Queueing Theory vs. Murphy's Law

 Fri, 17 Jul 2009

Well, I've just been through one of those weeks when every day seemed like Friday the 13th come on Halloween. And I swear this is due to either Murphy's Law or a statistical phenomena called queueing theory. Either way, sometimes, events just seem to cascade.

I have a fairly good size wireless network that covers several hundred square kilometers here in Chiriqui Province in the Republic of Panama. A couple of towers are located such that there may be one or two hops from tower to tower to get to some clients. On occasion, these links fail -- I accept that, part of the business. But it's rare to get two links that to fail simultaneously, and the odds against three links, all part of the same chain, failing all at once reside with the astronomical. So guess what? Yep, three links in a row all on the same day. Two failed at the same time, along with a computer system and an air conditioner. These I can explain in six words: wonderfully stable third world electrical power. Meaning, I really need to also become a power company and get off the (commercial) grid.

More than likely, a power spike took down all this equipment. After all, it did all fail at the same time, and power spikes have been the bane of my wireless deployment. All but for two solar-only tower sites (one at the end of the failing chain) have given me no end of headaches when the power twitches. The solar-only sites have generally been the most reliable and maintenance-free. Go figure.

Because the solar site with the failed card is so isolated, I have a program running called a "ping watchdog" that reaches out every 10 seconds to see if the next radio in the chain is still there. If the far system disappears for more than 60 seconds, the radio reboots (just in cast the problem lies with itself). Apparently, a reboot finally caught up with a failing card and it was "so long and thanx for all the fish" time.

I still remain faced with the task of finishing repairs along the way so I can take the long, slow trip up the side of the mountain to replace the failed radio card. But I doubt any of this would have happened like this were it not for Queueing Theory's need for fulfillment. In fact, I have a theory that theories tend to become self-fulfulling prophecies of a kind.

So, I'd like to propose that as a theory: theories tend to prove themselves despite anecdotes to the contrary (or despite evidence of God's demise).

Now at one point I promised myself I would not discuss God. Partly because those invoking the name of God usually do so while committing acts God would be ashamed to be party to, and partly because I'm agnostic (whatever that has to do with it). And since religion takes on a "religious" tone (I believe _this_, please don't confuse me with facts), I tend to avoid it and folks who feel the need to "share" their religion with me. And by-the-way, share in this context is akin to MicroSoft's End User License Agreements (EULA), agree with me or become food for my lawyers.

Am I jaundiced? Not at all. It would, however, be nice for some of these religious fanatics to practice what they preach -- tolerance. But then, if that happened, we would have missed out on most of the longest, bloodiest, and unnecessary wars (and terrorist events) the world has ever seen (and who needs that?). Long live religious zealotry (sigh).

But I'm off track. Bottom line, if I didn't know about Murphy's Laws or Queueing Theory, I suspect chain failures like the one I just experienced, would never happen, just like the grocery store lines: the shortest line is always the slowest moving; when you came in, all lines were empty, but the less items you buy, the longer the lines will be when you go to checkout (so never go in to buy just one item).

David-

Queueing Theory vs. Murphy's Law

Feedback count: 0

Comments are closed for this story.