WSC 2005

WSC 2005 Final Abstracts

Simulation Case Studies D Track

Tuesday 3:30:00 PM 5:00:00 PM
Goverment and Military Case Studies

Chair: Colonel Michael McGinnis (West Point)

Temperature Control for Aerospace Test Systems- Modeling, Simulation and Design Case Study
Nagy N. Bengiamin (California State University Fresno) and Allen Parker, Larry Hudson, and Van Tran (NASA Dryden Flight Research Center)

Temperature testing of aerospace surfaces and materials requires raising the temperature via air heat transfer to over 2,500 F°. Quarts lamps or graphite heating elements emit infrared waves that create the needed heat flux at the surface of the specimen that is usually divided into zones for different desired heating profiles. Heating elements activation pattern, arrangement of the elements and the shape and material of the specimen influence the temperature at its surface at a given time. Transfer of heat by convection between zones, radiation from neighboring heating elements, dynamics of heat flux and air turbulences pose significant challenges when designing a universal test cell. The heating test usually follows a ramping up/down profile, which calls for implementing a tracking control system. This paper presents mathematical modeling, simulation and control system design practices in a case study format where typical test data are utilized. Inherent system non-linearity and practical implementation aspects are addressed.

The Army’s MARATHON Model
Mark W. Brantley and Steven A. Stoddard (Center for Army Analysis)

In order to support Army force structure analysis efforts, the Center for Army Analysis has developed a discrete event simulation called MARATHON. This model allows us to simulate the flow of active and reserve component units through their respective lifecycles. Each lifecycle begins with a non-available period (when AC units are reset and RC units are not available for Title 10 operations), followed by periods when units train until they are ready and available, deploy, recover, and transform (as necessary). MARATHON allows us to examine a variety of force structure options by illustrating gaps or redundancies in capabilities, as well as associated deployment tempos. Since the Army has adopted MARATHON to analyze its force structure decisions, we have also developed extensions to this simulation that have the capabilities to analyze how the force structure decisions impact the ability of the Army to provide Soldiers and equipment to its units.

Big Government Software Simulation Projects For Those Who Actually Have To Do It
Ralph R. Nebiker (Sonalysts, Inc.)

Practical tips for getting the job done. Software simulation development experiences the same problems, if not more, as do other large software development projects. Automated software development is the future of software engineering for large projects because manually supported coding and testing can't do the job, but automated software development is a big change for the government, which is based on managing people through document review. Change management and working with staffs slow to accept change is the biggest problem. Becoming proficient at abstraction was our biggest technical challenge. It was two years before we were able to stand on our own technically, and after three years some support personnel still didn't understand our process.

Wednesday 8:30:00 AM 10:00:00 AM
Modeling Internet Applications

Chair: Anand Ganti (Sandia National Laboratories)

Modeling Overloaded VoIP Systems
Martin J Fischer and Denise M. Masi (Mitretek Systems)

Since 9/11 many telecommunications systems have seen the need to deal with overloading during high traffic conditions. As the use of VoIP increases an important question that needs to be answered is: “Can the voice packets be delivered in a timely fashion when the there has been a significant increase in traffic?” In this paper we considered the problem of modeling VoIP systems in an overloaded condition. We look at the problem from a simulation and analytic point of view. We present analytic models for the packet latency and jitter and loss probability for the three prevelant disciplines being used for VoIP today: First Come First Served, Priority Queue and Weighted Fair Queueing Systems. In addition, we investigate how simulation languages like GPSS/H compare with respect to runtime with simulation models developed using Visual Basic for Applications and if their runtimes are acceptable for practical use.

Agent Based Consumer Choice Model of Broadband Internet Services Incorporating Geographical Limitations of Technologies
Conrad Mark Fonseca (Evans & Peck Pty Ltd)

Telecommunications providers are seeking to capitalise on the growth potential of broadband internet services in Australia. Choice of technology roll-out by providers must balance technology roll-out cost, technology limitations of broadband speed decreasing with distance from source, and consumer’s demand for internet services requiring increasing broadband speeds over time. Evans & Peck and Alcatel Australia jointly developed a consumer choice model which allows scenario analysis on the potential impact to providers over the next 10 years. Critical to the consumer choice behaviour in the model were the geographical aspects of consumer location in relation to technology sources (eg. telephone exchange) which ultimately setup a choice matrix of technologies and providers for that particular consumer. A user interface was created in which a map of the geographical area of a single telephone exchange could be overlaid with both the location characteristics of households (density and consumer type)and of technologies coverage.

Supercomputing Interconnects
Anand Ganti, Thomas Tarman, and Jason Wertz (Sandia National Laboratories)

This case study describes efforts at Sandia National Laboratories to model and simulate the interconnection network for the Red Storm supercomputer. Red Storm is a Massively Parallel Processor (MPP) machine that uses commodity processors (10,368 AMD Opterons) combined with a very high performance 3-D mesh interconnect. We consider a processor (compute node) with its associated NIC (Network Interface Card) as an atomic unit. The interconnect is built by connecting the atomic unit as a 3-D mesh with a cut-through in the Z-Axis. We model the processor, the NIC and the interconnect. We model the NIC at a functional level. Since we are only interested in the communication aspect, we model the processor as an object that can generate and receive messages according to a tunable random process. We simulate the communication pattern and evaluate the throughput and delays of the interconnect. We present in detail our modeling methodology, design, initial results, and model validation criteria.