I planned to take my summer break seriously and stop blogging until late August, but then I shouldn’t have looked at my Twitter feed (my bad), where the AI algorithms selected just the right morsel to trigger the maximum rantiness. I would strongly recommend you read the original tweet and all the responses first – it looks like it was a serious suggestion, not a trolling exercise (here’s a copy of the original idea in case the tweets get lost in the mists of time).

Done? OK, here’s my take on what the author of that inflammatory tweet was saying: “the networks I design need no redundancy, and thus I consider redundancy to be unnecessary.” Good for him, and we could have stopped right there, but I got the feeling it went further along the lines of “and I’m willing to die on this hill and to prove how misguided y’all are.” Ouch.

Now step back and think for a moment about the various requirements we’re facing when designing networks. It ranges from “my Grandma needs Internet” (kid’s tricycle) to hyperscaler data centers (aircraft carriers?) and air traffic control (rockets?). Claiming one could apply the lessons learned from one’s small corner of the world to all of them is potentially a bit arrogant.

However, there’s always a nugget (or two) in every public spat; one just has to have the nerve to look for it ;) In this particular case:

• State-sharing redundant solutions are way less reliable than non-redundant solutions. I’m looking at you, firewall clusters and multi-supervisor switches… and I can’t force myself to look at stackable switches.
• Trying to increase system resiliency by deploying complex redundant infrastructure benefits only sloppy application developers and infrastructure equipment manufacturers. You might end up with a brittle solution that will break faster than the non-redundant infrastructure components or the craplication running on top of it.
• In many cases, we build redundant solutions to cover our backsides. Nobody could blame me for a failure if I did everything in my power to prevent it, right?

Does that mean the author of that tweet was correct? Of course not – the only way to get a resilient system (which includes infrastructure and everything running on top of it) meeting whatever requirements you might have is to figure out the real requirements first, and then follow the excellent suggestion made by Albert Einstein: “Everything should be as simple as it can be, but not simpler.”