【Paper Study】Data Mining Meets Performance Evaluation: Fast Algorithm for Modeling Bursty Traffic



     A simple method to easily generate large traces and the resulting synthetic traces are very  realistic.
     It has the same desirable properties as the Poisson model.
     Its main goal of the paper is to describe the model and demonstrate its usefulness in a variety of domains.
     It is closely related to 80/20 law.
Related work
Poisson model
Hurst exponent
80/20 Law (Pareto principle)
b-model algorithm

INPUT: bias b, aggregation level n, total volume N
OUTPUT: Yt with 2^n points following the b-model
ALGORITHM: A stack is used to keep track of the data points.

1.Initialize the stack and push pair (0, N) onto the stack.
2.If the stack is empty, all the 2^n data points have been generated and the process ends.
3.Pop a pair (k, v) from the stack. If k=n, output v as the next data point and go back to Step 2.
4.Flip a coin. If heads, push pairs (k+1, v*b) and (k+1, v*(1-b)) onto the stack. Otherwise, push pairs(k+1, v*(1-b)) and (k+1, v*b) onto the stack. Go back Step 2.


After study this paper, I still don't know what is the pair (k, v). What does the pair indicate?
0 Responses