Multipath channel (MPC) connections allow you to code a single transmission group (TG) for host-to-host communication that uses multiple write-direction and read-direction subchannels. The subchannels can be assigned to one or more physical channels based on how you have defined your I/O subsystem to the operating system. Because each subchannel operates in only one direction, the half-duplex turnaround time that occurs with other channel-to-channel connections is reduced.

To define an MPC connection between two subarea nodes, code a channel-attachment major node (VBUILD=CA) with LNCTL=MPC on the GROUP definition statement. For more detailed information about defining subarea MPC connections, see z/OS Communications Server: SNA Resource Definition Samples.

If you code a TG in which the subchannels are divided between two physical channels, you can increase availability because the TG will have a path to use, even if one physical channel is down. Because each TG can use more than one channel, and because the turnaround time required for half-duplex is reduced, throughput is increased.

The subchannels on the physical channel are represented by the subchannel addresses coded on the READ and WRITE operands on the LINE definition statement in the channel-attachment major node. One READ subchannel in one host and the corresponding WRITE subchannel in the other host form a complete path. The READ subchannel address and the corresponding WRITE subchannel address must reference the same physical connection between the two nodes; the two addresses do not need to be identical. You should consider the ratio of inbound and outbound traffic for a host in deciding how many write subchannels and read subchannels you want to code.

If multiple write subchannels are used, traffic is distributed among the subchannels as traffic requests increase. Arriving data is resequenced before being presented to the application. Priority is assigned to the write subchannels in the order that they are defined.

If you want two TGs, with each TG dividing its subchannels across two parallel physical channels, you have to code two groups, one for each TG. The LINE definition statement for each TG will include subchannel addresses on the READ and WRITE operands from each of the two physical channels.

If a subchannel becomes inoperative, the TG remains active as long as one subchannel is available in each direction. If the last subchannel in either direction becomes inoperative, issue a VARY INACT command, correct the problem (all subchannels must be allocatable when activated or activation will fail), and then issue a VARY ACT command to reactivate the major node.