JWAT allows workload analysis using clustering techniques. Input data can be collected from any kind of textual file: support for standard apache log files is provided but users can specify customized formats. During input phase, users can provide the following extraction criteria:
- All: selects every observation
- Interval: selects only observations in a given range
- Random: given number of observations, selects that number at random from data
- N every K: picks at random N observations every K observations in input file (K>N)
The application support a complete environment for statistical analysis. In that environment is possible to calculate univariate and bivariate statistic and allow the drawing of frequency, quantiles and scatter plots.
Data can be normalized with the following transformations:
- Standard deviation
and trimmed to selected percentile.
Clustering is performed with the following algorithms:
JWAT provides also an interface to the similarity clustering tool CLUTO. This is important for workloads including qualitative data.
Data analysis is guided with a wizard interface.