With the growing size of simulation requirements, high-performance
scalable simulation systems are in high demand. Along with a
simulation model that has higher parallelism and scalability, the
obtaining and exploiting knowledge about the simulation system is
essential for high performance. In this research, performance models
of the simulation protocols are developed. Simulation model
properties are identified and abstracted for performance analysis; in
addition, the hardware factors of a simulation computer system are
also measured.

Two prominent simulation protocols are studied: the conservative
parallel simulation protocol and the optimistic parallel simulation
protocol. Detailed protocol-dependent factors are identified from a
detailed description of the simulation runtime mechanism. The
protocol-dependent performance models are derived as follows. For the 
conservative simulation protocol, the critical path of logic
processes governs the performance of the simulation system. The
protocol requires the safeness of a message to be identified in order
to meet synchronization rule and simulation correctness. The number
of safe messages and the size of the Earliest Incoming Time (EIT)
update interval are dependent on the critical parent. Performance 
results are shown in the physical time domain as well as the
traditional logical time domain.

The optimistic simulation system, such as the Time Warp Operating
System (TWOS), has its own specific synchronization of rollback and
recomputation. Rollback requires state saving. As well as state
saving overhead, rollback probability is identified as a core factor
of optimistic simulation system performance. Lazy cancellation
efficiency and simulation topology connectivity are two core factors
for optimistic simulation system performance. The research on
optimistic simulation system performance is extended to adaptive
simulation systems. For an adaptive simulation system, which combines 
conservative simulation and optimistic simulation, the rollback
probability of the optimistic part is analyzed from the detailed
description of the protocol and requirements for control mechanisms.

This performance research results provide important information about
the internal workings of simulation system, and consequently,
possible improvements can be made to simulation system performance.
The Earliest Outgoing Time based scheduling mechanism for
conservative simulation system is selected to demonstrate the
significance of this approach.