ALOJA is a project to explore Hadoop's
performance under different Software parameters, Hardware, Cloud
or On-Premise, and Job types. This site is under constant development and it is
in the process of being documented.
For more details, visit our welcome page.
For inquiries, feature requests or bug reports please contact us at: firstname.lastname@example.org
Predict Configurations: This tool uses the created models from Modeling Data to expand the space of possibilities for the selected attributes in the right filter box. It gets the best compatible model in cache and predicts all the possible software and hardware configurations for the selected filter, and returns a table with the results. It also creates a tree of attributes, indicating which ones separate better the faster from the slower executions.
To use the tool, select which executions you want to filter through the right filter box (make sure you created a compatible model in Modeling Data section). Then observe the predictions for each combination, also the tree of attributes. Optionally you can re-launch the prediction process using a different model in cache, from the Model information toolbox found at the bottom of the page.
|This tool will generate predictions for ALL of the combinations of configurations/executions|
|1 -||Select from the Filters Box (right box):|
1) The values for each attribute that will be added to the combinations (if no value selected, all will be added to the combination)
2) The model that will be used to predict all the combinations.
|2 -||Click on Generate Predictions, and wait until the data is processed. Take into account that the bigger the data-set selected, the longer can take to process.|
|3 -||Wait until the navigator refreshes, and processes the received data.|
|4 -||Results will appear as:|
a) A table with the executions selected
b) A chart displaying the predicted configurations, marking each prediction and averages, also with the possibility of adding the observed executions that coincide with the filter.
c) A tree with the executions, joined by common attributes. The order of the attributes is greedily selected by the ones that generate the least splits between faster and slower executions.