasing slightly)

"In the cloud if you can, on premises if you must" #FutureNet

"we want you to be agile" but they pay for availability and penalize for outages. Adjust your KPIs for what you are encouraging and align the pay. Loving these slides because they are very true! #FutureNet

Love the "Hold a Network Hackathon" and No-CLI fridays! Simply, and this is my opinion, managing networking should not be different than managing the rest of the infrastructure, thus take advantage of what the compute guys have been doing for years #FutureNet

The summary of recommendations from Joe Skroupa #FutureNet

Now a panel led by Cameron Haight, VMware Field CTO for Americas #FutureNet

Hearing Andrew Clay Shafer talking about how you can't do next generation infrastructure using traditional methods #FutureNet

"No one really knows the whole thing - when people say they are full stack, they write javascript on a webpage and on the server" *laughter*

Silos (even whole industries and consortiums) will try to solve problems within their own expertise domains. This was a fun joke until he said it's really being done with service mesh etc #Futurenet

Getting into more panel discussion now. Most people here agree on the agile methods and the future role of networking. It was a great idea to include Andrew Clay Shafer since he comes from the dev side and has tried working with legacy engineering in the past #FutureNet

Important points everyone agreed in regards to scale and operations - start small, build expertise, focus on consistency in operations. Once you can do it for 10 or 50 you can scale 10X and you will be able to handle the subsequent challenges. #Futurenet

Some other thoughts:
there's also a certain maturity to the CI/CD tools by now. lots of people that see the benefit even in small deployments

sometimes when people say "brownfield" they're really saying "I really don't know what's on my network" xD

#futurenet

disasters tend to happen when people try to automate what they can't even do manually. make sure you understand what you are trying to automate #futurenet

We're back from lunch and Tom Gillis is kicking off the afternoon panel but first some polls for the audience #Futurenet

By the way seeing the PollEV results as attendees answer has been a lot of fun #FutureNet

Now a panel with about how networks should be simple with Omar Baldonado, Dino Farinacci and Arpit Joshipurs #FutureNet

Arpit from the Linux Foundation tells us about everything the LF does

Details of where open networking is - some vendors don't want to lose their secret sauce/differentiator #FutureNet

Open source has been critical to get the most basic functions right - the vendors want to innovate at the higher levels of the stack. Seeing that there's a lot of resources available on the Linux Foundation website for open source networking. #FutureNet

A fun prediction for 2025 - we will still be talking about next gen networking xD #FutureNet

Omar Baldonado highlights that by having simple networks it's easy to onboard new people. If they take 2 years to learn it, they'll likely have moved on from that job by then #FutureNet

Remember - don't sacrifice manageability im the name of availability. It's a balance. As simple as you can that will meet the requirement but still remain easy to operate, troubleshoot, upgrade. #FutureNet

Next up Dino Farinacci - bugs and dreams #FutureNet

Dino is telling us about how blockchain is suffering the same vendor discrepancies that the networking market faces 40 years ago. He raises a lot of the contradictions in our network. #Futurenet

Bruce Davie discussing with Dino - old colleagues and networking veterans #Futurenet

John Pescatore on operations (SOC/NOC) #FutureNet

"It's becoming harder and harder to recruit network engineers." I agree. When I started my career I really wanted to be a rockstar Cisco neteng. Now, I wouldn't want to limit my options to just learning one vendor and in one role. Tough since there's a high barrier of entry

John makes a lot of sense and talks about why he left the analyst life. But I love that he is telling us that sometimes the difference in a big breach is operations - the level of ops are the best indicator of a company's IT security #FutureNet

SOCs and NOCs #FutureNet

Virtualization gives opportunity for real security #FutureNet

Restarting after a small break. John Pescatore hosts Forrest Bennett, Grant Dasher and Lane Patterson on Who should be the gate keeper of identity, network and security policy
#FutureNet

Here's their companies

Grant Dasher explains about Google is how everything is integrated through process. No mention of silos. And yet, there are controls throughout the process #FutureNet

How is security measured? There's simply no clear answer. Sure, basics like creating risk profiles and making sure to log, analyze and follow up, but you will create your own version according to business priorities and environment
#FutureNet

There's some KPIs like response rate, restore rate, reduction of attach surface. #FutureNet

Mike Frane - SD wan and branch networking #FutureNet

Is SD WAN living up to the hype? #FutureNet

And the last session of the day - IT org charts of the future #FutureNet with Albert Greenman and Stephan Massalt

This last panel is super interesting. Lots of questions. I honestly think you could do 3 hours of questions on org charts, processes, management, budgeting, team dynamics, documentation, compensation, etc etc etc #FutureNet

@threadreaderapp unroll please :)