Table of Contents
Analyzing the simulation result is a lengthy and time consuming process. The result of the simulation is recorded as scalar values, vector values and histograms. The user then applies statistical methods to extract the relevant information and to draw a conclusion. This process may include several steps. Usually you need to filter and transform the data, and chart the result. Automation is very important here. You do not want to repeat the steps of recreating charts every time you rerun simulations.
In OMNeT++ 4.x, the statistical analysis tool is integrated into the Eclipse environment. Your settings (i.e. your recipe for finding results from the raw data) will be recorded in analysis files (.anf) and will become instantly reproducible. This means that all processing and charts are stored as datasets; for example, if simulations need to be rerun due to a model bug or misconfiguration, existing charts need not be recreated all over again. Simply replacing the old result files with the new ones will result in the charts being automatically displayed with the new data.
When creating an analysis, the user first selects the input of the analysis by specifying file names or file name patterns (e.g. "adhoc-*.vec"). Data of interest can be selected into datasets using additional pattern rules. The user can define datasets by adding various processing, filtering and charting steps; all using the GUI. Data in result files are tagged with meta information. Experiment, measurement and replication labels are added to the result files to make the filtering process easy. It is possible to create very sophisticated filtering rules (e.g. all 802.11 retry counts of host[5..10] in experiment X, averaged over replications). In addition, datasets can use other datasets as their input so datasets can build on each other.