Add sources for buildtimes report.

### Copyright 2007 Steven J. Murdoch, 2009 Karsten Loesing

### See LICENSE for licensing information

UFACTOR = 1e6

## Subtract to timevals, maintaining precision

todelta <- function(startsec, startusec, tsec, tusec) {

  tsec[tsec == 0] <- NA

  dsec <- tsec - startsec

  dusec <- tusec - startusec

  return(dsec*UFACTOR + dusec)

}

parsedata <- function(filename) {

  t = read.table(filename, header=TRUE)

  ## Make sure that the table has column names

  if (is.null(t$startsec)) {

    t <- read.table(filename, col.names=c("startsec", "startusec",

      "socketsec", "socketusec", "connectsec", "connectusec",

      "negotiatesec", "negotiateusec", "requestsec", "requestusec",

      "responsesec", "responseusec", "datarequestsec", "datarequestusec",

      "dataresponsesec", "dataresponseusec", "datacompletesec",

      "datacompleteusec", "writebytes", "readbytes"))

  }

  tStart <- t$startsec*UFACTOR + t$startusec

  dSocket <- todelta(t$startsec, t$startusec, t$socketsec, t$socketusec)

  dConnect <- todelta(t$startsec, t$startusec, t$connectsec, t$connectusec)

  dNegotiate <- todelta(t$startsec, t$startusec, t$negotiatesec, t$negotiateusec)

  dRequest <- todelta(t$startsec, t$startusec, t$requestsec, t$requestusec)

  dResponse <- todelta(t$startsec, t$startusec, t$responsesec, t$responseusec)

  dDRequest <- todelta(t$startsec, t$startusec, t$datarequestsec, t$datarequestusec)

  dDResponse <- todelta(t$startsec, t$startusec, t$dataresponsesec, t$dataresponseusec)

  dDComplete <- todelta(t$startsec, t$startusec, t$datacompletesec, t$datacompleteusec)

  cbWrite <- t$writebytes

  cbRead <- t$readbytes

  results <- data.frame(tStart, dSocket, dConnect,

                        dNegotiate, dRequest, dResponse,

                        dDRequest, dDResponse, dDComplete,

                        cbWrite, cbRead)

  invalid <- t$datacompletesec == 0

  results[invalid,] <- NA

  invalid <- t$dataresponsesec == 0

  results[invalid,] <- NA

  return(results)

}

plotboxes <- function(e1, e2, labels, title, ylim=c(NA,NA)) {

  range <- 1.5

  MinY <- ylim[1]

  MaxY <- ylim[2]

  ## Find how many points this will cause to be skipped

  skipped <- c()

  labels[1] <- paste(labels[1], " (", length(na.omit(e1)), ", ", length(e1) - length(na.omit(e1)), ", ", length(e1[e1 > MaxY]), ")", sep="")

  labels[2] <- paste(labels[2], " (", length(na.omit(e2)), ", ", length(e2) - length(na.omit(e2)), ", ", length(e2[e2 > MaxY]), ")", sep="")

  e1[e1 > MaxY] <- NA

  e2[e2 > MaxY] <- NA

  ## Plot the data

  boxplot(e1, e2, frame.plot=FALSE, axes=FALSE, ylab="Time (s)", range=range,

          ylim=c(MinY, MaxY), xlab="Configuration size (# runs, # timeouts, # points omitted)", main="",

          pars=list(show.names=TRUE, boxwex = 0.8, staplewex = 0.5, outwex = 0.5))

  axis(1, at=1:length(labels), labels=labels, lwd=0)

  axis(2, las=1)

}

plotdensity <- function(data1, data2, title, maxX, dens) {

if (dens == 1) {

  plot(density(na.omit(data1/1e6)), col="red", axes=FALSE, xlim=c(0,maxX), main=title, xlab="Time [s]")

  lines(density(na.omit(data2/1e6)), col="blue")

} else {

  plot(density(na.omit(data2/1e6)), col="blue", axes=FALSE, xlim=c(0,maxX), main=title, xlab="Time [s]")

  lines(density(na.omit(data1/1e6)), col="red")

}

median1 <- median(data1/1e6, na.rm=TRUE)

median2 <- median(data2/1e6, na.rm=TRUE)

abline(v=median1, col="red", lty=2)

abline(v=median2, col="blue", lty=2)

adj1 <- 0

adj2 <- 1

if (median1 < median2) {

  adj1 <- 1

  adj2 <- 0

}

mtext(paste("md =", format(median1, digits=3)), side=3, at=median1, col="red", adj=adj1)

mtext(paste("md =", format(median2, digits=3)), side=3, at=median2, col="blue", adj=adj2)

axis(1, lwd.ticks=1, lwd=0)

axis(1, at=c(0, maxX*65/60), labels=FALSE, lwd.ticks=0)

legend(x=maxX, xjust=1, y=par("usr")[4], fill=c("blue", "red"), ncol=1, legend=c("build-time patch", "unpatched"), bty="n")

}

plotdensities <- function(filename, data1m, data2m, maxm, densm, data1g, data2g, maxg, densg, title) {

pdf(filename, width=8, height=6)

par(mfrow=c(2, 1))

par(mar=c(4.3,0.1,3.6,0.1))

plotdensity(data1m, data2m, paste(title, "(50 KiB, moria)"), maxm, densm)

plotdensity(data1g, data2g, paste(title, "(50 KiB, gabelmoo)"), maxg, densg)

dev.off()

}

m_orig <- parsedata("moria-50kb.data")

m_circ <- parsedata("moria-mike50kb.data")

median(na.omit(m_circ[,6]))

g_orig <- parsedata("gabelmoo-50kb.data")

g_circ <- parsedata("gabelmoo-mike50kb.data")

m_orig <- subset(m_orig, tStart >= m_circ$tStart[1])

g_orig <- subset(g_orig, tStart >= g_circ$tStart[1])

plotdensities("connected.pdf", m_orig[,6], m_circ[,6], 5, 2, g_orig[,6], g_circ[,6], 5, 2, "Time to connect to website")

plotdensities("firstbyte.pdf", m_orig[,8], m_circ[,8], 15, 2, g_orig[,8], g_circ[,8], 15, 2, "Time until receiving first response byte")

plotdensities("download.pdf", m_orig[,9], m_circ[,9], 30, 2, g_orig[,9], g_circ[,9], 30, 2, "Time to complete request")

\documentclass{article}

\usepackage{url}

\usepackage[dvips]{graphicx}

\usepackage{graphics}

\usepackage{color}

\begin{document}

\title{Dynamically Adjusting Circuit Build Timeouts}

\author{Karsten Loesing}

\maketitle

\section{Introduction}

Current Tor clients use a fixed timeout when deciding whether to give up on

building a circuit or not.

In this experiment, Tor clients adapt their circuit build timeout from

observations of previously successful circuit build times.

\section{Results}

The first measured time is the time to

connect to the website. This step requires the Tor client to create a

circuit using the specified exit and send a \texttt{CONNECT} command to the

exit relay. As soon as the exit relay replies with a \texttt{CONNECTED}

cell, the connection to the website is established.

Figure~\ref{fig:connected} shows kernel density estimates of the connection

times.

\begin{figure}

\centering

\includegraphics[width=.85\textwidth]{connected.pdf}

\caption{Kernel density estimates of times to connect to website}

\label{fig:connected}

\end{figure}

The second measured time is the time between starting a request and

receiving the first byte of the response. Figure~\ref{fig:firstbyte} shows

these times for the three requested file sizes.

\begin{figure}

\centering

\includegraphics[width=.85\textwidth]{firstbyte.pdf}

\caption{Kernel density estimates of times until receiving first response byte}

\label{fig:firstbyte}

\end{figure}

The third measured time is the overall time to complete a request. This

time include all steps from connecting to the website until having received

all bytes of the response. Figure~\ref{fig:download} shows the

distributions of these times.

\begin{figure}

\centering

\includegraphics[width=.85\textwidth]{download.pdf}

\caption{Kernel density estimates of times to complete request}

\label{fig:download}

\end{figure}

\end{document}