TY - JOUR
T1 - Virtual PNNI network testbed
AU - Perumalla, Kalyan
AU - Andrews, Matthew
AU - Bhatt, Sandeep
PY - 1997
Y1 - 1997
N2 - We describe our experiences designing and implementing a virtual PNNI network testbed. The network elements and signaling protocols modeled are consistent with the ATM Forum PNNI draft specifications. The models will serve as a high-fidelity testbed of the transport and network layers for simulation-based studies of the scalability and performance of PNNI protocols. Our models are written in the new network description language TED which offers two advantages. First, the testbed design is transparent; the model descriptions are developed separately from, and are independent of, the simulation-specific code. Second, TED is compiled to run with the GTW (Georgia Tech Time Warp) simulation engine which is supported on shared-memory multiprocessors. Therefore, we directly obtain the advantages of parallel simulation. This is the first complex test of the TED modeling and simulation software system. The feedback from our experiences resulted in some significant improvements to the simulation software. The resulting PNNI models are truly transparent and the performance of the simulations is encouraging. We give results from preliminary simulations of call admission, set-up and tear-down in sample PNNI networks consisting of two hundred nodes and over three hundred edges. The time to simulate ten thousand call requests decreases significantly with the number of processors; we observe a speedup factor of 5.3 when 8 processors are employed compared to a single processor. Our initial implementations demonstrate the advantages of TED for parallel simulations of large-scale networks.
AB - We describe our experiences designing and implementing a virtual PNNI network testbed. The network elements and signaling protocols modeled are consistent with the ATM Forum PNNI draft specifications. The models will serve as a high-fidelity testbed of the transport and network layers for simulation-based studies of the scalability and performance of PNNI protocols. Our models are written in the new network description language TED which offers two advantages. First, the testbed design is transparent; the model descriptions are developed separately from, and are independent of, the simulation-specific code. Second, TED is compiled to run with the GTW (Georgia Tech Time Warp) simulation engine which is supported on shared-memory multiprocessors. Therefore, we directly obtain the advantages of parallel simulation. This is the first complex test of the TED modeling and simulation software system. The feedback from our experiences resulted in some significant improvements to the simulation software. The resulting PNNI models are truly transparent and the performance of the simulations is encouraging. We give results from preliminary simulations of call admission, set-up and tear-down in sample PNNI networks consisting of two hundred nodes and over three hundred edges. The time to simulate ten thousand call requests decreases significantly with the number of processors; we observe a speedup factor of 5.3 when 8 processors are employed compared to a single processor. Our initial implementations demonstrate the advantages of TED for parallel simulations of large-scale networks.
UR - http://www.scopus.com/inward/record.url?scp=0031359832&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0031359832&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:0031359832
SN - 0275-0708
SP - 1057
EP - 1064
JO - Winter Simulation Conference Proceedings
JF - Winter Simulation Conference Proceedings
T2 - Proceedings of the 1997 Winter Simulation Conference
Y2 - 7 December 1997 through 10 December 1997
ER -