<aside> 🚧 This document will be completely rewritten according to the experiment plan written by Marcin. So, I don’t read the details in this document right now.

</aside>

I implemented several queuing mechanisms described by design doc written by Marcin. This document contains a series of simulations to find the optimal queuing mechanism.

Mix Propagation Time Comparison

Before starting detailed analysis, I measured the mix propagation time (from the sender to the last mix) of each message, and compared the results between queuing mechanisms.

• The mix propagation time means the total time taken for a message to reach the last mixnode after it was sent by the sender.

All the details can be found at the PR. Here, I mention only important configurations to understand the simulation results.

• Num nodes: 6
• Network topology: Random
  • Peering degree: 3
  • That is, each node chooses 3 nodes randomly from the entire network. The simulator ensures that no node is isolated.
• Max network latency: 100ms
• Mix path length: 5
  • For easier analysis, I configured all nodes to use the same mix path length when creating Sphinx packets.
• Min queue size: 10
  • i.e. the $B$ parameter defined in the designed doc.

The following simulations change the parameters below to understand the property of each queuing mechanism.

• GTR: How many messages are emitted via a single connection per second
• Sender lottery probability: Probability of being selected as a real message sender by running a single lottery
  • Lottery interval: 1 sec (fixed)

Low GTR & Low real message rate

• GTR: 10
• Sender lottery probability: 0.001

[Mix Propagation Times (sec)]
           NONE  PURE_COIN_FLIPPING  PURE_RANDOM_SAMPLING  PERMUTED_COIN_FLIPPING
count  7.000000            6.000000              6.000000                6.000000
mean   1.122000            6.663500              4.207000                4.403333
std    0.106276            0.867867              0.978162                1.395482
min    1.009000            5.771000              2.609000                2.840000
25%    1.024500            6.082500              3.981750                3.429250
50%    1.157000            6.350000              4.315000                4.108000
75%    1.174500            7.306750              4.458500                5.274250
max    1.290000            7.891000              5.626000                6.498000