Stream Processing

Stream processing has the below characteristics:

• Unbounded data processing: As applying to unbounded data, the stream processing itself is also unbounded. The workload can distribute evenly across time compared to batch processing.
• Low-latency, near real-time: stream processing can process data once it is produced to get the result in a very low latency.

On the edge side, the majority of data are born as continuous streams such as sensor events. With the wide application of IoT, more and more edge computing nodes need to access the cloud network and generate huge amount of data. In order to reduce the communication cost, reduce the data volume of data on the cloud, and at the same time improve the real-time data processing to achieve the purpose of local timely response and also local timely data processing in case of network disconnection, it is necessary to introduce real-time stream processing at the edge.

• When aggregating events to calculate sum, count or average values.
• When searching for a pattern across a series of events.

The state information can be found or managed by:

• Windows