For pre-sorted maintenance request data by DateFiled, which Group By mode would likely provide better performance?

• A. Process at End
• B. Process When Group Changes
• C. Process When Group Changes (Advanced)
• D. Pre-sort data before grouping

Answer: (B)

When grouping by a sorted key, you want to act as soon as you know a group has finished, without waiting to collect everything. If DateFiled is already in sorted order, all records for the same date come together, so you can accumulate the current date’s results and emit them as soon as the date changes. This is exactly what the group-change processing mode does: it keeps the current group’s aggregates while features with that date stream in and flushes them when the group boundary is detected. That avoids buffering an entire date’s worth of records and reduces memory usage and latency, delivering better performance.

Process at End would require holding all features for a date until the end of that date’s block, which can use more memory and time. The Advanced variant is a more specialized option for certain boundary rules and isn’t necessary here. Pre-sorting the data is a preparation step, but given the data is already pre-sorted, using the group-change mode leverages that order for efficient streaming processing.