itm-doc.txt

it is very important

$Id: itm-doc.txt,v 1.1 1996/10/04 13:29:13 ewz Exp $

Name: itm 

Function:  Driver to create flat random graphs and two forms
	   of hierarchical graphs (N-level and transit-stub).

	   See models.ps for a more complete description of
	   the methods.  This file describes the syntax and
	   (limited) semantics of the parameter specification file.

Usage:  itm <spec-file0> <spec-file1> ....

Details:

File Format
-----------
    
    Each argument is a file containing ONE set of specs for 
    graph generation.  Such a file has the following format:

       [# comment line]
       <method keyword> <number of graphs> [<initial seed>]
       <method-dependent parameter lines>

    <method keyword> is one of:
	"geo":  flat random graph 
        "hier": N-level hierarchical graph
	"ts":   transit-stub graph

    <number of graphs>: number of graphs of specified type to generate
    <initial seed>: initial random number seed; optional

    Any line preceeded by # is a comment and is ignored by itm.


Method-dependent Parameters
---------------------------

    All of the methods make use of the following <geo_parms>
    line, which specifies the parameters for a flat random
    graph.  Six different types of flat random graphs are supported.
    (See models.ps for more detail on parameters and edge
    generation methods.)

    <geo_parms> ::= <n> <scale> <edgemethod> <alpha> [<beta> <gamma>]

       <n>: number of nodes in graph
       <scale>: one-sided dimension of space in which nodes are distributed
       <edgemethod>: method for generating edges; valid range 1..6
		1: Waxman 1 
		2: Waxman 2
		3: Pure random
		4: Doar-Leslie
		5: Exponential
		6: Locality
	<alpha>: random graph parameter (0.0 <= alpha <= 1.0)
	<beta>: random graph parameter (0.0 <= beta)
	<gamma>: random graph parameter (0.0 <= gamma)


   The <method-dependent parameter lines> are as follows:
  
   <"geo" parms> ::= <geo_parms>

   <"hier" parms> ::= <number of levels> <edgeconnmethod> <threshold>
                      <geo_parms>+  {one per number of levels}

	<number of levels>: number of levels in hierarchy
	<edgeconnmethod>: method of resolving edges  
		(NOTE: probably should just use method 0)
		0: random
		1: use non-leaf node of smallest degree 
	 	2: use node of smallest degree
		3: use first node with degree less than <threshold> 
	<threshold>: see above 

   <"ts" parms> ::= <# stubs/xit> <#t-s edges> <#s-s edges>
                    <geo_parms>       {top-level parameters}
                    <geo_parms>       {transit domain parameters}
                    <geo_parms>       {stub domain parameters}

	<# stubs/xit>: avg number of stub domains attached per transit node
	<#t-s edges>: number of extra transit-stub edges
	<#s-s edges>: number of extra stub-stub edges


Output File Naming Convention:
------------------------------
	
    The i'th graph created with the parameters from file "arg" is placed
    in file "arg-i.gb", where the first value of i is zero.


Example of Usage:
-----------------

The following is a valid specification file called r10:


	# <method keyword> <number of graphs> [<initial seed>]
	# <n> <scale> <edgemethod>  <alpha> [<beta>] [<gamma>]
	geo 3 
	10 10 3 .2 

If we run 'itm r10' then three pure random graphs of 10 nodes
each will be created.  The nodes of each graph will be generated
in a 10 by 10 (logical) grid.  The probability of an edge is 0.2, as
given by the <alpha> parameter; <beta> and <gamma> are not needed
for this method.  The output files will be named: r10-0.gb, r10-1.gb, 
r10-2.gb.  (This specification file and the resulting output files
can be found in sample-graphs/rand/r10.)


Another Example of Usage:
-------------------------

The following is a valid specification file called t100:

        # <method keyword> <number of graphs> [<initial seed>]
	# <# stubs/xit> <#t-s edges> <#s-s edges>
	# <n> <scale> <edgemethod> <alpha> [<beta>] [<gamma>]
	# number of nodes = 1x4x(1+3x8) = 100
	ts 10 47
	3 0 0
	1 20 3 1.0
	4 20 3 0.6
	8 10 3 0.42
	

If we run 'itm t100', then 10 transit-stub graphs of 100 nodes each
will be created, starting with initial seed 47.  Each graph will have
three stub domains per transit node, with no extra transit-stub
or stub-stub edges.  The line "1 20 3 1.0" says that there
will be one transit domain.  The next line "4 20 3 0.6" specifies
transit domains have (on average) four nodes, and an edge between each
pair of nodes with probability 0.6.  The last line "8 10 3 0.42" says
that each stub domain will have (on average) eight nodes, and edge
probability 0.42.  The graphs will be named t100-[0-9].gb.