Benchmarking Network Protocols

12 March 2019

In any distributed virttml simulation enviromnent, the underlying network architecture and its protocols play an important part in its performance. Tiffs paper describes tile different underlying protocols used in file support of the RTI implementation in the Federated Simulations Development Kit (FDK). The communication FM and MCAST modules were modified to support different protocols. Tile performance of two different protocols: TCP and a new Lightweight Reliable Multicast, called Pseudo Reliable Multicast Protocol (PRMP), running on top of two different network architectures Ethemet and Asyncltronous Transfer Mode (ATM) were compared. The latter protocol was developed specifically to support the distributed virtual simulation enviromnent. Furthermore, in the case of the ATM network architecture, the use of native ATM-API was also implemented and its performance compared with file other protocols. The benclunarks used to compare their performance are Latency and Time Advance Request Benchmark. The results show that PRMP outperforms the other protocol teclmiques when the number of subscribers are large and when file bandwidth is limited. But it has some additional latency overhead, due to additional processing required to provide tile reliability. needed by the sender and receivers. Comparing the network architecture, the benclunark performance of the above protocols operating on top of 100BaseT switch network performs much better than over ATM network, although the transmission speed is much higher in file case of the latter.

Introduction

Distributed Virtual Simulation Environment (DVSE) is an important strategic technology for linking simulations of various types, at multiple locations to create a realistic, complex, “virtual world” for the simulation of highly interactive activities. One major use of DVSE is the Distributed Interactive Simulation (DIS) [1,2] system for military simulations such as war-gaming, where geographically distributed hardware and personnel interact with each other as if they were in actual combat situations. The US Department of Defence (DOD) has adopted the High Level Architecture (HLA)[3,4] as the framework and the RunTime Infrastructure (RTI) as the infrastructure to support the DVSE.

One of the major issues faced by developers of DVSE is the communication requirement. Increasing the bandwidth or transmission rate does not necessarily give better performance. The efficiency/appropriateness of the protocol to support DVSE is also an important factor that affects its performance.

Over the past decade two network architectures have made significant impact: 100BaseT[5] and Asynchronous Transfer Mode(ATM)[6]. They have changed the transmission rate of networks which used to be in the 10s of megabits to 100s of megabits. They have both been widely accepted and deployed in various organisations.

With regard to the part on protocol, Interact Protocol (IP)[7] has been widely used in network communication. IP was developed for general purpose transmission of data. IP native multicast is a very useful mechanism for group communication found in DVSE and a number of researchers[17] have used reliable multicast in DVSE. However, such protocols are not tuned for DVSE. We have proposed and implemented a light-weight reliable multicast protocol, Pseudo Reliable Multicast Protocol (PRMP) [8], which runs on top of IP. It was designed specifically to support group communication found in DVSE.

In our paper, we will focus on the performance of the different network architectures, 10/100 BaseT Ethernet and ATM network as well as the performance of different protocols in DVSE. Section 2 will introduce the Federated Simulation Development Kit which provides the RTI for the DVSE. Section 3 will introduce the different protocols and network architectures. Section 4 will report on the benchinark results. Conclusion drawn from the experiments are found in section 5.