Data Grid Rendering Thread vs Large Result Set Option: Which Approach Best Balances Memory Usage and Interactive Features?

The Trade-off: Daemon Threads vs. Streaming

To balance memory usage and interactive features, the Large Result Set option is the superior choice for stability, while the daemon rendering thread is a mitigation for memory leaks in the standard interactive grid. There is currently no documented method to stream large result sets while retaining full client-side interactivity (sorting and inline editing), as these features require the dataset to be resident in the JVM heap.

Analysis of Rendering Approaches

The choice between these two mechanisms depends on whether the priority is data manipulation or system stability:

• Daemon Rendering Thread: This approach allows the JVM to reclaim resources more effectively and prevents the rendering thread from blocking JVM shutdown. It reduces the impact of memory leaks but does not eliminate the high heap requirement of the Data Grid.
• Large Result Set Option: This implements a streaming mechanism. By bypassing the full load of the result set into memory, it eliminates the leak risk and drastically lowers the memory footprint, but it strips away interactive capabilities because the client does not possess the full dataset to sort or edit.

Addressing the Technical Constraints

Regarding the specific concerns of daemon behavior and automation:

• Cleanup Risks: Setting the rendering thread to daemon by default is unlikely to interfere with critical database cleanup, as the rendering thread handles the display of data, not the transactional state of the session. However, any local client-side cleanup tied specifically to the thread's lifecycle could be terminated prematurely.
• Automatic Thresholds: Automatically enabling the Large Result Set option based on a row threshold would prevent crashes but would create an inconsistent user experience where interactivity disappears unpredictably. Manual selection remains the safest way to ensure the user is aware of the loss of sorting/editing.

Verification Steps

To determine which approach is necessary for your specific environment, perform the following checks:

1. Memory Delta: Use JVisualVM or JConsole to monitor the JVM heap. Compare the memory spike when loading 100k rows in standard mode versus the Large Result Set mode.
2. Interactivity Check: Execute a query with >100k rows using the Large Result Set option and attempt to click a column header for sorting; verify that the feature is disabled.
3. Version Validation: If using a version prior to 21.2, verify if closing a result tab continues to hold memory, indicating the need for the daemon thread fix.

Diagnostic Detail Needed: To refine this recommendation, please provide the current -Xmx (Maximum Heap Size) setting in your sqldeveloper.conf file, as the threshold for "leaking" is relative to available memory.