The session 1 parameters are defined in ‣.
Experiment 5
Using only the first 10000 sent messages to calculate ordering coefficients. Other metrics (such as latency) have been calculated from the full result.
Experiment 6
$T=10000$
This is a summary of analyses that are described in more detail in the following sections.
PureRandomSampling and PermutedCoinFlipping have similar latency, which is much lower than other queue types, in both Experiment 5 and 6.PureCoinFlipping has more than 2x latency of PureRandomSampling or PermutedCoinFlipping.NoisyCoinFlipping and NoisyCoinFlippingRandomRelease have the worst latency.NoisyCoinFlippingRandomRelease shows the lowest ordering coefficient. PureRandomSampling and PermutedCoinFlipping have similar coefficients but they are about 100 times higher than NoisyCoinFlippingRandomRelease.NoisyCoinFlippingRandomRelease, PureRandomSampling, and PermutedCoinFlipping show the similar lowest ordering coefficients.PureCoinFlipping has much higher ordering coefficients than the three queue types above. It is sometimes higher than NonMix (especially its min value).NoisyCoinFlipping has the worst ordering coefficient in both experiments.PureRandomSampling and PermutedCoinFlipping offer the best balance between latency and mixing quality.NoisyCoinFlippingRandomRelease shows the lowest ordering coefficient but its latency is one of the worst.<aside> 💬
The following sections show the detailed plots and tables for the analyses of latency and ordering coefficients. Due to the space limitations in Notion, the plots may not be very readable. If it’s the case, it may be easier to download the attachment below and open plots in the larger screen. The attachment contains plots and tables for all experiments.
</aside>
For accurate comparisons, the following analyses fixed all other parameters except the x-axis field (e.g. num_mixes or peering_degree) to those in the highest paramset.
num_mixes
