4.2 GenSyn DataReporter
- Post Processing of Data
The GenSyn DataReporter is a graphical user
interface (GUI) implemented in java for post-
processing of packet traces. The motivation was
to simplify and automate the post-processing of
packet traces collected during network tests
using the GenSyn traffic generator4). The post-
processing includes management of input files,
input parameters to scripts and output files gen-
erated from scripts. Thus, it was seen that a sup-
port system was needed. As shown in Figure 25,
a central host post-processes the packet traces
captured by the DAG monitors or tcpdump. Data
reduction on each of the monitoring machines is
being considered [EmVi01]. The post-process-
ing involves huge amounts of measurement data
and is therefore very computation intensive. The
actual processing of measurement data is imple-
mented by using perl, awk and gnuplot scripts.
Figure 24 illustrates the software developed for
post-processing of measurement data.
Note that the computation for large trace files is
rather time consuming. Therefore, the computa-
tion is performed only once for each combina-
tion of selected options whenever possible by
keeping intermediate files.
4.2.1 Centralized Post-processing
The storage unit is a central location where all
measurement data collected for a GenSyn test
scenario is stored. The measurement data col-
lected for the test scenario includes packet traces
(tcpdump or DAG) containing information about
packets sent to and from each of the GenSyn
machines in the test network (e.g. TG1 – TG10).
The GenSyn DataReporter requires a directory
structure as illustrated in Figure 26. That is,
before the DataReporter can be run the following
steps must be performed:
- Create the directory structure;
- Move rawdata files (DAG or tcpdump traces)
to the correct directory; - Create the GenSyn_machines file with map-
ping from logical name (and directory) to IP
address.
However, a directory structure with several lev-
els is recommended to keep track of various test
scenarios. Ideally, the scripts that collect and
move measurement data should create the neces-
sary directories and files automatically.
4.2.2 Performance Parameters
Computed by DataReporter
The statistics derived from the collected packet
traces are classified as point-to-point and multi-
point, respectively. Point-to-point statistics are
computed by correlating information in two
packet traces. Examples of such statistics are
average unidirectional delay, unidirectional
packet loss ratio and average throughput com-
puted over a bin of a specified duration. Multi-
point statistics are defined from observations at a
single measurement point (a single packet trace),
e.g. number of bytes sent and received to/from
various remote machines. Note that the point-
to-point statistics provide detailed information
about unidirectional performance, while multi-
point statistics are less computation intensive
but can provide an overview of a given scenario
in less time. It may be noted that adding other
statistics and graphs is only a matter of changing
the scripts and GUI.4.2.2.1 Multi-point Performance
The performance as observed at a single mea-
surement point is reported. These statistics are
computed:- Average pps5)and kbps6)sent to various des-
tinations from a selected traffic generator; - Average pps and kbps received from various
sources by a selected traffic generator.
These statistics are computed over bins of a
given duration as a function of time. In addition,
various statistics are computed over the entire
measurement period.Note that the multi-point performance statistics
only provide an overview as observed at a single
measurement point and are not suitable for eval-
uate end-to-end performance. However, the5)pps denotes packets per second.
6)bps denotes bits pr second.
Figure 26 Directories and files
Directory required by GenSyn DataReporterTG1 TG10 GenSyn_machinesTG1
10, 16, 78, 34
...TG10
- 32
Rawdata GraphsText- • •