TracNav menu





Experiment Management Interface (EMI) Manual for EMI 0.6.x

Contents

Accessing EMI

Currently EMI is deployed as a separate portlet within the Gridsphere portal. In order to access the portlet it is required to have a valid user account on the portal. For Virolab the portal is hosted at https://portal.virolab.org and its administrators should be contacted in order to obtain information on getting an account. After successfully logging into the portal a separate tab called "EMI" holds the EMI interface (see Fig. 1). Notice, that if you are using EMI in the standalone mode the layout may be different. Further detailed description of how the EMI interface works can be found in the sections below.

Fig. 1 Experiment Management Interface (EMI) - general look

EMI widgets explained

The EMI interface consists of three main visual components called widgets, each of which is described in the subsections below. They can be viewed by clicking a corresponding tab which can be seen in Fig. 1.

Repository Widget

The Repository Widget allows for experiment browsing by contacting external experiment repositories and downloading a list of available experiments with detailed information (experiment versions, licenses and user feedback) and presenting the list to the user. An example of an experiment item can be seen in Fig. 2. An experiment item can be expanded to reveal experiment details by clicking the experiment's name. If there are many experiments the list is divided into pages which can be managed by a pagination menu available in the upper part of the repository widget.

Fig. 2 Experiment details displayed in the Repository Widget

As seen in Fig. 2 each experiment has one or more versions (e.g. 0.4 or 0.9 ). The versions can be changed by clicking on their corresponding tabs. Every version has a separate details window in which information such as Author email, Organization or Publication date are displayed. Additionally, for each version a collection of action buttons is visible (bottom left part of the screenshot) which consists of the following items (starting from left):

  • Execute experiment - executes the currently viewed experiment version,
  • View main script - shows the main script of the experiment,
  • View license - shows the license text of this versions of the experiment,
  • Feedback - displays the experiment feedback editor.

Execution Widget

The Execution Widget is responsible for presenting the actual state of experiment execution and retrieving user data. An example of this widget is presented in Fig. 3.

Fig. 3 EMI Execution Widget

The list shows execution items, each in a respective state. The execution states can be one of the following:

  • initialized - the execution has been accepted by the execution engine,
  • input required - in this state the experiment requires the user to input some data (it can be completing a web form or providing a file),
  • running - in this state the experiment is being processed in the remote execution engine,
  • finished - when experiments finish their execution they are put into this group,
  • finished with error - if during the execution a critical error occurs this state is used.

Each of the execution items can be expanded to present execution log or data input widget in case user input is required. Below is an example of an expanded execution item, which is in the input required state.

Fig. 4 User input widget.

After providing the necessary data and clicking the Send data button the execution will continue and the state will be changed to running. For convenience execution items can be filtered by check boxes situated in the upper part of the widget. When inputing data the result reuse mechanism may be used to input previously generated result. This can be accomplished by clicking the Result reuse... button.

Results Widget

The third widget (see Fig. 5) is used to display the experiment results.

Fig. 5 EMI Results Widget

Results are grouped into two categories. The most recent results, right after an experiment finishes the execution, are put into the Current results group and are marked with a reddish color. When viewed they become gray and can be moved to Trash. Additionally, it is possible to empty the trash by using the Empty Trash button. Results removed from the trash cannot be restored. In the upper part of the widget a filter section can be seen. It can be used to filter results by their semantic type or content.

Experiment execution

Running an experiment

To execute a particular version of an experiment, experiment details need to be displayed in the repository widget by expanding the experiment's description and by clicking a version tab inside. The execution button is the first button in the collection of action buttons situated in the lower left part of the experiment details panel. After clicking it the running experiments box in the execution widget will be updated. Before running any experiment its license has to be read and the I have read the license check box needs to be checked. Then the licence needs to be accepted and if required an execution comment can be provided in the next window. The same window lets users to specify the log level of the experiment.

Providing user input

When user input is required during an experiment execution its status changes to input required and after expanding the corresponding execution item in the Execution Widget a data form is presented. After filling in the requested data they can be sent by clicking the Send data button. When the data reach the remote engine the experiment item changes its status back to running.

Analyzing experiment output

When the experiment execution finishes the output is presented to the user, depending on the experiment execution, as a result item in the result widget or as an experiment log in the execution item of a given experiment. In the case of a result being displayed in the Result Widget the payload of the result can be viewed by clicking the View result button. An experiment may be aborted at any time by clicking the Abort button inside the execution item.

Feedback mechanism

In order to provide feedback to the experiment developer, the feedback editor executed by clicking the Feedback button in the Actions menu of every experiment version box in the repository widget should be used. This requires however the user to have write permissions to the experiment repository. After filling in the tickets or comments about a particular version of an experiment the repository feedback file is updated accordingly. Also, the feedback window is updated.

Fig. 6 EMI experiment ticket editor.

In Fig. 6 the ticket editor is depicted. In order to submit a new ticket the New ticket button should be clicked after which the fields become editable. After inputing the ticket's contents and selecting its type the Save ticket needs to be clicked. The list of tickets will update automatically. To change the status of a ticket it should be selected from the list of tickets and the form will update. After changing the status of a ticket it should be confirmed with the button below.

Fig. 7 EMI experiment comments editor.

In Fig. 7 the comments editor is presented. It is used to submit user's free thoughts about an experiment. In order to submit a comment author's name and a comment should be provided in the New comment tab and Add comment button should be clicked. The list of comments below the New comment tab will be updated automatically.