mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekay
abf99d05fc
Restructured from flat .tex/.pdf files into per-paper subdirectories. Added Antje Worring as co-author on all papers. Synced PDFs to pdfs/ for download site.
44 lines
1.7 KiB
TeX
44 lines
1.7 KiB
TeX
\section{Empirical Evaluation}
|
|
|
|
\label{sec:evaluation}
|
|
|
|
\subsection{Setup}
|
|
|
|
We deployed Flare on a 500-node testnet (same infrastructure as the Wave evaluation)
|
|
with parameters $k = 20$, $\alpha = 0.8$, $\beta_2 = 14$, and round interval 50ms.
|
|
|
|
\subsection{Certification Latency}
|
|
|
|
Under no Byzantine faults, the median certification time was 6 rounds (300ms), with
|
|
99th percentile at 12 rounds (600ms). The conviction counter reached $\beta_2 = 14$
|
|
without any resets in 97.3\% of transactions, indicating that once the population
|
|
converges, consecutive supermajority rounds are highly reliable.
|
|
|
|
With 20\% Byzantine validators, the median increased to 10 rounds (500ms), with the
|
|
99th percentile at 22 rounds (1.1s). The increase was due to occasional conviction
|
|
resets caused by adversarial equivocation during the convergence phase.
|
|
|
|
\subsection{Skip Rate}
|
|
|
|
The skip rate (fraction of transactions experiencing at least one conviction reset)
|
|
was:
|
|
\begin{itemize}[nosep]
|
|
\item 2.7\% with 0\% Byzantine validators (due to natural sampling variance)
|
|
\item 8.4\% with 10\% Byzantine validators
|
|
\item 18.1\% with 20\% Byzantine validators
|
|
\item 34.6\% with 30\% Byzantine validators
|
|
\end{itemize}
|
|
|
|
Even at 30\% Byzantine, transactions eventually certified within 35 rounds (1.75s),
|
|
confirming liveness.
|
|
|
|
\subsection{Throughput}
|
|
|
|
Flare achieved 2,650 TPS, slightly lower than Wave's 2,800 TPS due to the additional
|
|
certificate propagation overhead. The certificate messages add $< 1\%$ to total
|
|
bandwidth since they are issued once per transaction.
|
|
|
|
%% -----------------------------------------------------------------------
|
|
%% 8. RELATED WORK
|
|
%% -----------------------------------------------------------------------
|