You are monitoring an Azure Stream Analytics job that processes data from an IoT hub. The job's output to Azure Synapse Analytics is experiencing high latency. The job's SU% utilization is at 90%. Which action will most likely reduce the latency?
More SUs provide more processing capacity, reducing latency.
Why this answer
The job's SU% utilization is at 90%, indicating that the current Streaming Units (SUs) are nearly saturated, causing a processing bottleneck. Increasing the number of SUs allocates more compute resources (CPU and memory) to the Stream Analytics job, allowing it to process incoming IoT data faster and reduce the latency to Azure Synapse Analytics. This directly addresses the high utilization issue, which is the most likely root cause of the latency.
Exam trap
The trap here is that candidates often confuse output-side tuning (like partitioning or sink configuration) with the actual processing bottleneck, overlooking that high SU% utilization directly indicates the Stream Analytics job itself is the limiting factor.
How to eliminate wrong answers
Option B is wrong because decreasing the watermark delay interval would make the job emit results more frequently, but it does not increase processing capacity; with SU utilization already at 90%, this could worsen backpressure and increase latency. Option C is wrong because increasing the late arrival tolerance window only allows the job to handle out-of-order events for a longer period; it does not improve throughput or reduce latency caused by resource saturation. Option D is wrong because increasing the number of partitions in the output table improves write parallelism on the Synapse side, but the bottleneck is the Stream Analytics job's processing capacity (90% SU utilization), not the output sink's partitioning.