z/OS MVS Programming: Resource Recovery
Previous topic | Next topic | Contents | Contact z/OS | Library | PDF


Distributed resource recovery

z/OS MVS Programming: Resource Recovery
SA23-1395-00

The resources that a work request updates can be distributed — reside on more than one system. Figure 1 shows the programs that participate in distributed resource recovery, also called distributed syncpoint processing.

When resources reside on multiple systems, a part of the application must run on each system. Using the ATM example presented earlier. assume that the resource manager for the savings account database runs on system A, and the resource manager for the checking account database runs on system B. We know from the earlier example that the ATM application has three main responsibilities:

  1. Communicate with the ATM user
  2. Update the savings account database using the resource manager on system A
  3. Update the checking account database using the resource manager on system B

Assume that the part of the application running on system A (APPL-A) always communicates with the ATM user.

When the ATM user requests the transfer of money from savings to checking, APPL-A sees the request, and uses the resource manager on system update the savings account database. Completing the transaction means that APPL-A must communicate with the resource manager on system B to update the checking account database. Thus, part of the application (APPL-B) must reside on system B.

APPL-B listens for a signal from APPL-A, and the signal tells APPL-B to use the resource manager on system B to update the checking account database. Another way to implement the application would be for the signal from APPL-A to actually initiate APPL-B.

In either case, APPL-A communicates with APPL-B through a communications resource manager (CRM), such as APPC/MVS. The main function of a CRM is to allow applications to communicate across systems, but it also allows RRS on one system to communicate with RRS on another system.

Keep this brief description of distributed processing in mind. It applies to the two models for processing distributed resources: the peer-to-peer model and the client-server model. Which model is appropriate depends on the application and how its resources are distributed.

Figure 1. Transaction — Syncpoint Processing
REQTEXT

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014