The bittersweet last day of VMworld. This year I'm doing something a little different and attending VMware's Future:NET. Future:NET is a sister conference to VMworld that focuses on networking. The day was broken down into sessions that alternated between panel and single presenters. Here they are with a few of my notes.
Wizards of Networking Predict Shifts by 2025
The day kicked off with a very interesting panel, with five leaders in networking providing their predictions for what networking will look like by 2025-2030. There were a few different ideas or variations on common themes, so the notes may appear to contradict themselves as they're the collection of ideas from several people.
11 year ago networking was primary closed and proprietary. Like the mainframe business of the 1980's.
There had been 7,000 RFCs to consider and implement into network solutions. This was complex and cumbersome.
Since that time we've had SDN, NFV, OpenFlow and other networking innovations.
The era from 2008-2019 saw network owners take control of their software.
The era from 2011-2021 saw and will see network owners take control of packet processing.
- Programmable switches have effectively the same performance as non-programmable switches.
By 2030 expect about 400 Tb/s per switch.
As of 2030 everything will have been programmable for 10 years.
We will no longer think in terms of protocols but software.
Software engineering practices will be brought to bear on network programming.
In 2030 networks will be programmed by many, but operated by few.
Network definitions are starting to move out of documents (ex. Word) and into code (ex. python, go).
There are five phases of networking:
human-driven
workflow-driven
event-driven
machine-driven
self-driving
Open Network is core to innovation.
Networking will be "self-driven, like the cars are being done, because it's safer, not cheaper" (i.e. don't focus on the OpEx).
Quality is most important in the network, followed by simplicity and manageability.
Networking is:
Identity
Connectivity
Security
Scale & Availability
Is orchestration and abstraction the right conversation?
Priorities have changed in the last 10 years; apps are now the primary driver.
In 5 years we'll stop focusing on pipes and instead focus on communications.
Application plan and network plane are treated today like they're different universes (hint: they're not).
Future-proof principles:
disaggregation
commodity & open source
network management (that is safe & comprehensive)
As mentioned, this was an excellent panel discussion and could've taken most (or all) of the day on it's own.
Network or Infrastructure Focused Data Centre
This session focused on whether it's possible to move I&O to an SLA consumption model.
Moving to an SLA-based infrastructure model requires management of risks (don't avoid them).
Provider SLA penalties are typically ineffective, and considered by many providers as "just the cost of doing business".
If SLAs don't work, then what?
We've gone from data centres to centres of data.
- "Where does the data live?"
Expect up to 80% of on-premises private cloud to be deployed by Kubernetes by 2023 (less than 5 year away!).
Infrastructure is just cost of goods sold.
RIP: MTBF and MTTR.
- Don't use these as PKIs.
During the next network refresh look at the network management model and practices before considering vendor and products.
From Automation to Operations
Another panel discussion, this time regarding automation. Most of the panelists were in agreement, so there wasn't a large variety of points to make note of. Here are a few highlights.
Adopting technology is changing peoples' behaviours.
Agile requires version control.
Guiding principle:
- Make the "right thing" the "easy thing".
Simplifying Networks
A small panel discussion regarding the simplification of networks. The core tenet of the conversation was, well, keep it simple.
Simple must equal maintainable.
Making something simple is hard, keeping it simple is hard too.
SOCs or NOCs?
This presentation spoke to whether SOCs, NOCs or both should have control and responsibility for networks.
By 2025 SOCs and NOCs will either:
Be replaced be "Resiliency Centres",
Stay separated (as they are today), or
Integrate with / overlay each other.
"Automation doesn't solve any problems that are not already solved."
Who Should Have Control?
A panel session about who the keepers of the network should be.
FedEx is bringing in PAM, identity, and OKTA.
FedEx uses all major cloud platforms.
Inside FedEx, the InfoSec team owns PAM and OKTA. Specifically an Identity team within InfoSec.
At Google, internal applications don't use DNS at all, they have a completely separate naming system.
At Global Webscale, they don't have any Palo Alto Networks-like systems in production as that type of solution is much too expensive at scale.
Global Webscale turned to Yubikey internally as a temporary security solution while they work on a long-term solution.
Global Webscale underwent a very complex role-based access control (RBAC) initiative.
Google's NOC is part of Site Reliability Engineering (SRE).
What's Next in Branch?
A look at what's happening with edge networking.
Is the running of the network crucial to business success? No, more often than not.
5G will not be the answer for edge networking connectivity for another 5-7 years.
IT Orgs of the Future
Last, but not least, a small panel discussing how the organizational charts of the future will need to look.
"Your DNA will be different is you're building services."
Swisscom is starting to see business developers, where developers are located in business units outside of IT.
Try out changes to organization structure via a lighthouse project before rolling them out corporate-wide.
The End… for Now
That wraps up another VMworld 2019 conference. As always it was tiring, exhausting, overwhelming, and worth every single minute of it. I heard someone mention that just because you get better at doing something, that doesn't mean it gets easier. The same is true of VMworld. Attend a few times and you figure out efficient routines, how to set your schedule, learn to prioritize what's important to you, etcetera. That makes each moment as optimum as you can, however there are still a lot of moments, and a lot of legwork (literally; I walked over 125k steps from days 0 thru 5).
So, I'm tired (as all attendees must be) but happy and grateful for yet another successful conference. And I can't wait to do it all over again. See you next year San Francisco!