table of Contents
1. Stream computing is what?
Flow is calculated with respect to a batch of it, we've learned that Mapreduce batch, it belongs to the off-line calculation, calculation of data are past a certain point in time, and we have developed a software management system, also queried past some point of data entry. Then lost to calculate it, it is time to enter the entry began to calculate, and calculation speed is also very fast, up to milliseconds after completion will be able to calculate real-time feedback, or stored. This calculation is for general traffic, ah, the electricity supplier, ah, ah the weather and other requirements of real-time push scene.
2. What are the calculated flow
Flow calculation includes Storm, Spark Streaming, Samza, Flink.
| name | the company | Applicable scene | Types of |
|---|---|---|---|
| Storm | Streaming | Flow Calculation | |
| Spark Streaming | Apache | Suitable for off-line computing and real-time computing needs at the same time. | Hybrid Computing |
| Samza | Spark and the combined stream is processed Kafka | Flow Calculation | |
| Flink | Apache | Fast processing | Hybrid Computing |
| Hadoop (this does not include) | Apache | Off-line calculation |
3. several structural differences calculated flow
Storm architecture
Spark Streaming Architecture
Samza architecture
