OSI/MHS Configuration and Management Manual

Sizing and Tuning Your OSI/MHS Subsystem
OSI/MHS Configuration and Management Manual—424827-003
8-6
Using Parallelism in OSI/MHS
MS Groups
You can use parallel processing to optimize performance of your message store in
either or both of the following ways:
•
Define multiple MS groups and distribute the MS APPLs across them. This
configuration typically provides better performance than having one MS process
serve all message-store users.
•
Associate multiple MR groups with the same MS group, using the MS group
attribute MR-COUNT. This configuration can prevent or alleviate bottlenecks
between MR and MS groups.
GI Groups
You can define multiple GI groups and GATEWAY objects to satisfy different
application requirements. Here are some possibilities:
•
Design your application to use different gateway interface processes (GIPs) for
sending and receiving messages.
•
Run only enough GPI applications per gateway to receive the load of messages
anticipated for the gateway. Running extra applications is likely to undermine
rather than improve performance, because if multiple applications attempt to
transfer in the same message, one will obtain the message and the others will
cause TMF aborts, degrading performance.
Multiple GPI applications are likely to be required to receive the messages entering
a P1-exit gateway, because a P1-exit gateway receives every message that enters
the system from any adjacent MTA. (A single GPI application is sufficient only if
traffic is very light.)
•
If multiple GATEWAY objects will be open simultaneously, define at least one GIP
per GATEWAY object. Define as many additional GIPs as are needed to handle
the traffic through the gateways. (Multiple GIPs can serve the same GATEWAY
object at the same time.)
For information about the design of gateway applications, see the Gateway
Programmatic Interface (GPI) Programming Guide.
MTA and ROUTE Objects
As described in Section 2, Management Environment for OSI/MHS, you use the MTA
object to define adjacent MTAs, and ROUTE objects to reach the same user through
several different adjacent MTAs. Thus you can distribute outgoing traffic among MTAs.
Alternatively (or additionally), you can define multiple MTA objects all of which refer to