Excerpts from the book
X-Events, Resilience, and Human Progress
John L. Casti
Roger D. Jones
Michael J. Pennock
Click to Buy Paperback
HOW RESILIENT ARE YOU?
The first item on the road to a measure of resilience is to remember that resilience is a context-dependent systemic property. It depends on the type of extreme event (X-event) you want to be resilient against. This point must be kept uppermost in mind, since it does no good to build resilience against a terrorist bombing when what happens is a hurricane or an Internet crash. So to set the stage for our Four As resilience measure, let’s assume we want to be resilient to an Inter- net failure. So that’s our target X-event. But even that’s not enough. We have to specify a few more details about the failure before we start measuring how well our protection is working. These details include whether it’s a purely in-house IT failure of a server or some other part of the local information-processing infrastructure or the failure goes outside your own organization. In the latter case, we then must ask if it’s local in the sense that the failure rests in your local ISP or if it’s global in the sense that the failure extends beyond simply your own provider. We can further subdivide the problem. But this is probably enough to get the general idea, which is that you have to have a very good idea of what you’re trying to be resilient against before you can even begin to put protection in place.
Once the matter of exactly what X-event is of concern is settled, we can bring out the Four As: Awareness, Assimilation, Agility, and Adaptivity and assign an integer from 0 to 10 to each category, reflecting how well-prepared we are for the X-event in the activities described by that category. So if our X-event is an in-house internet failure, we can first look at whether we had in place any kind of procedure for giving an early-warning signal for that particular event. In other words, how was our Awareness of an in-house failure? If we had a good early-warning of the failure, then we’d give a high mark, say 9 or 10, to Awareness; if we had no warning at all, then for Awareness we would assign a low value, say 0 or 1.
We now continue to Assimilation and ask how well we were prepared to survive the Internet failure and still keep at least a skeletal system going. Again, a high number means we did pretty well, a low number says we did poorly. We then continue with Agility, our ability to examine the situation after the Internet failure and look for opportunities that the failure created. Finally, we consider how Adaptive we are in the sense of being able to move into a new way of carrying out our in-house computing created by the failure.
In this manner, we obtain a set of four numbers (A1, A2, A3, A4) characterizing how resilient our system is to an Internet failure. If all of these numbers are large, great. The system is very resilient. But if even one of the numbers is small, then we have a potential problem and our resilience level is low. So perhaps the risk-averse way to measure resilience against an in-house Inter- net failure is to just take the smallest of the four numbers A1, A2, A3, A4. Or maybe the average of these numbers. Or even the variance of the four numbers. Which is the best way to reflect the resilience against this type of X-event also depends very much on the specific system we’re managing and the details of the hypothetical X-event.
Now recall that so far we’ve spoken about just one specific X-event. But there may be many more that concern the system manager. And for each of these X-events of concern, there is another set of four numbers characterizing how prepared we are for that X-event using the Four As categories. In the end, if we have N X-events, we’ll end up with a collection of N sets of four numbers. We then have to somehow combine that set of 4-tuples into a single number that reflects our estimate of the overall resilience of our system. Again, there are many ways to create that single measure, ranging from taking the minimum of all the other minima to some kind of statistical measure like an average. What works best is an empirical question that’s currently under investigation.