SFWRENG 4E03 Lecture Notes - Lecture 15: Jackson Network, Mathtype, Exponential Distribution
11 Oct 2016
School
Department
Course
Professor
Document Summary
Mean value analysis continued (get stuff here and before) Performs better than balance equations or jackson network, but gives less info. Choose a reference, such as v1 = c i. Let (cid:540)(m 1) denote the rate at which i see arrivals to any node with m 1 jobs in the system. i. 1 i (cid:541)1 = 2, (cid:541)2 = (cid:541)3 = 1. Normal distributions can be weird, since it looks like bell curve. It also generates negative numbers, so chop off the tails on either side of your distribution at the same distance: chop off: ,0 and 2*mean, , a. k. a. bounded uniform distribution. The higher your variance, the worse your system will perform. In the real world, they generally use exponential with minimum value. This is because it has more variance than the other 2 distributions. Pareto distribution: an exponential which doesn"t start at 0 (a. k. a. zipfian) Just think: 99% controls 50% and 1% controls the rest.
