Implementation consists of two main parts.

Stream generator is implemented as testing facility for data generation. It runs on one host and can be configured to listen to n different ports and generate streams of input data for the main application. It generates infinite streams of data, in chunks of random number data. Maximum count of elements in a chunk is configurable. It is possible every stream to be generated with different speed.

Stream combiner works on n streams, to which it is connected via network sockets. Host and port for every stream is configurable. There is no restriction that all streams should come from the same host.

All streams are merged into one and this resulting stream is fed into processing part. I pipeline of processing steps is assembled and data flow thru. Data comes in as XML and transforms in internal immutable object. Next a window with duration of 250 ms (this should be tuned) and all data items in this window are grouped by timestamp. Result is bunch of substreams holding the same timestamp and different amount. On every substream a reduction step is performed - meaning that all amounts in the substream are summed up and from every substream only one value is generated. These steps are performed in parallel. All resulting data items are the emitted in the main stream. Last step is transforming every data item from internal object to JSON representation.

##Configuration

The configuration is done in file named application.yml, which is part of the distribution or can be provided externally. List of host/port pairs are given, under unique names according the YAML syntax. It is important that these name are unique because are used as keys in a map.

The configuration is done in file named application.yml, which is part of the distribution or can be provided externally. List of host/port pairs are given, under unique names according the YAML syntax. It is important that these name are unique because are used as keys in a map. Maximum number of elements is also part of the configuration. Server port is defined, and has to be different form the one used by the combiner if both applications run on the same host.