Loading Records in EventFlow

"How many records can I load in Eventflow?"

Read the following report on Ten Strategies for Data Focusing with Temporal Event Sequences

Memory Usage

"What to do when the dataset is large and does not load into EventFlow?"

The memory use limit depends on the configuration of the Java virtual machine. The Java virtual machine limits the Heap memory it can use. Currently the default threshold is set around 1 GB for a typical computer with 4GB of RAM. The memory use is correlated with the total number of events in the dataset and rises linearly. Some conditions in the dataset affects the memory use more than others. We have observed that the number of events per record increases the memory use more than the number of records. With a standard configuration we have been able to load a dataset with 100000 events.

Steps you can take to improve scalability with repect to memory usage are the following:

  1. Increase the memory limits on the Java Virtual Machine: 
    java -Xmx3000m  -jar /path/to/EventFlow.jar

    - Here 3000 is the maximum heap memory size, configured to 3000MB.

  2. Consider trimming the dataset: Remove event categories you are less interested in or start without including event category or record attributes.

Known Problems and Troubleshooting

Listed below are issues that have been found, but are still in the process of being fixed. However, we have provided temporary solutions for several of them.

If you have found any issues with EventFlow that are not listed below, please email us at eventflow.umd@gmail.com.

  1. Security Settings: Mac

    2. Description:
    Launching EventFlow on Mac for the first time may cause an alert of the following:

    Solution:
    1. Open your System Preference

    2. Click Security & Privacy

    3. Unlock your settings

    4. Under General tab, click Open Anyway

  2. Sometimes the Timeline panel stop rendering for no reason

    3. Description:
    The Timeline panel is blank when the dataset is not empty. We found it a few times but could not reproduce the bug. It also rarely occurs. Please let us know if you know how to reproduce the bug.

    Temporary solution:
    Double-click on the scrollbar of the timeline panel will force the Timeline panel to redraw.

  3. Large dataset issues

    4. Description:
    The system finds it hard to handle exceedingly large and complex data set. Please refer to Section 1.1 of the Scalability Report on this issue.

    Temporary solution:
    Currently, the system can warn the users when the dataset exceed the capability of the system and prompt them to filter the dataset by hiding some event categories. The progress bar can also keep users informed of the data loading and tree building process.
    Besides, if the visualization is too complex to do analysis, users can resort to features like the Choose 2 methods, Category Aggregation, Interval Merging, Windowing, and the Simple Search methods.
    Note that since the memory management in java is not sufficiently controlled programmically, the memory warning may act in a unreasonable way. At this time, you should try to extend the JVM memory as instructed in Section 1.1 of the Scalability Report, or try to reperform the operation again after the wasted memory gets freed by JVM.

  4. Exporting from Excel creates invalid format

    5. Description:
    EventFlow gives an invalid format warning when loading a file that was exported from Excel.

    Solution:
    This often happens when using the "Unicode Text" option to save as .txt instead of "Tab delimited file." To correct this, be sure to choose the file format "Tab Delimited Text (.txt)" when saving in Excel.

  5. The overall event distribution is not supported

    6. Description:
    As you may know, users can choose to display the overall distribution of events in bar chart, which turns out to be very useful when doing analysis in a wholistic scale and comparing event distributions. Since EventFlow is designed to tackle event sequence aggregation problem, rather than event aggregation, it currently has no feature for overall event distribution.

    Temporary solution:
    EventFlow supports distribution of adjacent events. Users can also view the distribution of non-adjacent events using the Distribution Options tool. If you find the above feature can't solve your problem, we temporarily refer you to our LifeLines2 System.

  6. The system runs out of memory when loading multiple datasets

    7. Description:
    If a dataset is already loaded, and the user tries to open another file, EventFlow may run out of memory because Java's garbage collector has not cleared sufficient memory.

    Temporary solution:
    Close EventFlow completely and open the new dataset.

  7. Some features in LifeLines2 are not available in EventFlow

    8. Description:
    If a user is familiar with our older version, LifeLines2, he/she may notice that some features are not available in EventFlow (e.g. summarization feature).


    For more information, please see our OLD Scalability Report.


    << Previous Next >>

    Web Accessibility