Coercion of datetime columns to appropriate objects makes this a little easier.
d <- read.csv('https://dl.dropboxusercontent.com/u/14322966/20140401.csv')
d[, 1] <- as.POSIXct(d[, 1], format='%Y-%m-%d %H:%M:%S')
d[, 3] <- as.POSIXct(d[, 3], format='%Y-%m-%d %H:%M:%S')
xlim <- range(do.call(c,d[c(1,3)]))
plot(d[, 1:2], type='l', xlab='Time', ylab='Throughput (Mbps)',
xlim=xlim)
par(new = TRUE)
plot(d[, 3:4], type='l', col=2, axes=F, xlab='', ylab='',
xlim=xlim)
axis(4)
box(lwd=2)
Calculation of the range of datetime across the 2 columns allows for consistent xlims across the two plot calls.