Stream Data Platform - AF uses both in-memory processing and incremental computational processing, which allows it to quickly process large sets of time-sequenced data.
- In-memory processing
- With in-memory processing, data is processed while it is still in memory, thus eliminating unnecessary disk access.
- When processing large data sets, the time required to perform disk I/O can be significant. By processing data while it is still in memory, Stream Data Platform - AF avoids excess disk I/O, enabling data to be processed faster.
- Incremental computational processing
- With incremental computational processing, a pre-loaded query is processed iteratively when triggered by the input data, and the processing results are available for the next iteration. This means that the next set of computations does not need to process all of the target data elements; only those elements that have changed need to be processed.
- The following figure shows incremental computation on stream data as performed by Stream Data Platform - AF.
Figure 1-4 Incremental computation performed on stream data
![[Figure]](figure/zd010300.gif)
- As shown in the figure, when the first stream data element arrives, Stream Data Platform - AF performs computational process 1. When the next stream data element arrives, computational process 2 simply removes data element 3 from the process range and adds data element 7 to the process range, building on the results of computational process 1. This minimizes the total processing required, thus enabling the data to be processed faster.