While the method of capture varies based on the source of the data, only two methods are used to manage transient data. A fundamental premise of the product architecture is to avoid "landing" the data from the point of capture to the application of CDC data to target datastores. This is accomplished by using either z/OS system LogStreams or a proprietary memory mapped storage system referred to as CDCStore. In both cases the Publisher component utilizes TCP/IP to transport data to subscribing Engines:
While not part of a true Capture/Publisher framework, TCP/IP can be used by an Apply Engine to communicate directly with other Apply Engines on the same system or across the network to an Engine running on a different operating system platform.