OSI/MHS Configuration and Management Manual

Sizing and Tuning Your OSI/MHS Subsystem
OSI/MHS Configuration and Management Manual—424827-003
8-5
Average and Maximum Message Size
periods and the delivery requirements for those messages (for example, the
proportions of urgent, normal, and nonurgent priority messages).
Average and Maximum Message Size
Sizing and throughput are affected by the average and maximum message size. The
types of body parts and attachments also affect how messages are processed and
therefore what overhead is associated with each message.
Large messages can have a dramatic impact on performance. It is best not to send
them during peak traffic periods.
Message Store Use
In addition to the number of messages destined for the message store, sizing and
performance are affected by the subscribed attributes for APPL objects and the length
of time a message resides in the message store before it is retrieved.
Checking Sizing and Distribution
You also need to look at both the requirements of your message handling applications
and the other work each system is doing. For example, a system to which you forward
all OSI/MHS accounting event messages for processing by billing applications
potentially requires more disk space for EMS and less disk space for the MS database
than a system that is managed by remote applications but supports a large number of
message-store users. The information in this subsection represents a selection of
information relevant to sizing MHS configurations.
Using Parallelism in OSI/MHS
You can define multiple instances of OSI/MHS groups to increase throughput.
MR Groups
It is usually desirable to define multiple MR groups. Such groups can send and receive
messages in parallel and also provide fault tolerance:
•
If you define multiple MR groups that receive on the same address, the failure of
an MR or RTS process (or of the CPU in which it runs) does not prevent adjacent
MTAs from using the address for access to OSI/MHS.
•
If you define multiple MR groups that receive on different addresses, adjacent
MTAs can retry association requests on different links should any one address
become unavailable.