does any successful system need REDUNDANCIES to function properly?

Question

some examples: bridge supports that will prevent a catastrophe when one component fails... a predator/prey relationship in an ecosystem to control the population... checks and balances to theoretically control the power of a government...

does a system need a way to regulate itself to perform its intended functions properly? how long can a system without these safeguards built in operate at all?

Answer 1

Not all systems require redundancies to function properly and safely. It depends on the likelihood of failure and the severity of the consequences of failure.

If life and limb are at risk when the system fails, redundancy is a good idea and is often required by law.

In most structural solutions (for example, an airplane wing or a highway bridge), the means of ensuring safe operation is called a Safety Factor, which usually means building the critical parts of the structure to be stronger than required to some specified multiple. For example, a commercial airliner is built with a safety factor of ten (10). That is, the framework can withstand 10 times the load required in normal flight.

Safety factors of 10 also apply to many parts of a highway bridge. The problem is that corrosion can eat into the safety factor. The solution is inspection, and also humility, which government safety agencies often fail to exercise.

Answer 2

Redundancy is mostly a contingency measure, it attempts to provide better reliability by keeping the option of internal replacement of a defective part without affecting the total operations. It cannot be compared with the prey/predator thingy, because thats a dynamic relationship.

A good system needs a way to regulate itself else it might over exert itself to disaster. For example excess inputs needs to be rejected, or excess outputs needs to be controlled else some part of the system will be starved and so on.

The electricals in a house can function without the fuse, but it is better to have the fuse in as a safety precaution. That should be a good example for you.

Answer 3

Be it human operated or totally autonomous systems, they are never perfect. A system that is totally perfect would violate the laws of physics.

The performance of every system is constrained by asymptotic behaviour. When you step on the brakes the perfect thing to happen would have the car stop right that moment. Fourier analysis on an abrupt square wave would prove that such abrupt perfection to stop even a little momentum requires infinite energy.

Having established that a perfect system is inherently imperfect, we would have to call into question the reliability of a system. A water bottling factory wants to bottle up and sell us bottles of 0.5 litre of water. The bottling factory is completely automated. Will every bottle have exactly 0.5 litre. The decades of Quality Engineering tells us that would not be the case. The label on the bottle says "0.5 litre" by the virtue of "0.5 litre" being the average measured over millions of bottles sold. Some might have 0.51 litre while others might have 0.489 litre.

If the bottling company wants to project an image of precision, it might spend a lot of money to ensure that the volume bottled is between 0.4999999 and o.5000001 litre. Is it worth it?

Similarly, in the deployment of redundancy. Say a system has an offline rate of 0.95. Which means, on the average of every 100 hours of operation, it has to be offline for 5 hours either due to failure or maintenance. One of the most straight forward way to increase the reliability of that system is to have a redundant system. Now, we could wonder and say we have two systems running in parallel at 0.95 reliability. We have got it covered more than 100% - not so quick, because we have to consider the probability of both machines going down at the same time. So after considering all the peripheral possibilities we might conclude that we managed to increase the reliability to 0.998. If that two hours out of a thousand is so important, why don't we deploy a third machine to improve the reliability to 0.999998? How about another redundant machine to mitigate it further? When should we stop? That is the asymptotic dilemma.

We stop aligning more machines into the system when the cost of failure equals the cost of mitigation. That is, the expected cost and effects of failure is equal to the cost of improvement.

Whether one believes in Evolution or Creation or both, you will notice that this Universe has been "designed" (by Evolution or otherwise) to operate "imperfectly" and with redundancies. As demonstrated by better answers before mine, the genius of this Universe is that it's imperfect. It is this imperfection and their spontaneous redundancies that give rise to mutation, speciation and consequently diversity. Whether one believes in the fallacy of "the original sin", the Creator would have been bored to death if every rose was the same perfect shade of red. Without sin or deviation, the Universe would not be as beautiful as it became. Without imperfection, there would not even be roses.

**I wish to comment of Irv's differentiation between balancing and redundancy. In some schemes of redundant processing, redundant processors not only check on each other's calculation, they also compete with each other for work. Merging the roles of fail-safe and load-balancing. Many naturally occuring redundancies are both competitive and cooperative to reach an equilibrium. Therefore, redundancy and equilibrium are often inseparable effects.

Answer 4

Interesting question!
I think redundancy is one of the basic properties of complex systems. In biological systems, it provides the basic material for evolution by supplying "extra" elements that can be coopted for different functions. In man-made systems, redundancy ensures that there will be a backup element when the main ones fail.

However, maybe a concept closer to what you're referring to is *degeneracy*, i.e. the ability of structurally different elements to perform the same function. Thus, diverse mechanisms in the bridge that might prevent it from collapsing, so that, if electronic devices fail, there will be mechanical elements that still do the job. Biologically speaking, our genetic code is degenerate, and this ensures less mistakes in translation. Food webs are examples of redundancy and degeneracy.
And yet another concept is that of negative (or positive) feedback (such as predator-prey), and the flow of information in these complex systems. These are necessary to control and self-regulate how these work.

While these properties are necessary for a system to regulate itself, I don't know if they're mandatory for it to "just" work. I guess if they're absent, it takes much more external monitoring, control, and additional input (subsidy) for the system to function in the long term. Or at least that's the case in man-made agricultural systems compared with natural ecosystems.

BTW I'm very interested in these matters, and thanks to your question I found a couple of papers that I will read. In case you're interested:
http://www.its.caltech.edu/~theory/edelmangally.pdf
http://www.indiana.edu/~cortex/degeneracy.pdf

Answer 5

You confuse redundancy with dynamic balances.
Predator / prey is a dynamic balance.
The UPS for your computer to maintain power and save your files in the event of a grid failure is a redundancy.
"Balance" is almost by definition, a delicate thing, subject to disturbance. Since we live in a random universe full of disturbances, anything that lasts must have some stabilizing influence.
Redundencies abound in nature. You have two kidneys don't you?

Answer 6

Natural counterexample - The moon has managed to stay in orbit, without backup systems, for quite some time now.

Engineered system counterexample - Deep submergence pressure hulls. No backup/redundancy possible. If it fails at all, you're dead.

Answer 7

All systems were humans are the main, need checks and
balances to keep it honest.