Möbius Data Manager
Möbius provides the ability to store solver results in a PostgreSQL database. Instructions for setting up a database to work with Möbius are detailed in the Installation page. Once the results database has been successfully set up, the user can store any results from simulation and analytical solver executions in it.
Several third-party utilities exist for accessing and managing general data in PostgreSQL databases. PGAdminIII (http://www.pgadmin.org) is a very popular graphical administration tool, and PostgreSQL also distributes a simple text-based command-line tool. However, these tools work on databases without any notion of how or why the data is organized the way it is.
The Möbius Data Manager tool seeks to fill the role of a graphical interface on Möbius results databases in a way that is intuitive and familiar for Möbius users. The Möbius Data Manager allows users to connect to Möbius results databases and manage, organize, and export the data they contain. Möbius users can launch the Möbius Data Manager by clicking on the UtilitiesMobius Data Manager menu item in the Project Manager window.
Connecting to the Database
When you first launch the Möbius Data Manager, you will immediately be prompted to enter the connection details for the database you’d like to work with. These are the same details you entered in the Database Settings dialog window.
The Host field should have either the fully qualified domain name or the IP address of the computer running the PostgreSQL database. The Username and Password fields should contain the PostgreSQL user credentials. The Database field should contain the name of the Möbius results database on the PostgreSQL server.
Once you have entered valid details, click Login to connect to the server and begin the Data Manager session.
To log in to a new connection, you may start the new connection by clicking on the FileNew Connection menu item in the Möbius Data Manager.
Saved Connection Details
Some connection information can be saved between sessions so that the connection details don’t need to be typed in each time. By default, the Data Manager always stores the details from the most recent connection. The next time the Data Manager is launched, the details from the last session will be automatically filled into the login dialog.
For the sake of security, the Möbius Data Manager never stores passwords. You must always enter your password when using stored connections.
Connection details can also be manually saved with a user-specified connection name. In the Login window, highlight the text in the Connections field and replace it with a name, for example, “Work DB Server.” Fill in the remaining details and then click the Save Connection button. The next time you are prompted to login, the new saved connection details will be in the drop-down list. Selecting it in the list should populate the fields with the saved information.
Should a set of saved connection details no longer be needed, you can easily delete it. Begin by selecting the connection from the drop-down list and click the Delete Connection button to permanently remove it from the list. This does not affect the database server or any data on it.
Once connected to a results database, the Data Manager presents the data as a tree in the upper-left panel of the window. The topmost node represents the database server. The next level represents the projects that have data stored in this database. This level branches based on unique Möbius project names, so two separate projects that are named identically will show up as just one node in the tree. In a situation like that, the results will likely branch apart at lower levels in the tree, but it is typically a good idea to give your Möbius projects unique names.
The third level in the tree represents the solvers in the project. Each node is a conjunction of the solver name and the run name (for analytical solvers, this is the output file name), which are both defined in their respective solver windows in Möbius. The next level represents individual solver executions, which are labeled with the date and time when the run completed.
The fifth level always consists of two nodes: reward variables and experiments. They are simply two ways of grouping the same data. They allow the user to view only data for a specific reward variable or data for a specific experiment. Expanding the reward variables node provides a list of reward variables defined in that solver execution. Expanding the experiments node provides a list of experiments defined in that solver execution.
At the individual solver execution time stamps (fourth level) and all nodes below them, you can view the data associated with the nodes by double left-clicking on them or right-clicking and selecting Open Table from the context menu. Depending on the node, certain subsets of the data for the solver execution are presented. For example, if you open the Experiment 1 node, all the data from Experiment 1, across all reward variables, is presented. If you open the unreliability reward variable, all data for that reward variable, across all experiments, is presented. The data is displayed in a table in the right pane.
The columns in the data tables show all the stored attributes for the associated results. Some data cannot be easily displayed with a single value, such as Mean. If you double-click on the value, a dialog appears that states the value, the lower bound, the upper bound, and the bound type for that value. This data can also be exported as described in Section 1.5. Global variable parameters, which are defined in Möbius studies, are also displayed as columns on the far right end of the table.
Every time any solver is run in Möbius while the database feature is enabled, results are added to the database. When refining models and experimenting with different parameters, this can lead to collecting a multitude of unnecessary results in the database. The Data Manager allows users to easily delete as much data as necessary. This does delete data from the database itself, so care should be used when deciding which data can be removed.
To delete data, right-click on any node in the tree (aside from the root node) and select Delete. This will permanently remove from the database the data associated with the node and all of its children.
Making the Most of the Data Manager Interface
The Möbius Data Manager graphical user interface has been designed from the beginning to be extremely flexible. Any of the panels that have tabs in the upper left corner can be moved, resized, maximized, minimized, stacked, or popped out. This is particularly useful during comparison of results from several runs.
By left-clicking and dragging the tab from one of the table panels, you can move the panel. If you hover over the right side of another panel or the entire window while dragging the tab, a docking arrow and the silhouette of the panel in its prospective position will appear. Table panels can be docked along any edge of any other panel or the main window. There is no limit to the various configurations a user can build by dragging and docking the panels. It is also possible to stack panels by dragging the tabs next to other tabs in a different section. When you drag a tab, a silhouette and a different cursor indicates the prospective position of the tab until you release the mouse button.
You can also pop the data panel out into a new window by either dragging the tab outside of the Data Manager window or right-clicking on the tab and selecting Detatched from the context menu. Right-clicking on the tab again and unchecking Detatched or dragging the tab back into the Data Manager window will pop it back in.
If a particular data panel requires more space, double-clicking on its tab will maximize it in the Data Manager window and minimize all other panels. Double-clicking the tab again will restore all the panels to their previous state. Maximizing, minimizing, and restoring a panel can also be done by right-clicking on the tab.
Finally, you can close data panels by clicking on the X on the right side of the tab or right-clicking on the tab and selecting Close.
The Möbius Data Manager allows access to the Möbius results database using an intuitive, tree-based structure. However, viewing data like this requires users to live with the structure defined by the database. To overcome that, the Möbius Data Manager provides a notion of data sets. In the lower left pane of the Data Manager application is the data sets panel. Data sets provide a flexible place to locally gather data from one or more results databases and organize and export the data.
Adding and Removing Data
Begin by clicking on the New button of the Data Sets panel. A new, empty data set is now listed. Next, drag and drop nodes from the tree above onto the data set below. This copies the data from the database and stores it locally in the data set. To view the contents of the data set, double-click on it, and a new data panel will appear.
In the data panel, all the data of the data set is displayed. Heterogenous data is merged as effectively as possible. You can delete rows by highlighting them, right-clicking, and selecting Delete Row from the context menu.
All the currently open data sets are listed in the Data Sets panel. Data sets are saved locally as .mds files. To save a data set, select it from the list and click the Save button. This also allows you to rename the data set, because the name that appears in the list is the same as the filename without the .mds extension.
When you no longer want to work with that data set, you close it by selecting it from the list and clicking the Close button. Should you want to work with it again, click the Open button and locate the file using the open dialog window. All data in the data set is stored locally in the file and can be transferred between machines and users through email or some other method. No database connection is necessary in order to work with the data in a data set, which means that data sets provide an excellent way to work on data in an offline setting.
At the individual execution nodes and lower levels, you can export data associated with the nodes. Right-clicking on one of these nodes and selecting Export begins the process. The first window that appears provides a list of available columns for the data. You should check any columns you would like to include in your export and click the Export button.
A file-chooser dialog then appears, and you must provide a file name and path for the export file. The export file is formatted as a comma-separated values file, so it is appropriate to end the filename with the .csv extension, but it is not required. Once exported, the file can be easily used by many applications, such as Microsoft’s Excel and OpenOffice.org’s Calc.
You can also export data from a data set. To begin the process, select the open data set that you want to export and click on the Export button.