# Azure Stream Analytics Job

> Stream Analytics jobs bill per streaming unit per hour whenever the job is running, regardless of how many events arrive. ZopNight discovers each job's SU allocation and state via Resource Graph, compares 60 days of input-volume metrics against that allocation, and surfaces jobs metering full SU rates on quiet sources.

Source: https://zop.dev/integrations/azure/streamanalytics
Updated: 2026-08-19

---

Stream Analytics jobs process real-time event streams, billed per streaming unit per hour while the job runs. Jobs left running against quiet event sources bill full SU rates for nothing.

## Streaming units meter on time, not throughput

A running job charges for its allocated streaming units by the hour, and the meter is indifferent to traffic: a job allocated several SUs processes a trickle of events at exactly the same hourly cost as a flood. That decoupling is the whole cost story for this type. The spend question is never "how many events did we process" but "how many SU-hours did we hold, and were they warranted". Those two numbers drift apart the moment an event source goes quiet.

## Matching SU allocation to event volume

Discovered via Azure Resource Graph with SU allocation and job state. Azure Monitor metrics (60-day lookback) show input volume, Cost Management billing attributes spend, and idle jobs surface in review. Putting allocation next to two months of input volume makes the mismatches obvious: jobs whose sources went dark when an upstream device fleet was retired, or allocations padded for a launch spike that settled long ago.

## Why this type is visibility-only

ZopNight treats Stream Analytics as discovery and cost visibility only. A stopped job halts its outputs and, on restart, resumes from where its offsets allow. Whether that loses, duplicates, or delays events depends entirely on the pipeline's design. That judgment belongs to the pipeline's owners, so the platform surfaces the idle-job evidence rather than automating a stop.

## Stream Analytics jobs that overspend

The common leaks: jobs consuming from decommissioned IoT hubs or event hubs, running indefinitely on zero input; test and demo jobs started during development and never stopped; and SU counts sized for worst-case burst scenarios that the 60-day input history shows never arriving.

## Job state and SUs in the Azure portal

Azure portal → Stream Analytics jobs lists every job with its Status (Running, Stopped, or Failed), and opening a job shows its streaming unit allocation under Scale, the two facts a cost review needs side by side.