Back Of The Envelope Calculation

Also known as a Fermi Approximation - one of the most useful tools any engineer, scientist, manager, mathematician or programmer can have.

How much water flows out of the Mississippi in a day?
On average, what thickness of tire do you lose each time it goes around?
How many piano tuners are there in London?
If everyone in your city threw a ping-pong ball into your living room, how deep would they be?
How many intelligence-bearing planets are there in the universe?
- N=R*fs*fp*ne*fl*fi*fc*L
  - Yes, very good, now estimate each of those figures.
    - Let me empty out my phone bill first
    - 1 <= N < 10^22 (and a preemptive shush to the wag that will imply Earth has no intelligence)
How many telephone numbers are in use in the world?
- Not sure but it will dramatically increase once Ip6 and ENUM are standard

The most powerful mathematical calculations are those done "on the back of an envelope".

(In a pinch, napkins will suffice. ;-)

Q: You guys can pay 10,000 people in 8 hours. To do the hourly payroll you'll have to do 10 times that many in the same time. Why do you think you can do it?

A: GemStone 5.x is supposed to be 4 times faster than the version we use now, and we're assuming we'll get only a factor of 2. We did an experiment a while back paying on 5 CPUs instead of 1. It went five times faster. We have 12 CPUs available on the machine, and we're assuming we'll get only a factor of 8. GemStone 5 plus paying in parallel should give us a factor of at least 16 in performance. We only need 10. That's why we're not worried.

The BigDesignUpFront consultant and several inexperienced coders were killing the project with complex multi-threaded architecture, application-defined locking schemes, signals between processes, etc. "What volume are you expecting?" I asked, and "how large are your transactions?" "Lots and lots" they said. "You just have no idea how much work we'll be doing!!!" they said. So I looked at their transactions and counted their customers: A few hundred lines per transaction, with relatively simple processing. Several thousand customers, but with work process rules that limit when and how they can gather (and hence send us) the data. Result: Roughly one transaction every three minutes, with no reason to believe there will be significant "peak periods" of high load. "I think Oracle can handle that," I said.

(Now to kill the multi-threading, synchronization, background processing, and other hairy nonsense that was shoveled in...)

The great mathematician Stanislaw Ulam recounted that the tables of the (Scottish) café in Lvov (Ukraine) had marble tops, so they could write in pencil, directly on the table, during their discussions.

See http://encyclopedia.thefreedictionary.com/Scottish%20Caf%E9

CategoryJargon