Designing User Interfaces Spring 1999

1
SE 767-NT Software Performance Engineering Robert
Oshana Lecture 17 For more
information, please contact NTU Tape
Orders NTU Media Services [phone]
oshana_at_airmail.net
tapeorders_at_ntu.edu
2
Performance measurement concepts

• Performance measurements may be static or dynamic
• May need to make measurements indirectly
• Approach depends on the problem

3
Terminology

• States
• Events
• We monitor states
• Software monitors
• Hardware monitors
• We record events
• Program event recording
• System event recording

4
Terminology

• Internal versus external
• Granularity
• Record all
• Summarize
• Tally data between events
• Must then run programs to analyze the data and
  report results

5
Factors that may affect measurement

• System perturbation
• Capture ratios
• System overhead
• Measurement timing
• Reproducible results
• Representative time periods
• Averages for typical behavior
• Workload generation

6
Data collection techniques and tools

• Many computer systems offer data collection tools
  with their systems
• Commercial products also available
• Computer measurement group (CMG)

7
Data collection techniques

• System monitors
• Program monitors
• System event recorders
• External program event recorders
• Internal event recorders

8
UNIX sar

• sar - system activity reporter
• In the first instance, sar samples
  cumulative activity counters in the operating
  system at n intervals of t seconds, where t
  should be 5 or greater. If t is specified with
  more than one option, all headers are printed
  together and the output may be difficult to read.
  (If the sampling interval is less than 5,
  the activity of sar itself may affect the
  sample.) If the -o option is specified, it saves
  the samples in filename in binary format. The
  default value of n is 1.

9
UNIX sag

• sag - system activity graph
• The sag utility graphically displays the
  system activity data stored in a binary data
  file by a previous sar run. Any of the sar data
  items may be plotted singly or in combination,
  as cross plots or versus time. Simple
  arithmetic combinations of data may be specified.

10
Measuring SPE data

• Workload data
• Data characteristics
• Path characteristics
• Software resources
• Processing overhead
• Computer resource usage

11
SE 767-NT Software Performance Engineering Robert
Oshana End of lecture For
more information, please contact NTU Tape
Orders NTU Media Services [phone]
oshana_at_airmail.net
tapeorders_at_ntu.edu