dstat.csv

First thing I did was parse the time:

df$time <- strptime(df$time, "%m/%d/%Y %H:%M")

It's a bit confusing that all of these have the same timestamp. For that reason, I numbered them:

df$rowid <- 1:nrow(df)

Then you can plot the series of sys, for example with:

ggplot(df, aes(rowid, sys)) + geom_area(fill='red')

Theoretically, you can stack them:

ggplot(df, aes(rowid, usr)) + geom_area(fill='green') + geom_area(fill='red', aes(sys)) + geom_area(fill='blue', aes(idl))

But for some reason, when I add wai, it goes a little weird

They don't all add up to 100, which is a little weird. I'm a little dubious of percentages in general:

> summary(df$sys + df$idl + df$wai + df$usr)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
  88.13   93.03   99.94   96.91  100.00  100.00 

Oh, obviously because I did the wrong aes:

ggplot(df, aes(rowid, usr)) + geom_area(fill='green') +
    geom_area(fill='red', aes(y=sys)) +
    geom_area(fill='blue', aes(y=idl)) +
    geom_area(fill='purple', aes(y=wai))

ggplot generally wants "long format" data, so it's far easier if you rotate it. I used reshape2's melt for this. Try this:

df.melted <- melt(df, id.vars=c('rowid'), measure.vars=c('usr', 'sys', 'idl', 'wai'))

ggplot(df.melted, aes(rowid, value, fill=variable)) +
    geom_area() + theme_bw() + scale_x_continuous(labels=formatSeconds) +
    labs(x='', y='Percentage')

I really like jitter plots for looking getting a feel of where the bulk of measurements are irrespective of time.

e.g. if you looked at scatter plots of all the variables:

ggplot(df.melted, aes(rowid, value)) + geom_point(size=1) + facet_wrap(~variable)

you can see a lot of patterns and get some feel for where things are, but a jitter plot sort of rotates this whole thing so that you're looking down the X axis and time no longer matters:

ggplot(df.melted, aes(variable, value)) + geom_jitter(size=1)

the X is fuzzy, but grouped in columns atop the variable. That is, it has no actual meaning within those bounds, but the Y does. You can quickly see where the bulk of the values lie on the Y axis by variable.

Starting with the data above minus the first six lines:

library(ggplot2)
library(humanFormat)

df <- read.csv("~/Desktop/dstat.csv")
df$rowid <- 1:nrow(df)

df$time <- strptime(df$time, "%m/%d/%Y %H:%M")

df.melted <- melt(df, id.vars=c('rowid'), measure.vars=c('usr', 'sys', 'idl', 'wai'))

ggplot(df.melted, aes(variable, value)) + geom_jitter(size=1)