Simulation Study for a Broadband Multimedia VSAT Network

Yi Qian, Rose Hu, and Hosame Abu-Amara

Nortel Networks 2201 Lakeside Blvd., Mail Stop 992-02-E70 Richardson, Texas 75082, USA Phone: 972-685-7264 Fax: 972-685-3463 Email: {yiqian, rosehu, hosame}@nortelnetworks.com Abstract We present a simulation approach for assessment of quality of service and bandwidth utilization for a meshed broadband VSAT system over a geostationary (GEO) satellite. The model includes a bent-pipe satellite with a number of broadband VSATs carrying multimedia traffic. A multi-frequency time division multiple access (MF-TDMA) is used with an integrated call admission control (CAC) and bandwidth on demand (BOD) algorithm for uplink from the VSATs. Traffic source models of various multimedia services are implemented along with the CAC and BOD scheme by using OPNET. Using this simulation platform, we measure the quality of service for different applications and uplink bandwidth utilization. 1. Introduction Widespread deployment of global corporate Intranet enhances multimedia information exchange between different branches and regional offices of international corporations. Intranet will also drive LAN-based cooperative work in design, engineering, and manufacturing (CAD/CAE/CAM). The multimedia information exchanges will require higher data rates for voice, data, and video traffic in inter-LAN applications. The aggregate data rates of the VSAT inter-LAN interconnections are expected to be in a broad range between 1.5 Mbit/s to 100 Mbit/s. To meet the requirement for more capacity and increase in reach, new generation systems are being designed to offer more throughput and support a larger number of VSATs, which adds more complexity to the network control and resource management. Simulation of VSAT networks that meets these service requirements will be critical in designing the most elegant and efficient system. We present a simulation study for a VSAT network in this paper. In Section 2, the GEO-based high capacity multimedia VSAT network is introduced, the uplink MF-TDMA scheme is described, and an integrated CAC and BOD algorithm for the MF-TDMA access is presented. Section 3 gives modeling approaches for source traffic and the VSAT network simulation. Section 4 discusses the simulation results. Section 5 presents conclusions of this study. 2. A GEO-Based High Capacity Multimedia VSAT Network 2.1. Network Architecture and Uplink Multiple Access Structure Figure 1 illustrates the network structure of the GEObased high capacity multimedia VSAT network we investigate. This is a full-meshed network. The GEO satellite is bent-pipe. As shown in Figure 1, user links support access to individual users hosts and LANs via VSATs. The user links are MF-TDMA links on the uplink, while they are TDMA on the downlink.

User Links

LAN

FAX

VSAT VSAT

Full-meshed satellite Network up to 8 x 2 Mbps

VSAT VSAT

ATM Network

VSAT
Local Corporate FR Network

VSAT

Master VSAT
PABX

LAN (Ethernet) Network Management System

FAX

Figure 1. The GEO-Based High Capacity Multimedia VSAT Network

The Network Management System is in charge of most of the signaling and management functions in the VSAT network. We suppose there is a Master VSAT in this system, which is one of the VSATs and is selected to perform the functions of the Network Management System, in addition to being able to function as a data and traffic node. We focus our performance study on VSAT terminal uplink access. A mix of frequency and time sharing access (MF-TDMA) is baselined for the satellite uplink, where the VSAT terminals access narrowband carriers of a fixed bandwidth (b kHz) on a time shared basis, with a time slot dimensioned on the basis of one single ATM cell transmission. The VSAT uplink MF-TDMA structure is shown in Figure 2. There are m time slots in a frame of a TDMA channel, and B channels in the MFTDMA scheme. One 48 byte (384 bits) ATM cell every 0.192 milli-second corresponds to a 2 Mbps channel. Minimum user assignment of 1 time slot (cell) per frame gives 2 Kbps access rate. An integrated CAC and BOD algorithm is used for the MF-TDMA uplink access. We will briefly describe the algorithm in the next sub-section. The details of the CAC and BOD method are described in [1]. The CAC and BOD Controller in this system is in the Network Management System. As it is shown in Figure 2, one portion of the first channel is called the control channel and is dedicated to the Master VSAT. The timeslots in the control channel is referred to as control slots. Data traffic from the Master VSAT will be restricted to the timeslots after the control slots. Thus, the Master VSAT will always have less than 2 Mbps available for data transmission.
Channel 1 Timeslot 1 Channel 2 Channel 3 Channel B

2.2. Integrated CAC and BOD Algorithm Each type of application is classified in terms of ATM transfer capabilities, i.e., CBR, rt-VBR, nrt-VBR, ABR, or UBR. CAC on the multiple access uplink will statically allocate to each connection j an amount of resource (SRj) that depends on the connection, traffic descriptor, and requested Quality of Service (QoS). Possible values for SRj could correspond to the following capacity: for CBR--PCRj; for rt-VBR--ERj; for nrt-VBR--SCRj; for ABR--MCRj; and for UBR--0. PCR is peak cell rate, SCR is sustainable cell rate, and MCR is minimum cell rate, as specified by ATM Forum and ITU-T specifications. ER is effective rate or effective bandwidth [2]. CAC on the multiple access uplink books for connection j additional resources BRj, in addition to the resource allocated statically (SRj). Unlike the statically allocated resources case, the connection does not immediately claim the usage of the booked resources. Rather, it is the BOD process that grants the booked resources based on the connections requests. If a connection does not ask for its booked time slots, then the time slots become available for assignment on best effort basis. The amount of booked resources BRj depends on the connection type, traffic descriptor, and requested QoS for j. BRj could correspond to the following capacity: for CBR--0 (no BOD for CBR); for rt-VBR--0 (no BOD for rt-VBR); for nrt-VBR--ERj SCRj; for ABR--0 (BOD for ABR uses best effort); and for UBR--0 (BOD for UBR uses best effort). CAC on the multiple access segment grants admission to connection j only if the sum of what is to be statically allocated to j (i.e., SRj) and what is to be booked for j (i.e., BRj) is less than the total capacity of the multiple access uplink, minus the sum of the already allocated capacity, and minus the sum of the booked capacity for all ongoing calls k on the link, i.e., only if:

Control Slots

Timeslot n Timeslot n+1

Timeslot m

SR + BR + SR + BR C
j j k k k k

(1)

Figure 2. VSAT Uplink MF-TDMA Structure

where CT is the total capacity of the multiple access uplink. In addition to condition (1), because each VSAT terminal is equipped with only one antenna, we have a second constraint for the CAC:

A VSAT is not allowed to transmit simultaneously on two different frequencies with the same timeslot (2) Note that SR, BR, and CT can either be capacity or time slots. For the simulation study, we use SR, BR, and CT as time slots. For the MF-TDMA scheme described in Figure 2, CT = m B n time slots, where n is the number of control slots. In the rest of this paper, we use the term SATS (statically allocated time slots) and BATS (booked allocated time slots). We specify the number of SATS and BATS for each type of source traffic we use in Figure 4. For VSAT connections, after connection j is accepted by the uplink of the network, and depending on its type, it will use the BOD protocol to request resources in addition to what has been statically allocated to it (i.e., in addition to SRj) on a need basis. We assume in the following that the BOD is performed for each connection individually. An existing connection j entitled to use BOD can make BOD requests at any time. As it is shown in Figure 3, the BOD process consists of five phases: 1) BOD Computation of Need Phase; 2) BOD Request Signaling Phase; 3) BOD Controller Computation Phase; 4) BOD Response Signaling Phase; and 5) VSAT Slot Assignment Phase. Periodically (e.g., in the rest of this paper, every 0.192m milli-seconds), the Master VSAT broadcasts via satellite a Burst Time Plan (BTP) table to all the VSATs in the system. The BTP table contains a list of all time slots that can be used for the source connections for all VSATs.

3. Simulation Modeling Approach In this section, we present the modeling approach for source traffic and for the VSAT network simulation. 3.1. Source Traffic Modeling

For our study, many different applications were identified for possible GEO-based VSAT services. A two level traffic model is used: Session Level and Burst/Cell Level. Session Level is modeled as a Poisson process, i.e., session inter-arrival is an exponential distribution. Session duration is also modeled as an exponential distribution. For many applications, Burst/Cell Level is modeled inside each session as an ON/OFF 2-state discrete-time Markov model. At each ON state, the actual cell emission pattern is specified. A detailed multimedia traffic study has been done for the satellite system [3]. Using the method described in [3], a list of 13 most dominant applications are identified for the satellite system and modeled in the VSAT network simulation here. Each of these applications is also identified in terms of ATM transfer capabilities: CBR; rt-VBR; nrt-VBR; ABR; and UBR. Figure 4 shows the list of 13 types of applications and the statically allocated and booked resources used in the simulation study.
ATM Applications Type 1: Voice Type 2: Custom Calling Services Type 3: Fax Type 4: Video Type 5: Data Dissemination Type 6: Web Access Type 7: Telnet Type 8: File Transfer Type 9: Email - Text Type 10: Email - Image Type 11: Email - File Attachment Transfer Capability CBR CBR rt-VBR rt-VBR nrt-VBR ABR ABR UBR UBR UBR UBR UBR UBR 32 32 31 142 4 1 1 0 0 0 0 0 0 0 0 0 0 28 0 0 0 0 0 0 0 0 SATS BATS Peak Uplink Bandwidth (Kbps) 64 64 64 384 64 64 64 64 64 384 64 64 384

u eq

es

es

po

e ns

Type 12: Email - Audio Attachment Type 13: Email - Video Clip Attachment 2 4

Connection j

accept reject

CBR rt-VBR nrt-VBR ABR UBR Queue 1

Figure 4. Traffic Parameters for All Applications

1 CAC Queue 2

5
VSAT

BOD

3 Queue 3 Master VSAT

3.2. VSAT Network Simulation Modeling Our VSAT network simulation model consists of the following steps and is illustrated in Figure 5.

Figure 3. Integrated CAC and BOD Method

Step 1: A two-level traffic generator is used in the simulation: a Session Level Traffic Generator and a Burst/Cell Level Traffic Generator, as shown in Figure 6. The Burst/Cell Level Traffic Generator will be used in Step 3. The Session Level Traffic Generator is used for the CAC simulation. For each session generated, a CAC request is queued in Queue 2 of Figure 4. Step 2: Run CAC algorithm and update BTP table periodically. Step 3: A Burst/Cell Level Traffic Generator then generates traffic for every call connection j that is accepted by CAC. A queue (Queue 1 as described in Figure 3) is maintained in the VSAT to keep count of all the cells that the traffic generator is generating. The cells generated are immediately put in the corresponding sub-queue in Queue 1.
Session Level Traffic Generator Step 1 & 2

Step 4: A BOD Computation of Needs process is invoked immediately for all cells in the sub-queues that do not have assigned time slots. The BOD computation of needs is done when cells arrive from the traffic generator. Step 5: In response to all the BOD requests processed by the BOD controller in the previous time interval, the BTP table is broadcast to VSATs via the BOD Response Signaling Phase. BOD requests that arrive too late to be processed within a time interval are queued in Queue 3 of Figure 3 to be processed in the next time interval. Step 6: When the VSAT receives the BTP, a VSAT Slot Assignment process is started. Based on the BTP, the cells are removed from the queues based on connection identifiers in the BTP. The slot assignments in the VSAT are done on a VC-basis. 4. Results and Discussions The simulation is implemented in OPNET [4]. Extensive simulation experiments have been done for various scenarios of traffic mixes. In the following, we show part of the results for a spot beam in a given region in a future year at a normal busy hour of a working day. Various performance measures can be collected from the simulation runs. In the following, we only show part of the results of a simulation experiment: uplink throughput, SIT queuing delay, and end-to-end queuing delay. Figure 7 shows the uplink throughput which is measured every 0.192m milli-seconds as the number of time slots that are actually used divided by total number of time slots in a BTP table (mB-n slots). As shown in Figure 7, the steady state BOD throughput is about 74.27%. Figure 7 shows that the remaining unused 25.73% of all bandwidth is completely due to the unused SATS which are statically reserved for CBR, VBR, and ABR connections. In order to guarantee QoS for a connection, the reserved SATS for a connection cannot be released to the usage of other connections in the duration of this connection. Therefore, the BOD scheme is able to efficiently utilize all available bandwidth and gain the highest throughput.

run CAC algorithm update BTP table

Burst/Cell Level Traffic Generator Step 3 Queue 1

BOD Computation of Need & Request Signal Queue 3

Step 4

Step 5

BOD Controller Computation & Response Signal collect delay and jitter statistics

Step 6

Figure 5. VSAT Network Simulation Model

Call Request

SESSION LEVEL GENERATOR

CAC CONTROLLER

Accept

Reject

BURST/CELL LEVEL GENERATOR Discard

Traffic

VSAT Queue

Figure 6. Traffic Generator

Figure 8 shows average, minimum, and maximum queuing delays and end-to-end delays for each type of traffic. The queuing delay is the time that an ATM cell waits in Queue 1 (Figure 3) in the VSAT for uplink access. The end-to-end delay is the interval between the time an ATM cell is generated in the source VSAT and the time it arrives to the destination VSAT.

5. Conclusions Satellite is an attractive vehicle for the transport of multimedia and broadband services. Satellites can be integrated in the core of wide-area networks, in broadband access, or in enterprise and private networks. Satellites enable the network operator to customize the network to the transport of specific mixes of applications, simplify congestion control and network management, allow direct user-to-user interaction, and provide superior security. They also have the advantage of high scalability and ease of deployment. Also, GEO satellites can transport traffic from long distances to the gateways with uniform delay that is independent of terrestrial distances. In order to get a better understanding of broadband multimedia satellite network design and network planning, performance evaluation techniques need to be developed for the satellite networks. This paper presents a framework of a simulation model to evaluate the performance of a GEO-based high capacity multimedia VSAT satellite network. Based on the simulation results, several issues can be effectively addressed. For example, the system designers can determine how to mix different applications to obtain maximum capacity utilization and network revenue, guarantee the QoS for all applications with limited network resources, and implement buffers of sufficient size to reduce loss. 6. References [1]. Catherine Rosenberg, End-to-End Resource Management for ATM On-Board Processor Geostationary Satellite Systems, Proceedings of 4th Ka Band Utilization Conference, Venice, Italy, pp. 481488, November 1998 [2]. K.W. Ross, Multiservice Loss Models for Broadband Telecommunication Networks, SpringerVerlag, London, 1995 [3]. Jeff Babbitt, Yi Qian, and Hosame Abu-Amara, Global Traffic Generation, Modeling, and Characterization Methodology, Proceedings of 8th International Telecommunication Network Planning Symposium, Sorrento, Italy, pp. 321-326, October 1998 [4]. MIL 3, Inc., OPNET Manuals, 1997

Figure 7. Uplink Throughput The maximum end-to-end delay for the CBR application is in the order of 400 milli-seconds. The mean end-to-end delay for VBR applications is between 300 and 800 milli-seconds, and mean end-to-end delay for ABR and most UBR services is in the order of seconds. The VSAT system simulation model we developed here can be used as a powerful tool to experiment with different types of traffic mixes and the CAC/BOD reservation schemes so that it can guarantee QoS for all applications.
Applications

Average Minimum Maximum Average Minimum Maximum Queuing Delay Queuing Delay Queuing Delay ETE Delay ETE Delay ETE Delay (Seconds) (Seconds) 0.186 0.186 0.000 0.045 0.002 1.897 3.905 0.500 0.500 0.500 0.500 0.500 0.500 (Seconds) 0.186 0.186 0.281 0.097 0.557 7.947 17.477 12.104 43.509 40.081 175.067 30.497 53.029 (Seconds) 0.436 0.436 0.296 0.326 0.727 2.791 5.058 2.611 8.498 14.581 48.192 9.389 14.461 (Seconds) 0.436 0.436 0.250 0.295 0.252 2.147 4.155 0.750 0.750 0.750 0.750 0.750 0.750 (Seconds) 0.436 0.436 0.531 0.347 0.807 8.197 17.727 12.354 43.759 40.331 175.317 30.747 53.279 0.186 0.186 0.046 0.076 0.477 2.541 4.808 2.361 8.248 14.331 47.942 9.139 14.211

Type 1: Voice, high-quality Type 2: Custom Calling Services Type 3: Fax Type 4: Video Type 5: Data Dissemination Type 6: Web Access Type 7: Telnet Type 8: File Transfer Type 9: Email - Text Type 10: Email - Image Type 11: Email - File Attachment Type 12: Email - Audio Attachment Type 13: Email - Video Attachment

Figure 8. Delay for Each Type of Traffic