Tuning a UNIX domain manager to handle large numbers of fault-tolerant agents

The performance of domain managers on UNIX is impacted if they are overloaded with jobs. Improvements can be obtained by modifying the kernel parameters. The precise settings differ according to operating system, and you might need to test different settings to obtain optimum performance.

The following is an example of the kernel settings for Linux (kernel t3.10.0-514.e17.x86_64) to handle 500000 jobs per day workload:
data seg size=unlimited
scheduling priority=0
file size=unlimited
pending signals=124946
max locked memory=64
max memory size=unlimited
open files=105000
pipe size=8
POSIX message queues=819200
real-time priority=0
stack size=10240
cpu time=unlimited
max user processes=16384
virtual memory=unlimited
file locks=unlimited