Today’s network-centric military relies heavily on mobile ad-hoc networks. Net-centricity is a force multiplier that relies on adaptive communication technologies and dynamic network quality of service (QoS) to enable mission-critical applications. In order to build next-generation communication systems, system developers need to accurately predict end-to-end performance. But while traditional network simulations assumed near-perfect communications, they don’t reflect the reality of the battlefield.

In a modern-day combat theater, decision making depends on information flow. Combat networks pass data that are subject to delays. These delays are unequal depending on routing protocols, terrain, environmental effects, connectivity, hops, priorities, available bandwidth and traffic. The perceived truth of the state of a network can significantly lag the truth on the ground.

Because so many factors go into determining network performance, the simulation of such systems is computationally very demanding. To be an expert on next-generation wireless networking requires a complex set of expertise. This includes an understanding of mobile communication, quality of service, software defined radios and the network-centric services that support them. Network and equipment tests traditionally required months to perform all the calculations, or much of the details were abstracted, rendering the results useless. Still other solutions modeled only small scale networks (200 or fewer devices), whose performance is not extrapolatable to the larger scale networks expected to be deployed in combat.

Fidelity and Scalability

High-fidelity network models are hard to achieve in simulation, yet they are critical to effective analysis of adaptive communications (Figure 1). Existing tools use abstractions that hide critical effects. Scalability is also a big issue with network simulation. Models must project to operational size for meaningful results, but existing tools do not have sufficient processing power. Existing simulation programs, written with legacy sequential processing code, can only simulate a maximum of about 200 devices, and fidelity drops as you approach that number.

Application-centricity is key to making network simulation meaningful and relevant. Rather than providing results in terms of delay or packet delivery rate, simulations are most valuable when they can test protocol and device impact on end-to-end performance. Existing tools can’t support simultaneous multiple applications, cross-layer interactions or other application-centric approaches to simulation.

An example of a tool that supports high-fidelity emulation and simulation for predicting network performance is Scalable Network Technologies’ (SNT) QualNet product. It brings the speed and scalability of parallel processing to wireless network modeling and simulation applications for military and commercial wireless network design and deployment. QualNet enables developers to digitally represent an entire network—devices, software, transmitters, antennas, terrain effects, atmospheric effects and human interaction effects. One can now represent every variable that will affect the performance of your real network in QualNet without trading accuracy for scalability, or vice versa.

With real-time simulation speed, network and equipment tests that traditionally required months to perform all the calculations can now be performed in minutes, with real-time speed and real-network behavior. QualNet allows developers to move from months to minutes. QualNet scales up and simultaneously offers ultra-fidelity at 50 or 5,000 nodes. That means it’s possible to get the same accurate representation of a network whether you’re testing 50 nodes or 5,000.

The tool’s highly scalable kernel is natively multiprocessing and thread safe. QualNet runs on all leading embedded operating systems and hardware, including multicore processors, multi-processor architectures, clusters and supercomputers. Figure 2 shows an example of how parallel processing affects simulation speed in a 700-node detailed wireless simulation. QualNet also allows integration of customer-developed models. It can run the most detailed protocol models so users can directly incorporate their code into QualNet without abstracting or re-writing it.

Standards Enhance Quality

Key enablers of traditional off-the-shelf network simulators have been open standards such as DIS, HLA, SDF and PDEF. They have enabled free market competition to drive quality up and prices down. QualNet supports those standards and interfaces with other simulations readily. It is also embeddable as a Communication Effects Server (Figure 3) in diverse network-centric planning, training and demonstration systems. By serving as a simulator of a specific type of network, where only one particular stack is relevant, QualNet can be optimized even further for leaner and meaner network simulations. By being natively parallel, QualNet has eliminated the trade-off of network simulation in the past, which meant giving up one for the other—such as speed, scalability and fidelity.

Scalable Network Technologies
Los Angeles, CA.
(310) 338-3318.
[www.scalable-networks.com].