Critical Chain - Eliyahu M. Goldratt [92]
"These computers are an excellent way to procrastinate," Mark agrees. "We were dealing with minute contentions, things that if we had to do them manually, we would never have bothered. But we removed all contentions. Then, of course, as Ted predicted, we had to fight with all the project leaders."
"To cut a long story short, we squeezed agreement," Fred summarizes days of fierce arguments into one sentence.
"Then reality showed us what fools we were. Any speculations on what happened?" Mark asks the class.
Everybody is thinking hard, but nobody comes up with any ideas. Not even Ted.
Mark doesn't wait long. "Did you ever see a step that finished somewhat late?" He gives them a clue. "One small deviation in one step and BOOM—you get the domino effect, contentions all over the place. We found ourselves wasting all our time sorting out fights. Ted, you called it a nightmare? You are absolutely right."
"I can see it clearly," Ted agrees. "It could easily happen in my place. So what did you do?"
"Before we see the solution," I say, "can you tell us the conceptual mistake you made?"
"We were treating estimates as if they were real," Ruth answers.
"What do you mean?" I prod.
"Suppose a step was supposed to take ten days. You know there is a good chance it might take seven or fifteen, but we fed ten days into the computer. Then we treated this number as holy."
"I still don't understand," I say.
"We regarded contentions of three days as significant even when the estimate for the path was thirty days."
"Basically," Fred clarifies further, "we fell into the trap of thinking that eight times eight equals exactly sixty-four; we were trying to be more precise than the noise. Everybody was fighting about contentions which, left alone, could have been easily absorbed by the buffers."
"As a result," Mark summarizes, "we were constantly changing schedules for nothing, and by that creating real problems."
"I understand," Brian says. "And thanks for telling us about it. Now I know what not to do myself, but I don't know what to do. We cannot ignore contentions.
"Absolutely," Mark agrees. "We cannot ignore contentions in one project, we definitely can't ignore them when we look at all projects."
"You see the dilemma we were caught in? Fred asks. On the one hand, we had to consider contentions, but on the other hand, when we did, we ran into the nightmare."
"So what did you do?" Brian is eager to find the answer.
"We called Professor Silver in to help."
"Which was totally unnecessary," I emphasize. "You knew the answer. You were just too lazy to realize it."
"That's unfair!" Ruth rebels. "Even after you showed us, it took me some time to digest."
"The answer was taught to you by Professor Fisher in your production course, and then it was elaborated on by Professor Wilson in your systems course."
I know that I'm unfair. I spent weeks working on it myself before I figured it out. But I want to open my students' minds to the possibility of transferring a good concept from one field to another.
"We are dealing here with resource contention problems," I start to explain. "Have you seen the same problem in your production course?"
"Of course," Brian says. "Every time there is a queue of tasks in front of a machine and the priorities are unclear, we have resource contention; a few tasks are fighting to be processed at the same time by the same resource."
"Exactly," I say. "And how do you handle such a case? You have learned that it is foolish to try to schedule the work of each and every machine. What are you supposed to do?" "Identify the bottleneck," Charlie says.
"And then?"
"Then exploit it; schedule the sequence of work for the bottleneck."
"By that," I say, "you have eliminated any contention