Complexity and the Triangle of Entanglement
Where will simplification come from?
Working at a startup can be insanely consuming. As I am currently working on evangelizing the use of AI/ML to dramatically reduce alerts and trouble tickets, I sometimes worry I am not keeping up with what is going on in core networking. Then I read an article like this, and my mind is put at ease: “On Applicability of MPLS Segment Routing (SR-MPLS)”. The narrative sounds eerily familiar to the industry narrative a couple of years ago. I’m guessing there is no quick convergence in site, especially for the tens of thousands of routers already deployed. I’ll check in a couple of years and see if things have changed yet ;-)
One of the great things about the space I work in now is that while there are of course consequences stemming from already deployed tools, at the end of the day, if you can get information “routed” to you, you can still add a great deal of value without any of the existing tools having to do anything different.
You can consolidate and normalize many different data types / sources; apply machine learning to generate alerts with less false positives; consolidate 30-100 related alarms under one incident ticket, and more. At the end of the day, it's just software fun with data, and no network equipment is harmed in the filming of the movie.
Long time ago, in a galaxy far, far away, there was a network with a single end-to-end protocol. However, the people were not happy with it. So, they add NAT, dense topologies with millions of interfaces, optical, Ethernet, IP/MPLS, service overlays, SD-WAN and more. Worse still, the network became really, really, successful, because of, or despite, all these changes, so there were so many more endpoints on the network now. Physical endpoints, virtual endpoints, machines, people, things, ... Then one day the people said “Wow, our beautiful network looks a little more complex than it used to be”.
In truth, all of life is more complex than it was a generation ago. Simplicity might be the battle cry in networking today, but it is also the battle cry in life itself. We’ve come a long way very fast across so many dimensions: the evolution of news, always on communication / social media, online retail, most major car makers pumping out EVs, and the New York Giants are 5-1. Oh, the NY Jets are also 4-2. The rate of change is almost impossible to absorb.
So where IS simplification going to come from in networks? As I implied in the three-olive network architecture framework, the answer is not going to be the same for everyone, because different network operators have different capabilities & capacities across services, operations, and the network itself. Each networking team must assess realistically where it is in terms of capabilities and capacities in these three areas and/or where they are going to place large bets / make investments.
You want to simplify a network in a way that pushes the complexity into operations? Well, you better be prepared to either invest in operations or experience service degradation. You want to improve service KPIs? Great, where are you going to invest to do that? You want to keep you installed equipment in place? Ok, what are the implications of doing that? The larger network in any network, is the triangle of entanglement between operations, equipment / endpoints, and services.
Is this a shameless plug for investing in operations tools? Possibly, I do have a conflict of interest there, though I did not when I created the three-olive martini framework - it was just kind of intuitive. However, whether biased or not, it does seem like a reasonable question to ask: what is likely to get results faster, a) investing in operations tools, b) waiting for the SR-MPLS / SRv6 thing to settle out, c) waiting for networks to become simpler by some other means or d) upgrading all the equipment in the network to support a desired network architecture? That question I will leave as an exercise for the reader.