R - histogram and scatter plot

histo.R

library (ggplot2)
library (gridExtra)

# arg1 - online csv
# arg2 - base filename

options <- commandArgs(trailingOnly = TRUE)

basename = options[2]

csv_online = read.csv(options[1])
print("Data loaded")
csv_online$timeStamp = (csv_online$timeStamp - min(csv_online$timeStamp)) / 1000
print("Data modified")

min_resp <- min(csv_online$Latency)
max_resp <- max(csv_online$Latency)
pctl99_resp <- quantile(csv_online$Latency, probs=0.99, names=FALSE)
pctl90_resp <- quantile(csv_online$Latency, probs=0.90, names=FALSE)
pctl50_resp <- quantile(csv_online$Latency, probs=0.50, names=FALSE)
stddev <- sd(csv_online$Latency)
total_resp <- length(csv_online$Latency)
failed_resp <- sum(csv_online$responseCode != 200)
failed_perc <- failed_resp/total_resp * 100
long_resp <- sum(csv_online$Latency > 300)
long_perc <- long_resp/total_resp * 100
ok_resp <- sum(csv_online$responseCode == "200")
default_resp <- sum(csv_online$bytes != 4269 & csv_online$responseCode == "200" )
duration <- max(csv_online$timeStamp)
print("Data extracted")

bounds_df = data.frame(
                Duration=duration,
                Min=min_resp,
                Max=max_resp,
                Median=pctl50_resp,
                PCTL90=pctl90_resp,
                PCTL99=pctl99_resp,
                StdDev=round(stddev, digits=2)
                )
req_df = data.frame(
                    Total=total_resp,
                    Failed=failed_resp,
                    FailedPercentage=round(failed_perc, digits=2),
                    Long=long_resp,
                    LongPercentage=round(long_perc, digits=2),
                    OkResp=ok_resp,
                    DefaultResp=default_resp
                       )

scatter <- ggplot(csv_online, aes(x=timeStamp, y=Latency)) + geom_point(alpha=1/6, size=1.5, aes(colour=success)) +
      labs(x="timestamp (s)", y="response time (ms)", title="Response time over time")

        

histo <- ggplot(csv_online, aes(x=Latency)) + geom_histogram(binwidth=25, origin=0) +
      scale_y_sqrt() +
      scale_x_continuous(limits=c(0,max_resp)) +
      labs(x="response time (ms)", y="number of responses", title="Response Time Distribution (25ms bins)")

bounds_table <- tableGrob(bounds_df, show.rownames=FALSE, cols=expression(
                                                                          "Dur (s)",
                                                                          "Min (ms)",
                                                                          "Max (ms)",
                                                                          "Median (ms)",
                                                                          "90th PCTL (ms)",
                                                                          "99th PCTL (ms)",
                                                                          "Std Dev"))

req_table <- tableGrob(req_df, show.rownames=FALSE, cols=expression(
                                                                    "Total # Req.",
                                                                    "# Failed Req.",
                                                                    "% Failed Req.",
                                                                    "# Req. >300ms",
                                                                    "% Req. >300ms",
                                                                    "# 200 OK",
                                                                    "# Def. Cont."))
dpi <- 72
png(paste("results/", basename, ".png", sep=""), width=dpi*9, height=dpi*14, res=dpi)
grid.arrange(arrangeGrob(bounds_table, req_table, ncol=1), arrangeGrob(histo, scatter, ncol=1), main=paste(basename, options[1], sep=" "), ncol=1, heights=c(1, 4))
dev.off()

response_times.R

library (ggplot2)
library (gridExtra)

# arg1 - online csv
# arg2 - offline csv
# arg3 - filename postfix

options <- commandArgs(trailingOnly = TRUE)

basename = options[3]

csv_online = read.csv(options[1])
csv_online$timeStamp = (csv_online$timeStamp - min(csv_online$timeStamp)) / 1000

csv_offline = read.csv(options[2])
csv_offline$timeStamp = (csv_offline$timeStamp - min(csv_offline$timeStamp)) / 1000

histo_max = max(c(max(csv_online$Latency), max(csv_offline$Latency)))

histo_online <- ggplot(csv_online, aes(x=Latency)) + geom_histogram(binwidth=100, origin=0, fill="green") +
      scale_y_sqrt() +
      scale_x_continuous(limits=c(100,histo_max)) +
      labs(x="response time (ms, excludes first 100ms bin)", y="number of responses", title="Origin Online Response Time Distribution (100ms bins)")

histo_offline <- ggplot(csv_offline, aes(x=Latency)) + geom_histogram(binwidth=100, origin=0, fill="red") +
      scale_y_sqrt() +
      scale_x_continuous(limits=c(100,histo_max)) +
      labs(x="response time (ms, excludes first 100ms bin)", y="number of responses", title="Origin Offline Response Time Distribution (100ms bins)")

scatter <- ggplot(csv_online, aes(timeStamp, Latency)) + geom_point(data=csv_online, size=2, colour="green", alpha=1/4, label="Online") +
      geom_point(data=csv_offline, size=2, colour="red", alpha=1/4, label="Offline") + 
      labs(x="sample point (s)", y="response time (ms)", title="Scatter")

pdf(paste("results/", basename, ".pdf", sep=""), paper="a4")
grid.arrange(histo_online, histo_offline, scatter, main=basename, ncol=1)
dev.off()

histo.R generates this;