So, would you propose a dedicated physical wire for every service?
You have to begin weighing the differences between physical vs. logical separation in network design at a point. MSR did two really good studies on network reliability, when they found that larger Aspen trees are significantly more reliable than smaller ones. If you're to build a large tree for every system, you're going to have a prohibitively expensive infrastructure.
In addition to this, the complexity of having multiple bespoke systems for every time someone calls something special, and dedicated results in higher system, and network administration cost. Because the scales of various networks are going to be different, their implementation will probably vary greatly as well, giving your engineers massive headaches. It's been shown that engineers, and software result in the highest number of failures in production systems.
Lastly, physical separation is no longer an easy excuse for security. At some point, that separation, intentionally, or unintentionally is going to get broken. It's best to begin operating from an idea that shared infrastructure will be more resilient, cheaper, and easier-to-operate.
> So, would you propose a dedicated physical wire for every service?
That's actually in a twisted sense what the article is advocating. Instead of relying on commodity communications, they advocate the city putting in their own infrastructure so they can do things like this.
The article is generally good, but that section is strange. You unequivocally do not need to run high speed internet to homes in order to move towards a "smart city" - wiring up those things they mention is orders of magnitude cheaper than running fiber to every home (funny how surveillance cameras is a suddenly a good thing, by the way).
You have to begin weighing the differences between physical vs. logical separation in network design at a point. MSR did two really good studies on network reliability, when they found that larger Aspen trees are significantly more reliable than smaller ones. If you're to build a large tree for every system, you're going to have a prohibitively expensive infrastructure.
In addition to this, the complexity of having multiple bespoke systems for every time someone calls something special, and dedicated results in higher system, and network administration cost. Because the scales of various networks are going to be different, their implementation will probably vary greatly as well, giving your engineers massive headaches. It's been shown that engineers, and software result in the highest number of failures in production systems.
Lastly, physical separation is no longer an easy excuse for security. At some point, that separation, intentionally, or unintentionally is going to get broken. It's best to begin operating from an idea that shared infrastructure will be more resilient, cheaper, and easier-to-operate.