C Program for simulating blocking probabilities in multiservice loss systems
----------------------------------------------------------------------------
Version 2.0 , 29. Sept. 2000
Author: Jan Hlinovsky, jan.hlinovsky@hut.fi
Organisation: Laboratory of Telecommunications Technology, 
              Helsinki University of Technology

Contents:
1. Copyright information
2. Compiling the program
3. Input file format
4. Running the program
5. References

1. Copyright information
------------------------
The program can be used and modified freely. No claims are made
about the correctness of the program and no liability is taken 
for any damage caused by the use of the program. 

2. Compiling the program
------------------------
In Unix, the program can be compiled using the make utility. 
Make sure you have all the following files:
        main.c
        ini.c
        read.c
        simu.c
        ini.h
        read.h
        simu.h
        structs.h
        Makefile
The Makefile is written to be used with gcc, so if you want to use some 
other compiler and linker, replace the corresponding lines in the Makefile. 
The program uses two functions for random number generation that are 
not part of the standard C library,  namely srand48() and drand48(). 
Check that you have these (e.g. "grep [sd]rand48 /usr/include/stdlib.h" ).
If you don't have these functions you may want to replace them with some
other random number generator that returns a value between 0 and 1.
If you have gcc and the above mentioned functions, just type "make" in
the directory where the source files and the Makefile are.

3. Input file format
--------------------
An input file must be a text file and have the following properties:

        a) content of an input file is (in following order):
                number of classes (integer)
                number of links   (integer)
                load vector rho   (real)
                capacity vector   (integer)
                matrix of bandwidth requirements; (integer)
                        NOTE:
                        nr of rows = nr of classes
                        nr of columns = nr of links
        b) Every scalar number, vector, or matrix row is on its own line.
        c) Lines beginning with '%' are ignored (comments).
        d) Following letters are ignored: []{},; (blocks, braces, comma, semicolon).
        e) Any number of whitespace (space, tabs, newlines) is allowed, except
           for rule b): no newlines inside a vector or a matrix row.

Examples of input files:
-----------------------
        The following file is a legal input file:
% beginning of example file 1 -----------
2
3
35 22
100 120 170
2 0 2
0 3 3
% end of example file 1 ------------------

        The following is also legal, and more understandable input file:
% begining of example file 2 -------------
%number of classes
2
%number of links
3
% load vector rho
[35, 22]

% capacity C
[100, 120, 170]

%B
[[2, 0, 2]
 [0, 3, 3]]
% end of example file 2 -------------------


4. Running the program
----------------------

The program is started with the command ./netwsim.
The input file may be given as an argument, otherwise it is asked for. 
Then the program asks for number of batches and batch size. 
The batches are used for calculating the 95% confidence interval, so the
number should be several hundred. The number of samples in one batch should
be a multiple of 1000, for optimal allocation of samples. 

After this the program will ask which class you are interested in.
The program will write the point estimate of blocking probability and its
95 % confidence interval for the class that is being simulated, and the 
amount of processor time used for the simulation. Note that for very long 
simulations (e.g. hours) the clock may "overflow".


example:
~ % ./netwsim inputfile

Enter number of batches:400

Enter number of samples in one batch (min. 1000):5000

For which class (number) you want to run the simulation?:2

Traffic class    Point estimate of Bk           95% conf int
==================================================================
         2             0.0130674          [0.0130644 , 0.0130704]

Simulation time: 19.490 sec



5. References
-------------
The program uses a sophisticated inverse convolution method for
calculating blocking probabilities. The method is described in
the following paper: 

    P. Lassila, J. Virtamo, "Nearly Optimal Importance Sampling for 
    Monte Carlo Simulation of Loss Systems", to appear in ACM TOMACS, 
    January 2001.

The paper is available from http://www.tct.hut.fi/tutkimus/com2/