You need to monitor the performance of an Azure Data Factory pipeline that copies data from an on-premises SQL Server to Azure Blob Storage. The pipeline runs on a self-hosted integration runtime. Which metric is most important to monitor to ensure the self-hosted IR is not a bottleneck?
Queue depth measures pending requests awaiting a free self-hosted integration runtime node, so a rising value directly signals the IR cannot keep pace with copy activity demand. This exposes the bottleneck constraint, whereas CPU or memory alone may not reflect queued work.
Why this answer
The self-hosted integration runtime (IR) queue depth metric indicates the number of activities waiting to be processed by the IR. A consistently high or increasing queue depth signals that the IR cannot keep up with the workload, making it a bottleneck. Monitoring this metric allows proactive scaling or optimization.
Exam trap
DP-203 often tests the distinction between pipeline-level metrics (duration, data read/written) and IR-specific metrics (queue depth, CPU); candidates may pick duration as a proxy for IR performance.
How to eliminate wrong answers
Option A is wrong because pipeline duration is an end-to-end metric that can be affected by many factors (source, sink, network), not specifically the IR. Option C is wrong because active connections do not directly indicate IR saturation; the IR can handle many connections if not CPU/memory bound. Option D is wrong because data read/written metrics reflect throughput but not whether the IR is the limiting factor; they are outcomes, not bottleneck indicators.