How zFBA works with IBM Connect:Direct
IBM® Connect:Direct® utilizes a TCP/IP connection with its remote partner for its normal application (FMH) communication path, and two zFBA devices on the DS8K for data transfer.
IBM Connect:Direct for z/OS® uses an IOS API to allocate, read, write, unallocate and erase the zFBA devices. IBM Connect:Direct for z/OS defines and manages the zFBA devices within its configuration in its NETMAP file and the remote IBM Connect:Direct partner is not required to define the zFBA configuration in its NETMAP file. IBM Connect:DirectUNIX only need to request that IBM Connect:Direct for z/OS allocate two zFBA devices for the COPY.
In a non-zFBA COPY, the data and all FMHs are sent using TCPIP in units where the maximum size is determined through the negotiation between two nodes.
In a zFBA COPY step, the data (and any checkpoint FMHs) are transferred with a pair of zFBA devices, in units of 32MB (the last one is usually smaller). The zFBA devices must be defined to be at least 32MB and any amount in excess of 32MB is unused by the IBM Connect:Direct zFBA feature. During the bulk of the transfer, the sending node is writing to one zFBA device in a pair; the receiving node is simultaneously reading the other zFBA device and writing the data to the destination data set. When the sender is done writing to a zFBA device, it communicates to the receiver over the TCPIP connection directing it to the new data. When the receiver has finished reading the data, it communicates back to the sender that the device is available and allows IBM Connect:Direct to reuse the zFBA device. The switching of zFBA devices occurs many times in a COPY step.