Providing Quality of Service over General Packet Radio Service: Admission Control, Radio Resource Reservation, and Scheduling

Ana Cristina B. Kochem, Eliane L. Bodanese

Programa de Ps-graduao em Engenharia Eltrica e Informtica Industrial Centro Federal de Educao Tecnolgica do Paran (CEFET-PR) Avenida Sete de Setembro, 3165 80230-901- Curitiba PR Brasil {cristina,bodanese}@dainf.cefetpr.br

Abstract. Mobile wireless networks aim to support different applications, especially those based on voice, video, and data. However, these applications differ a lot in terms of Quality of Service (QoS) requirements. We propose a combination of three different QoS mechanisms over General Packet Radio Service (GPRS) network elements, in order to manage efficiently the scarce frequency spectrum. The first mechanism is a Call Admission Control (CAC) where connections are admitted only if the network is capable of providing the required resources. The second is a Radio Resource Reservation mechanism that ensures bandwidth to ongoing high priority applications. Finally, a scheduling mechanism provides a more regular behaviour to services.

1. Introduction The explosive growth and innovative potential of wireless communication technologies have been causing a revolution on the communication market. Wireless networks have aimed to integrate voice, video, and data over the same network infrastructure due to user demand for different kind of services. However, the traffic generated by these media is highly diverse and imposes specific QoS requirements in terms of delay and bandwidth. To provide such requirements, mobile wireless networks need to deal with user mobility, scarce bandwidth, and interference. This paper proposes the implementation of admission control, resource reservation, and scheduling mechanisms over the GPRS air interface, in order to prevent congestion and provide better quality of service to GPRS subscribers. The paper is organised as follows: in the next section, the General Packet Radio Service is presented. Section 3 defines the term QoS in wireless networks. Section 4 describes the three proposed QoS mechanisms for GPRS networks. The simulation model is presented in Section 5. Simulation results are shown in Section 6. Conclusions and future work are described in Section 7.

2. General Packet Radio Service GPRS, defined by the European Telecommunications Standard Institute (ETSI), is a packet switched oriented service proposed for Global System for Mobile Communications (GSM) 2.5 generation [1]. GPRS and GSM networks make use of the same infrastructure and share the same radio resources. However, two new elements were added in GPRS in order to provide an end-to-end packet transfer: the Serving GPRS Support Node (SGSN) and the Gateway GPRS Support Node (GGSN). The SGSN routes packets to/from terminals within its service area and it is responsible for the mobile subscribers location, mobility management, link management, security, authentication, etc. The GGSN acts as a logical interface to external Packet Data Networks (PDNs) such as Internet, Intranet, and X.25 networks. Some of its functions include packet format translation, traffic classification, and address mapping (making possible the communication between different networks). The GPRS elements in the radio interface include the Mobile Station (MS), the Base Transceiver Station (BTS), and the Base Station Controller (BSC). The BTS and BSC together compose the Base Station Subsystem (BSS). 3. Quality of Service in Wireless Networks Quality of Service (QoS) is a broadband term used to describe the overall experience a user or an application will receive over a network [2]. The main research areas in QoS look for solutions that will allow networks to meet the performance requirements of different applications in terms of bandwidth, delay, network availability, and packet transfer reliability. To offer QoS for GPRS subscribers, ETSI has specified four QoS classes [3]: (1) Delay: defines values for end-to-end packet transfer between two entities; (2) Reliability: represents maximum values accepted by an application in terms of loss, duplicated, and out-of-sequence packet probability; (3) Service Precedence: indicates three levels of priority for a service: 1 (high), 2 (normal), and 3 (low). In case of congestion, a service with higher priority will receive a better treatment; (4) Throughput: defines the peak throughput in octets per second and the mean throughput in octets per hour at which packets are expected to be transferred across the network. Nine classes are defined for the peak rate, ranging from 8 kbit/s to 2048 kbit/s and nineteen classes define different mean data rates ranging from 100 octets per hour (~0.22

bit/s) to 50 Mega octetos per hour (~111 kbit/s). Table 1 describes some QoS requirements for a group of applications according to ETSI [4]. During a call establishment, the MS and the GPRS core network need to negotiate a QoS profile for the service being requested. The negotiated profile is stored in a Packet Data Protocol (PDP) context along with other information used to perform packet routing. Once the QoS profile is accepted, the GPRS network must provide the negotiated resources to the MS in order to satisfy the QoS requirements.
Table 1. QoS requirements for a group of applications Application Audio Streaming E-mail File Transfer (FTP) Telnet Video Streaming Voice Messaging Data Rate (kbit/s) 5-128 < 384 <1 20-384 4-13 Delay Variation < 2s < 2s < 1 ms Delay (start-up) < 10s Best effort (start-up) < 10s (end-to-end one way) < 250 ms (start-up) < 10s (end-to-end one way) < 1s

4. QoS Provisioning Mechanisms Although ETSI has specified four QoS classes for GPRS, there are no specifications about QoS management mechanisms, therefore this subject has received a special attention from researchers. In [5], the authors have examined the improvement in capacity and performance gain through QoS management in GPRS. In [6] the QoS control in GPRS is separated in two phases to achieve the specified performance requirements of a service subscription contract. The first phase includes the CAC and the resource reservation. The second phase applies scheduling and policing to packets in the BSS to regulate the traffic and ensure the required QoS during a packet transfer. In [7,8] scheduling disciplines applied to GPRS network were evaluated. However, those studies did not consider the use of mechanisms to meet the QoS requirements imposed by voice, video, and data services. 4.1 Call Admission Control in GPRS The SGSN and GGSN nodes are responsible to perform the Call Admission Control (CAC) mechanism during the QoS profile negotiation between the subscriber and the network [5]. In respect to the radio resources, the purpose of the CAC is to limit the number of connections in such a way that, once admitted, the probability of this connection encountering a congestion state is limited to a pre-defined level [9].

We have applied a connection setup algorithm based on [10] to achieve the required QoS for a set of applications. The algorithm performs CAC and resource reservation (described in next section) based on local and remote information, i.e., from the cell where the MS originates the connection request and from adjacent cells, respectively. Based on that information it is possible to determine whether to accept or reject an incoming traffic to be carried in high-speed wireless networks. We have adapted that algorithm to be used in a GPRS network. The CAC procedure applied in the GPRS nodes is summarised below.
IF (Precedence Class 1 OR Precedence Class 2) THEN // Normal or high priority application IF (new connection) THEN IF (mean_throughput available bandwidth in the current cell) THEN Reservation() // Tries to reserve the radio resources for the new connection ELSE // There is not enough bandwidth in the cell Reject new connection IF (handoff request) THEN IF (mean_throughput available or reserved bandwidth in cell) THEN Reservation() // Tries to reserve the radio resources for the handoff request ELSE // There is not enough bandwidth in the cell Reject handoff request ELSE // Precedence Class 3 - Low priority application IF (available bandwidth in cell > 0) THEN Accept connection Allocate the minimum {requested, available resources in cell} ELSE // no resources available Reject connection

When an MS tries to establish a new connection or requests a handoff, performing a PDP context activation, it provides its service precedence class and its mean throughput class, both defined in the QoS profile being negotiated. Then, the CAC is performed by the SGSN and the GGSN by checking the mean throughput class that defines the data rate or required bandwidth of the MS application. If there is enough bandwidth in the cell, the QoS profile will be accepted and the BSS will be responsible to reserve the resources to the MS depending on the mean throughput negotiated and the MS multislot class. The MS multislot class refers to the maximum number of time slots allowed for a downlink and uplink transfer [11]. High and normal priority connection requests are accepted only if the cell where the MS is located or moving into can provide the mean throughput required and, in case of a new connection, if the BSS is able to reserve resources in all neighbouring cells following a reservation policy described later. So, when an MS moves to a new cell, in a handoff procedure, it may have its resources already reserved. Low priority connection requests will always be accepted as long as there is some available bandwidth in the cell. Once the connection is accepted, the available

bandwidth in the cell, if it is less than the required one, will be allocated and no resource reservation will be made in neighbouring cells for low priority services. 4.2 Resource Reservation During CAC, the reservation() function is performed every time an MS requests a new connection or moves to a new cell. The MS tries to establish a Temporary Block Flow (TBF) when it has a packet to send/receive by sending a message to the BSS requesting the radio resources, i.e., one to eight physical channels (or time slots). The reservation() function algorithm is described below.
Reservation() // Performed every time an MS requests a new connection or moves to a new cell Try to reserve the radio resource(s) in all adjacent cells of the current cell IF (successful reservation OR handoff request) THEN Accept connection Establish TBF between the MS and the BSS to allocate the requested resources in the current cell ELSE // it is not possible to reserve the resources for new connection request*/ Reject new connection

Differently of the algorithm in [10], the handoff request will not be rejected if the resource reservation in adjacent cells cannot be performed. That modification was made to avoid blocking of ongoing calls. We assume the network can follow the MSs movement pattern and provide the necessary information for the prediction of its more probable direction of handoff. In [10], the authors propose that the users movement pattern can be expressed as a probability px in which an MS will perform a handoff to cell x. (1 - px) represents the probability that the MS will move to other neighbouring cell. To determine the amount of bandwidth to reserve in neighbouring cells, the algorithm requests to cell x the largest requested bandwidth of precedence class 1 or 2. It requests to the other neighbouring cells the largest bandwidth requested * (1-px). A parameter called maxb is used to indicate the maximum percentage of the cell bandwidth that can be reserved. So, if the bandwidth to be reserved is greater than maxb, only maxb will be reserved. The allocated resources in the cell are released when a connection is terminated or when an MS executes a handoff. In the latter case, the reserved resources in the neighbouring cells, which after handoff are no longer adjacent cells of the new cell, are also released. 4.3 Scheduling Once a TBF is established and the required resources are allocated, the second phase of the QoS provisioning is applied involving the scheduling and policing of the packets in the BSS to make the decision of which packets must be served, regulate the traffic and

ensure the negotiated QoS profile. When packets arrive in the BSS, they are distributed into one or more queues according, for example, to their service precedence class. In the queues, these packets are scheduled following a queue scheduling discipline. We have applied the queue scheduling discipline in the Medium Access Control (MAC) layer. In the MAC layer, a MAC header is added to each Radio Link Control (RLC) data block, forming a RLC/MAC block (also called a radio block). The radio blocks are then formatted into the physical layer for transmission. Each radio block is composed of four normal bursts in consecutive Time Division Multiple Access (TDMA) frames [11]. The Deficit Round Robin (DRR) and the DRR+ queue scheduling disciplines concepts [12] were used to schedule the radio blocks into the physical layer. We have modified these scheduling disciplines to be consistent with the GPRS functionality and to supply the QoS requirements of an application set. This modification was called MDRR+ (Modified DRR+) and its functions are: 1. Classification: the scheduler distributes the radio blocks to one queue dedicated exclusively to the flow to which the blocks belongs to based on the flow identification number (Fid) stored in the packet IP header; 2. Policy: a flow can increase its rate reducing the bandwidth portion of other flows. To solve this problem, the scheduler limits the data rate of each flow according to its negotiated mean throughput class; 3. Scheduling: the scheduler determines which radio blocks can be scheduled in the radio interface according to the negotiated service precedence class. There is one queue dedicated to each media flow. The packets are distributed into the queues according to their Fid. The service precedence class (priority) of each flow is used by the scheduler, to determine which flow has to be served first. Some array variables are initialised for each flow: 1. Quantum[Fid]: defines the portion of the bandwidth to be allocated to each flow (calculated by the mean throughput class value and converted to bits per second); 2. Priority[Fid]: determines the service precedence class of each flow; 3. DeficitCounter[Fid]: specifies the maximum number of bytes that each queue can transmit when it is visited by the scheduler. This variable is initialised with the Quantum[Fid] value and, each second, it is reloaded in order to compensate the queues whose packets could not be transmitted during a scheduler service cycle;

4. QuantumTime[Fid]: a timer that indicates when the DeficitCounter[Fid] must be reloaded with the quantum value; 5. InactiveTime[Fid]: a timer used to indicate when a flow is inactive for more than one second; 6. NslotsAloc[Fid]: indicates the number of time slots each flow can allocate (calculated by dividing the Quantum[Fid] by the data bit rate of the slots); 7. Reload[Fid] : indicates when DeficitCounter[Fid] must be reloaded (initially, set to 0). The timers QuantumTime[Fid] and InactiveTime[Fid] are initialized with 0. The value of their timeout is 217 to indicate the number of TDMA frames (0.004616s) the scheduler has to scan to complete one second. Those timers are added to the default DRR scheduling discipline to limit the data rate per second of each active flow. Initially, the algorithm tries to allocate NSlotsAloc[Fid] time slots to each flow. Next, the function Enqueue() is performed to place the new incoming packets into a queue dedicated exclusively to one flow (Fid). The parameters of each flow are set up at the beginning of each TDMA frame. If QuantumTime[Fid] has expired, the Reload[Fid] variable is set to 1 to indicate that the DeficitCounter[Fid] must be incremented with the quantum defined for the flow. Then, the QuantumTime[Fid] is set to 0. Otherwise, if the QuantumTime[Fid] has not expired yet, it is incremented by one. The next test, verifies if the flow has been transmitting packets. If so, this function is resumed. If it is active (DeficitCounter[Fid] ! =-1) but it is not transmitting any packets (Queue[Fid] == NULL) for more than one second (InactiveTime[Fid] >= 217), the flow is considered inactive. If the inactive time has not expired yet, it is incremented. Finally, the function Dequeue() visits each service precedence class looking for flows within it. Then, the algorithm verifies if the queue has packets to send. If the queue is not empty, the InactiveTime[Fid] is set to 0. If the timer QuantumTime[Fid] has expired, the DeficitCounter[Fid] is incremented by Quantum[Fid]. If the flow was inactive, DeficitCounter[Fid] is set with the value of the quantum. Then, the packets of the flow are transmitted as long as their size do not exceed the DeficitCounter[Fid] and if there is some slot allocated to the flow. When a packet is sent, the variable DeficitCounter[Fid] is reduced by the number of bytes in the packet. If the size of the packet exceeds DeficitCounter[Fid] or when the queue is empty, the scheduler will visit the next queue. A variable called stop is used when there is no time slot available and the flow could not

allocate its NSlotsAloc[Fid] in order to control the priority of the flows using the available time slots, 5. Simulation Model Figure 1 shows the simulation model of the GPRS network. In order to simulate the bandwidth reservation in adjacent cells and the system mobility (handoff request rate), the network is composed of a cluster of seven cells. The cells can exchange bandwidth information with each other. The cell model used for the evaluation of the scheduling disciplines (MDRR+ and FIFO) is composed of one GGSN, one SGSN, one BSC, one BTS, and six MSs. Only downlink traffic is being considered, therefore the GGSN represents the entry point of the traffic sources. The GGSN and the SGSN are responsible for the QoS profile negotiation along with the MS during the PDP context activation. The SGSN and the GGSN run the CAC algorithm when requests for new or handoff connections arrive. The BSS allocate radio resources for the cell and reserve resources in the adjacent cells if needed.

. . .

B
B T S

1 0 S

it / s
G S

0 0

i t / s
G

Figure 1 GPRS simulation model

As the air interface is considered the system bottleneck, we have applied a link of 100 Mbps between the fixed nodes. Delay on fixed links was not considered. In the air interface, there are six Packet Data Channels (PDCH) for GPRS download traffic (from a GGSN to an MS). The link capacity of each PDCH is 21.4 kbit/s (CS-4 - Coding Scheme 4) [11]. Consequently, the cell has a total capacity of 128.4 kbit/s.

We have used the software NS [13] to simulate our GPRS network and we have modified the code contributed in [14] to be able to have more than one MS transmitting packet fragments per frame TDMA. We have also modified the MAC layer for multislot operation, in order to support traffic with data rate higher than 21.4 kbit/s. In the chosen scenario, all MSs belong to multislot class 8 (4 time slots for reception and 1 time slot for transmission). Six traffic sources were simulated according to Table 2. We have used the FTP and Telnet applications already available in NS. The Finnish University and Research Network (FUNET) model was used to generate e-mail applications with inter-arrival times following an exponential distribution and the packet sizes following a Cauchy distribution (with maximum packet size of 10 kbytes [8]). Finally the video traffic was taken from a file adapted to NS [15] which generates H263 frames from the movie Jurassic Park I. In this paper, we have assumed a communication link free of interference and buffers with infinite capacity.
Service precedence class (priority) 2 (normal) 3 (low) 3 (low) 1 (high) 2 (normal) 1 (high) Table 2. Traffic source parameters Average connection Average bit Inter arrival holding time (min rate (kbit/s) times (s) max) 3 minutes 20 (CBR) 0.08 (constant) (1min max) 30 seconds 4 1 (exponential) (10s120s) 2 minutes < 110 (30s max) 3 minutes 0.512 1 (exponential) (30s max) 6 minutes 64 0.04 (25 frames (3min max) per second) 3 minutes 10.7 (CBR) 0.02308 (1min max) (constant) Average packet/ frame size (bytes) 200 (constant) 500 (Cauchy) 500 64 320 31 (constant)

Application Audio Streaming E-mail FTP Telnet Video Streaming Voice Messaging

6. Simulation Results This section presents the results obtained by the simulation of the proposed QoS mechanisms for GPRS. In the first scenario, each cell of the cluster has a maximum bandwidth of 428 kbit/s and a maxb of 15% is assumed. Figure 2(a) shows the new connection rejection rate versus the load of a cell measured in calls per second. Figure 2(b) shows the handoff request rejection rate versus the load in the cell. Figure 2 shows, as it was expected, that handoff requests are being favoured in relation to new connection due to the bandwidth reservation made for them.

(a)

(b)

Figure 2. New connection rejection rate (a) and handoff request rejection rate (b)

Another scenario was simulated in order to compare the performance of the GPRS network using a MDRR+ queuing discipline and then using a FIFO queuing discipline, both in the MAC GPRS layer in order to schedule the received packets at the BSS into the air interface in the direction of the MSs. The MDRR+ discipline (described in Section 4) schedules the packets according to their priority. Its main objective is to provide a fair distribution of bandwidth among all active flows in the cell. On the other hand, the FIFO discipline schedules the packets in the same order they have arrived at the BSS. Table 3 shows the simulation results. After 10 seconds of simulation a video application is admitted in the network. During CAC, the video application negotiate a throughput class of 111 kbit/s. However, the MSs can only receive up to 4 PDCHs (85.6 kbit/s) because of the multislot class they belong to. Therefore only 4 PDCHs are allocated for the video application. Telnet and e-mail applications are also later admitted, respectively. Both demand low data rate and they can share the same packet data channel. At 25 seconds of simulation, a connection request from a FTP application arrives. Initially, it tries to request a mean throughput of 111kbit/s (class 19), however at this time there is only one PDCH available in the cell. After negotiation, the available PDCH is allocated for the FTP application, because it is not delay sensitive and it is a low priority service. In this condition, the FTP application can reach a maximum data rate of 21.4 kbit/s. Unfortunately, some traffic sources can misbehave causing packet dropping and delay to other traffic.
Application Connection duration Negotiated Table 3. Simulation Results Throughput Maximum m Minimumpacket delay Maximum delay variation (s)

e a maximum (bit/s) n t h r o u MDRR+ g h p u t Audio Streaming E-mail FTP Telnet Video Streaming Voice Messaging 90 seconds (75s165s) 25 seconds (22s47s) 45 seconds (25s70s) 1 minute (20s80s) 3 minutes (10s190s) 2 minutes (50s170s) 17 class (22 kbit/s) 15 class (4.4 kbit/s) 17 class (22 kbit/s) 13 class (1.11 kbit/s) 19 class (111 kbit/s) 16 class (11.1 kbit/s) 19,200 20,800 4,000 4,000 18,400 20,800 512 2,048 51,024 81,848 10,713 10,763

FIFO

MDRR+

FIFO

MDRR+

FIFO

19,520 20,480 4,000 20,000 40,000 84,000 512 8,192 39,432 83,913 6,200 14,384

0.21464 5.954661 4.208082 0.050078 0.329245 0.01772

0.38701 1.440183 1.919670 1.393830 8.804382 1.109951

0.053864 1.874630 0.246330 0.018693 0.160291 0.00018

0.06078 1.118404 0.096900 0.889596 0.342070 0.41544

Analysing the results in Table 3, we can see the FIFO scheduler cannot provide the delay requirements of real time applications. For example, the delay variation of the voice messaging service is 415ms, which is much higher than the required 1ms. The endto-end one way delay of the telnet service is around 1.39s, which is also higher than the limit of 250ms. The proposed MDRR+ scheduler controls better the distribution of the bandwidth and comply with the requirements for voice messaging and telnet services. The MDRR+ scheduler brought another advantage, it provided better delay requirements to applications with medium priority, for example, the video application had an start-up delay of approximately 0.33s when MDRR+ was used and around 8.8s when the FIFO scheduler was used. Figure 3(a) shows the mean throughput at the air interface using FIFO scheduling and Figure 3(b) using MDRR+ scheduling. It can be seen (Figure 3(a)) that a misbehaved flow like the FTP application can steal an arbitrary portion of bandwidth of the cell compromising the delay requirements of other applications. The solution is to isolate the effects of misbehaved flows, the MDRR+ scheduler is capable of that, controlling the radio resources and distributing the available bandwidth in a fairer way. Figure 3(b)

clearly shows the distribution of bandwidth according to the negotiated QoS profile of each application. In this way, bursty flows like FTP have a much more regular behaviour.

(a) (b) Figure 3. Wireless throughput using FIFO (a) and Wireless throughput using MDRR+ (b)

7. Conclusions and Future Work This paper has investigated the performance of a QoS provisioning scheme for GPRS networks. Simulation results have shown that the proposed scheme composed of a call admission control (CAC) mechanism, bandwidth reservation, and a packet scheduler following the MDRR+ discipline can become a differential factor in providing quality of service to GPRS users. With the CAC, the connections are only admitted when the network is capable to provide the required radio resources. The MDDR+ scheduler forces applications to have a more regular behaviour and misbehaved flows are not allowed to exceed their negotiated portion of bandwidth. From the results of these first investigations, we could observe that a GPRS network using the proposed QoS provisioning scheme can become more capable of supporting applications with different QoS requirements such as data (Telnet, e-mail, and FTP), audio (voice messaging and audio streaming), and video (video streaming). As future work, further investigations are needed using a more powerful network simulator and a network with a more realist size of cells. The performance of the CAC mechanism and the bandwidth reservation can be optimised. A better investigation on

the effects of the maxb parameter over the rejection rate of new connections and handoff requests is also needed, in order to find a value that keeps a better trade off between their rejection rates. 8. References
[1] [2] C. Bettstetter, H.-J. Vgel, J. Eberspcher, GSM Phase 2+ General Packet Radio Service GPRS: Architecture, Protocols and Air Interface,IEEE Communications Surveys,Vol. 2, No. 3,1999, pp.2-14. NORTEL NETWORKS. Introduction to Quality of Service (QoS) White Paper, Sep. 2001, http://www.nortelnetworks.com/products/02/bstk/switches/bps/collateral/56058.25_022403.pdf (current July 29, 2003). Digital cellular telecommunications system (Phase 2+); General Packet Radio Service (GPRS); Service description, Stage 2, 3GPP TS 03.60, v. 7.7.0, Release 1998. Universal Mobile Telecommunications System (UMTS); Services & service capabilities, ETSI TS 122 105 (2002-06). 3GPP TS 22.105, v. 5.2.0, Release 5. P. Stuckmann and F. Muller, Quality of Service Management in GPRS Networks, Proc. IEEE International Conference on Networking, Colmar, France, Jul. 2001, pp. 276-285. S. Tsao, Quality of Service Control over GPRS Data Network, Proc. IEEE 52nd Vehicular Technology Conference, Boston, USA, Vol.3, Sept. 2000, pp. 1001-1007. J. Sau and C. Scholefield, Scheduling and Quality of Service in the General Packet Radio Service, Proc. IEEE International Conference on Universal Personal Communications, Italy, Vol.2, Oct. 1998, pp. 1067-1071. Q. Pang, A. Bigloo, V. C. M. Leung et al., Service Scheduling for General Packet Radio Service Classes, Proc. IEEE Wireless Communications and Networking Conference, USA, Vol. 3, Sept. 1999, pp. 1229-1233. M. Naghshineh, A. S. Acampora, QOS Provisioning in Micro-Cellular Networks Supporting Multimedia Traffic, Proc. of IEEE Computer And Communication Societies, USA, Vol. 3, Apr. 1995, pp. 1075-1084. C. Oliveira, J. B. Kim, and T. Suda, Quality-of-Service Guarantee in High-Speed Multimedia Wireless Networks, Proc. IEEE International Conference on Communications, USA, Vol.2, Jun. 1996, pp. 728-734. R. J. Bates, GPRS General Packet Radio Service, McGraw-Hill, USA, 2002. M. Shreedhar, G. Varghese, Efficient Fair Queuing using Deficit Round Robin, Proc. ACM SIGCOMM95, Cambridge, USA, Vol. 25, No. 4, Oct. 1995, pp. 231-242. The Network Simulator ns-2: Documentation. The VINT Project. http://www.isi.edu/nsnam/ns/ns-documentation.html (current Apr. 2002). UC Berkeley. Apr. 2002.

[3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15]

R. Jain, GPRS Simulations using ns-Network Simulator, masters thesis, Indian Institute of Technology Bombai, Dept. Electrical Engineering, India, 2001. F. H. P. Fitzek, M. Reisslein, MPEG-4 and H.263 Video Traces for Network Performance Evaluation, IEEE Network, Vol. 15, No. 6, Nov.-Dec. 2001, pp. 40-54.