Azure Stream Analytics: Real-Time Data Processing

Azure Stream Analytics processes millions of events per second with SQL-like syntax. No coding required for common streaming patterns.

Basic Query

SELECT
    DeviceId,
    AVG(Temperature) AS AvgTemp,
    MAX(Temperature) AS MaxTemp,
    System.Timestamp() AS WindowEnd
FROM IoTInput TIMESTAMP BY EventTime
GROUP BY
    DeviceId,
    TumblingWindow(minute, 5)

Window Types

Tumbling Window

Fixed, non-overlapping intervals:

TumblingWindow(minute, 5)
-- |---5min---|---5min---|---5min---|

Hopping Window

Fixed intervals that overlap:

HoppingWindow(minute, 10, 5)
-- 10-minute windows every 5 minutes

Sliding Window

Window slides with each event:

SlidingWindow(minute, 5)
-- Triggers on every event, includes last 5 minutes

Session Window

Groups events with gaps:

SessionWindow(minute, 5, 30)
-- 5-minute timeout, 30-minute max duration

Anomaly Detection

SELECT
    DeviceId,
    Temperature,
    AnomalyDetection_SpikeAndDip(Temperature, 95, 120, 'spikesanddips') AS SpikeAndDip
FROM IoTInput

Reference Data Join

SELECT
    i.DeviceId,
    d.DeviceName,
    d.Location,
    i.Temperature
FROM IoTInput i TIMESTAMP BY EventTime
JOIN DeviceReference d ON i.DeviceId = d.DeviceId

Output to Multiple Sinks

-- To Power BI for dashboards
SELECT * INTO PowerBIOutput FROM AggregatedData

-- To Blob for archival
SELECT * INTO BlobOutput FROM RawData

-- To Event Hub for downstream processing
SELECT * INTO EventHubOutput WHERE IsAnomaly = 1

Stream Analytics bridges the gap between raw event streams and actionable insights.