fio2gnuplot - Render fio's output files with gnuplot
fio2gnuplot [-ghbiodvk] [-t title] [-o outputfile]
[-d output_dir] [-p pattern]
[-G type] [-m min_time] [-M max_time]
fio2gnuplot analyze a set of fio's log files to turn them
into a set of graphical traces using gnuplot tool. Several flavor of
plotting are produced
- Individual 2D
Graph
- Each file is plotted in a separate image file with several option
- raw : Plot the exact reported performance. This plotting could be
difficult to read
- smooth :a smoother version of the raw print Using csplines option of
gnuplot, the rendering is filtered to get an easier to read graph.
- trend : an even smoother version of the raw print to get trends Bezier's
curves makes much more filtered plots The resulting graph helps at
understanding trends.
- Grouped 2D
graph
- All files are plotted in a single image to ease the comparison. The same
rendering options as per the individual 2D graph are used :
- Grouped 3D
graph
- All files are plotted into a single 3D graph. The 3D plotting generates a
'surface' to estimate how close were the performance. A flat surface means
a good coherency between traces. A rugged surface means a lack of
coherency between traces
- Mathemical
Plotting
- Average
graph
- A bar graph to show the average performance of each file. A green line is
added to show the global average performance. This green line helps at
understanding how far from the average is every individual file.
- Min graph
- A green line is added to show the global average of minimal performance.
This green line helps at understanding how far from the average is every
individual file.
- Max graph
- A bar graph to show the maximum performance of each file. A green line is
added to show the global average of maximal performance. This green line
helps at understanding how far from the average is every individual
file.
- Standard
Deviation
- A bar graph to show the standard deviation of each file. A green line is
added to show the global average of standard deviation. This green line
helps at understanding how far from the average is every individual
file.
Erwan Velu <erwan@enovance.com>