NASA-CR-199526 Modeliner and Simulation of the Data Communication Network at the ASRM Facility R. P. Nirgudkar, R. J. Moorhead, and W. D. Smith Department of Electrical and Computer Engineering Mississippi State University Mississippi State, MS 39762 A Full-length Paper submitted to IEEE SOUTHEASTCON '94 April 10-13,1994 Miami, Florida Contact Addresses? R. P. Nirgudkar R. J. Moorhead Graduate Student Associate Professor P. 0. Box EE P. O. Box 6176 Electrical Engineering Engineering Research Center Mississippi State University Mississippi State University Mississippi State, MS 39762 Mississippi State, MS 39762 Phone: (601) 325-2220 Phone: (601) 325-2850 Fax: (601) 325-2298 Fax: (601) 325-7692 E-mail: [email protected] E-mail: [email protected] The funding for the research described herein was provided under NASA grant NAG8-866. (NASA-CR-199526) MODELING AND N96-11953 SIMULATION OF THE DATA COMMUNICATION NETWORK AT THE ASRM FACILITY (Mississippi•<State Univ.) Unclas 21 P G3/32 0069842 Modeling and Simulation of the Data Communication Network at the ASRM Facility ^~-— Abstract „-,.*-x> This paper describes the modeling and simulation of the communication network for the NASA Advanced Solid Rocket Motor (ASRM) facility under construction at Yellow Creek near luka, Mississippi. Manufacturing, testing, and operations at the ASRM site will be performed in different buildings scattered over a 1800 acre site. These buildings are interconnected through a Local Area Network (LAN), which will contain one logical Fiber Distributed Data Interface (FDDI) ring acting as a backbone for the whole complex. The network contains approximately i i 700 multi—vendor workstations, 22 multi—vendor workcells, and 3 VAX clusters 1 t interconnected via Ethernet and FDDI. The different devices produce appreciably ? different traffic patterns, each pattern will be highly variable, and some patterns will be very bursty. Most traffic is between the VAX clusters and the other devices. Comdisco's Block Oriented Network Simulator (BONeS) has been used for network I simulation. The two primary evaluation parameters used to judge the expected ' network performance are throughput and delay. 1.0. Introduction The Advanced Solid Rocket Motor (ASRM) facility at Yellow Creek near luka, Mississippi is part of a National Aeronautics and Space Administration (NASA) program to substantially improve the flight safety, reliability, productivity, and performance of the space shuttle's solid rocket motors. The facility is government-owned but contractor-operated. Lockheed Missiles and Space Company Inc. (LMSC) is the prime contractor. The operation of the facility is directed by the subcontractor Aerojet ASRM division (AAD); RUST International Corporation (RUST) is responsible for the engineering and construction of the facility [1]. The main aim of this paper is to present the overall communication network structure for the ASRM facility and describe the modeling and simulation of the network. The section on 'ASRM Communication Network Structure' concentrates on the network connectivity, the cabling, and the protocols. That section also explains the flow of data in the network. The section on 'BONeS Modeling* gives an overview of the BONeS simulator and describes the different BONeS models developed to simulate the ASRM environment. The 'Analysis and Results' section at the end of the paper comments on the network expectations and the network evaluation parameters. The section also summarizes the various plots of Mean Delay and Throughput versus Traffic Intensity obtained using BONeS and some important conclusions are drawn. 2.0. ASRM Communication Network Structure 2.1. Main Computing Center Building 1000 (B_1000) provides an efficient means to plan, control, and document the manufacture of solid rocket motors for the ASRM project. All the workstations and workcells communicate only with the Operational Information System (OIS), the Business Information System (BIS), and the Area Supervisory Computers (ASCs) in B_1000; there is no peer—to—peer communication. B_1000 also provides a link between the business functions and the manufacturing functions of the facility. The interconnection between the devices in B_1000 is shown in Figure 1. The OIS is a VAX cluster consisting of two VAX 6000 computers, each with one FDDI adapter. In addition to this OIS VAX cluster, there are two VAX 4000 computers, each with one Ethernet adapter. The BIS is a VAX cluster consisting of one VAX 6310 and two VAX 6420 computers. Each of the ASCs are VAX 4400 with an Ethernet adapter. Building 1000 ^ ASC OIS •• ™ BIS ASC ^3 Devices w CIIS 4001 Cabletron _^_^_ I x / //// / I \ Building 1001 Building Building Building Building Building Building Building \ l -\ f I 'I Building 2087 2029 2030 2066 1016 2031 2029 1002 = FDDI Building Building Building 10BASE-FL 10BASE5 1025 1045 1010 10BASE2 Figure 1 Inside Building 1000 B_ 1000 has a Gandalf terminal server. The Gandalf terminal server is a large terminal server with a multitude of RS-232 ports. The Gandalf can support 12 separate Ethernet channels. It is the only terminal server throughout the campus. The OIS, the BIS, the ASCs, and the Gandalf terminal server in B_1000 are connected to the outside network complex by Cabletron Multi Media Access Centers (MMACs), the intelligent hubs. There is a fiber connection between the OIS and the Cabletron hubs. There is a copper connection between the BIS and its Cabletron hub, and between the ASCs, and their Cabletron hub. In addition to the devices already mentioned, B_1000 has nine more Cabletron hubs distributed in two switch rooms for the BIS devices. The BIS devices includes 32 printers, 25 CAD workstations, and 400 Macintosh computers connected to the BIS hub by 10BASET; 50 PCs connected to the BIS hub by 10BASET; 31 Engineering workstations on 10BASE2; and 289 dumb terminals connected via Asynchronous Data Interface to the Gandalf. The connections of the BIS devices is as shown in Figure 2. BIS HUB Connections Inside BJOOO / CAD N/W \ HUB i 30 Integraph 6 - 8 SGI / \ / nr-i / \ HUB flP. / ENGRN/W ] \ 8 to 10 servers / 32 MAC Laser Printers (Apple talk) <10 times/day ,3 K file (20 pages/request)) ^» , Figure 2 BIS Hub Connections 2.2. Network Cabling at the ASRM Site At the ASRM site all the outdoor cabling and much of the indoor cabling is optical fiber. Thin-wire coax, thick-wire coax, and twisted-pair are also used in B_1000. 2.2.1. Outdoor Cabling at the ASRM Site All the outdoor cabling is optical fiber. All optical fibers used are 62.5 / 125 micron multimode optical fibers. Every FDDI hub has at least three redundant paths, viz. Channel A and Channel B of FDDI and a 10BASE-FL backup. Also every FDDI hub has two redundant dual rings. Link Distance Number of Number of TVpe of the (feet) Workstations Workcells Link on the link. on the link. Link # 1 : (2029) 6700 29 04 Intensive and Non-Intensive Link # 2 : (2031) 4650 15 00 Intensive Link # 3 : (1016) 1450 27 16 Intensive Link # 4 : (1000) 00 (19+73) 00 Non-Intensive + BIS Devices. Link # 5 : (2066) 3550 09 00 Non-Intensive Link # 6 : (2030) 5000 12 06 Intensive and Non-Intensive Link* 7: (1012) 950 14 00 Non-Intensive Table 1: Distances of each hub from B_1000 The manufacturing intensive buildings have two FDDI data paths from B_1000 with automatic switch-over. One data path is buried, while the other is aerial. Buildings 1016,2029,2030, and 2031 are the manufacturing intensive buildings; each has a hub directly connected to a hub in B_1000. Buildings 2060 and 2076 are connected to the hub in building 2029. Each hub receives two pairs of fibers from the outside cable plant. All workstations and workcell devices will receive two fibers each from the respective hubs. All the manufacturing non—intensive buildings in the complex receive two fibers for their hub via the outside cable plant. All the workstations inside the buildings get two fibers each from the respective hub. Total OIS Nodes = 184 W/S 26w/c 4 Intensive Links and 5 Non-Intensive Links ENGINEERING/ COMPUTER (Intensive and Non-intensive) m ' LINK#1 •iS (Intensive and Non-intensive) | 2029 |— ' | 2060 | 1 | 2076 | 1 | 2028 1 04 w/c | 2082 | FINAL 29 w/s ASSEMBLY Figure 3 Hub Connections 2.2.2. Indoor Cabling at the ASRM Site For the indoor cabling in the manufacturing intensive buildings, the 10BASE-FL protocol is used, mainly because it allows lower light levels and 16 redundant data paths [7]. 2.3. Cabletron Devices At the ASRM site all the Cabletron MMACs will be MMAC-8FNBs, which support up to seven Media Interface Modules (MIM). The first slot in the MMAC will be the EMME multichannel management / bridge module. The Cabletron hubs provide the necessary security, routing, and redundancy [7]. The different MIMs used in the network at the ASRM are listed in Table 2. Name of the TVpe of the Protocol Number of Comments Card Card ports EMME Ethernet Bridge Ethernet 4 ports Used as a man- agement module in all the MMACS FDMMIM FDDI to Ethernet 8 ports Connects 10 Mbps Ethernet to 100 Bridge Mbps FDDI MT8-MIM DELNI Card Sports AUI Transceiver — FORMIM-22 10BASE-FL card Ethernet 12 ports Provides connec- tivity for 12 ethemet channels TPRMIM-36 10BASE-T Card Ethernet 24 ports Provides connec- tivity for 24 ethernet channels CXRMIM DEMPR Card Ethernet 12 ports Provides connec- tivity for 12 ethernet channels GX-M GatorStar Card LocalTalk 24 ports 24 port LocalTalk repeater with a LocalTalk to Ethernet router FDMMIM-04 FDDI Concentra- FDDI 4 ports Provides 4 con- centrator ports tor FDDI connections Table 2: Cabletron MIMs used at ASRM 2.4. Protocols used at the ASRM Site For the communication network at the ASRM site, two protocols are specified, FDDI and CSMA/CD. All the manufacturing intensive buildings are connected to B_1000 by links with FDDI protocol, and all the manufacturing non-intensive buildings are connected to B_1000 by links with 10BASE-FL protocol (i.e. CSMA/CD on optical fiber). The protocol inside the manufacturing intensive buildings and the manufacturing non-intensive buildings is 10BASE-FL. 2.5. Data Flow over the Network 2.5.1. Manufacturing Intensive Buildings The manufacturing intensive buildings have FDDI data paths from B_1000. The manufacturing intensive buildings viz. 1016, 2029, 2030, and 203 1 each has a hub directly connected to an OIS hub in B_1000. The workstations in these buildings communicate with the OIS 6000 computers. The workcells in these buildings communicate via the OIS hub with the two ASCs. 2.5.2. Manufacturin Non— Intensiv The manufacturing non-intensive buildings 1001, 1002, 1010, 1025, and 1045 have 10BASE— FL data paths from B_1000. All other manufacturing non-intensive buildings are connected to B_1000 through the nearest manufacturing non-intensive hub in buildings 2029, 2066, and 2030. The workstations in all the manufacturing non-intensive buildings communicate via the BIS hub with the two OIS 4000 computers. 2.6. Research Objective The main objective of the research is to simulate and analyze the network to determine its performance under different load conditions. Comdisco's Block Oriented Network Simulator (BONeS) is used to evaluate the performance of the network with the given topology and protocols. The two primary evaluation parameters that are used to judge the network performance are the throughput and the delay. The aim of the simulations is to look into the loading of the OIS, the BIS, the ASCs, and the network links due to the traffic generated by the workstations and the workcells over the entire site. 3.0. BONeS Modeling 3.1. Different Methods of Network Modeling F41 The network modeling can be done by several different means, each having its advantages and disadvantages. The first method is by developing a mathematical model of the network, normally using queueing theory. This model can then be used to provide results of the performance of the network. Due to the simplifying assumptions required to use with this type of modeling, it is often not the best possible model and often is not feasible. The second approach to analyze the performance of a network is to actually build the network. Although this approach provides very good results, it is normally very expensive, both in time and resources. The third approach is that of computer simulation. Using a computer simulation the user can model the network to as close to reality as desired. This approach is less expensive than building the network; however, the major disadvantage of this approach is insuring that the simulation model accurately models the real-world network. 3.2. BONeS Simulator F61 The Block Oriented Network Simulator (BONeS) provides an interactive graphical environment for simulation-based analysis and design of a broad range of communication networks. In the BONeS environment, the network model is specified in terms of the network topology, traffic, packet and message (data) structures, and protocol functions. The user constructs the network graphically and hierarchically using the building blocks from the BONeS model library. The user can also write the components of a model in C and incorporate them into the BONeS modeling environment. BONeS translates the network model into a C program, executes an