TS/MP 2.5 Release Supplement

For example, a user can configure ACSCTL file on two nodes in this way:
System 1:\sys1
[PATHWAY DOMAIN]
%DOM1 = \sys1.$exam1, \sys2.$exam2
System 2: \sys2
[PATHWAY DOMAIN]
%DOM1 = \sys1.$exam1, \sys2.$exam2
The example shows that on both systems %DOM has the same PATHMONs, and the domain sends
issued on both systems are served by either of the PATHMON processes.
NOTE: TS/MP 2.5 does not support the sync-upgrade feature for a domain which has a remote
PATHMON process.
PATHMON as a HIGH-PIN process
PATHMON runs as a HIGH-PIN process by default.
Improved Link Selection Algorithm for Low-Load Conditions
The link selection algorithm is modified to give preference to static server processes over dynamic
server processes. ACS will use dynamic server processes only when static server processes are not
available. This results in a faster release of the dynamic server links under low-load conditions.
Reduced Server Cleanup Time in PATHMON
In this TS/MP version, PATHMON server cleanup time is reduced. This in turn improves the server
startup time since servers cannot be restarted until they are stopped, thereby increasing the
availability of the server.
Enhanced Server Process Rebalancing after a CPU Reload
In the current implementation when a CPU fails, all the static server processes running on that CPU
(if AUTORESTART > 0) are automatically restarted as static servers on other configured CPUs.
When the CPU is reloaded these restarted server processes continue to run on the same CPUs
unless they are explicitly stopped or killed. No static server processes are started on the reloaded
CPU. This can lead to an imbalance of server distribution across CPUs and reduced capacity. The
following techniques are implemented to avoid imbalance of server distribution across CPUs and
reduced capacity:
• When a CPU fails, the PATHMON process automatically starts the static server processes (if
AUTORESTART > 0) from the failed CPU on the remaining CPUs as static server processes.
The PATHMON process tracks these restarted static server processes as special.
• When the CPU is reloaded, these special server processes are shifted to the reloaded CPU.
PATHMON will try to shift one server process per server class at a time. After that server
process is restarted on the reloaded CPU, the next special server process of that server class
is taken for rebalancing. This method results in minimal impact on server availability and
response time during the rebalancing phase.
In this way, after a CPU failure increased load per CPU is still accommodated by the use of special
server processes; and after CPU reload, static server processes are restarted on the reloaded CPU
while the special servers are stopped, thereby avoiding an imbalance (that is, static server process
distribution is similar to what it was before the CPU failure), enabling the full usage capacity of
that reloaded CPU.
This feature is configured using one of the three modes - AUTO, MANUAL or DISABLE. A new
Pathway attribute (SPREBALANCEMODE) is provided to set one of these modes. This parameter
30 Introduction to TS/MP 2.5