Where graph evolves over time.

Depends on #7 and being able to identify deadlock in a simulation.

Find references about deadlock, list and we can discuss here perhaps? (Might not be called this throughout)

These show the distribution slightly better than boxplots...

http://matplotlib.org/examples/statistics/violinplot_demo.html

Could you read through this, make sure its clear etc. @drvinceknight
I will be adding more on the lit review at some point when I find more papers

You will want tikz code, Sage and networkX might be able to produce the tikz for you (not sure but am curious to find out).

Call a deadlocked state something other than -1? As when we have more than 1 type of deadlock ill want to distinguish between then.

For the 2 node network, simulating for 1000 iterations:
Base parameters:
L1 = 4.0
L2 = 5.0
mu1 = 10.0
mu2 = 8.0
r12 = 0.25
r21 = 0.15
n1 = 3
n2 = 2

• L1, 2.5-14 in steps of 0.5
• L2, 2.5-14 in steps of 0.5
• mu1, 2.5-14 in steps of 0.5
• mu2, 2.5-14 in steps of 0.5
• r12, 0.1-1.0 in steps of 0.05
• r21, 0.1-1.0 in steps of 0.05
• n1, 0-6 in steps of 1
• n2, 0-6 in steps of 1

Perhaps look at a dummy example in each.

• Simulate MC
• Find median analytically
• Run simulation for much longer
• Plot difference for given parameter set and increasing the number of runs

Do so analytically (brute force: no need for anything smart).

Following #8 get the analytical version for other topologies:

• Smaller (single q with feedback)
• Bigger (2 queues with also having feedback loop on each queue...)