Because other Adaptive Servers involved in a distributed transaction may also coordinate remote participants, transactions can be further propagated to additional servers in a hierarchical manner. For example, in Figure 3-1, the client connected to ASE1 begins a transaction that executes an RPC on ASE2 and an RPC on ASE3. The coordination service for ASE1 propagates the transaction to ASE2 and ASE3.
Since ASE2 also has transaction coordination services enabled, it can propagate the transaction to additional remote participants. Here, ASE2 propagates the transaction to ASE4 where data is updated using CIS.
Figure 3-1: Hierarchical transaction coordination
In Figure 3-1, ASE1 is referred to as the commit node for the distributed transaction. When the transaction on ASE1 commits, the coordination service for ASE1 instructs ASE2 and ASE3 to prepare the transactions that it propagated to them. ASE3 indicates that its transaction is prepared when its local work is ready to be committed. ASE2 must complete its local work and instruct ASE4 to prepare its transaction. When the transactions are prepared in ASE2 and ASE4, the coordination service in ASE1 commits the original transaction. The instruction to commit subordinate transactions is then transmitted to ASE2, ASE3, and ultimately to ASE4, in the same manner as the instruction to prepare was transmitted.
