Generating a graph with multiple (sets of multiple sets of multiple) X-axis data sets

Question

I am looking for a way to generate a graph with multiple sets of data on the X-axis, each of which is divided into multiple sets of multiple sets. I basically want to take this graph and place similar graphs side by side with it. I am trying to graph the build a graph of the duration (Y-axis) of the same jobs (0-3) with different configurations (0-1) on multiple servers (each group with the same 8 jobs). Hopefully the following diagram will illustrate what I am trying to accomplish (smaller groupings are separated by pipes, larger groupings by double pipes):

|| 0 1 | 0 1 | 0 1 | 0 1 || 0 1 | 0 1 | 0 1 | 0 1 || 0 1 | 0 1 | 0 1 | 0 1 ||
|| 0   | 1   | 2   | 3   || 0   | 1   | 2   | 3   || 0   | 1   | 2   | 3   ||
|| Server 1              || Server 2              || Server 3              ||

Is this possible with either the GD::Graph Perl module or the matplotlib Python module? I can't find examples or documentation on this subject for either.

Answer 1

Here's some Python code that will produce what you're looking for. (The example uses 3 configurations rather than 2 to make sure the code was fairly general.)

import matplotlib.pyplot as plt
import random

nconfigs, njobs, nservers = 3, 4, 4

width = .9/(nconfigs*njobs)  
job_colors = [(0,0,1), (0,1,0), (1,0,0), (1,0,1)]

def dim(color, fraction=.5):
    return tuple([fraction*channel for channel in color])

plt.figure()
x = 0
for iserver in range(nservers):
    for ijob in range(njobs):
        for iconfig in range(nconfigs):
            color = dim(job_colors[ijob], (iconfig+2.)/(nconfigs+1))
            plt.bar(x, 1.+random.random(), width, color=color)
            x += width
    x += .1

plt.show()

This code is probably fairly transparent. The odd term (iconfig+2.)/(nconfigs+1) is just to dim the colors for the different configurations, but keep them bright enough so the colors can be distinguished.

The output looks like:

Answer 2

Recently, I saw a graph that I think does what you want using protovis

I have no experience with the program, but the graph was enlightening and I think would give you want you want.

Answer 3

MathGL can do it easily and it have Python interface too. See this for examples.