In Chapter3, we presented a description of Distributed Relative/Absolute Fair Throughput with Delay Support (DRAFT+D) protocol that provides support for relative throughput, absolute throughput, and absolute delay with a distributed safeguard mechanism. In this chapter, we will present a baseline performance evaluation of DRAFT+D for providing QoS support. The important findings are summarized as follows. Fair bandwidth allocation according to weight can be achieved with very low variation of throughput and delay. Specific bandwidth allocation is also achieved by absolute throughput support via the introduction of ω and θ parameters with very low variation. The simulation results confirm that the analysis in Section3.4.3of fair share and overload condition are correct. Specific delay support for the Head-of-Queue (HoQ) packet is also achieved via absolute delay support. The mechanism that limits excessive traffic from new RT-MSs can be accomplished in a distributed manner without requiring exchange of any information among MSs. Finally, the results confirm the effectiveness and robustness of DRAFT+D in an extremely saturated condition (10 Mbps in 2 Mbps WLAN).
To demonstrate the performance of DRAFT+D, we conducted our simulations using OPNET 9.1. The model reused the 802.11 DCF model available in OPNET. Unless otherwise specified, the following assumptions and parameters are used:
• Each MS works independently and cooperatively according to the provided specification. We assume that no MS attempts to cheat or gain illegal advantages over the other MSs. • MSs operate in Ad-Hoc mode which is sufficient and appropriate to evaluate the perfor- mance of the MAC mechanism. In Chapter 5, we will relax this assumption and show
the result in infrastructure mode where an AP is used to relay traffic among MSs. If there is the presence of AP in the network, the AP will use the same set of rules as DRAFT+D. Additionally, the AP is also responsible for relaying traffic among MSs or between WLANs and wired-networks.
• In this simulation, each MS has only one class of traffic per MS, however, our mechanism works for many classes of traffic per MS as well.
• All MSs are located within a single Independent Basic Service Set (IBSS) where every MS is able to detect a transmission from other MSs. We assume that hidden-terminals are not present and the Request to Send/Clear to Send (RTS/CTS) mechanism is not used. In Chapter 5, we will relax this assumption and show the results when RTS/CTS is used.
• Time variant behavior of the wireless channel is not considered. MSs are assumed to operate in a WLAN where the characteristics of the wireless channel is static. The channel is also assumed error-free. In Chapter 5, we will relax this assumption and evaluate the performance in erroneous channels.
• We perform most of the simulations on a 2 Mbps WLAN to strategically minimize the simulation times into manageable durations. In Chapter5, we will relax this assumption and evaluate the performance in 5.5 and 11 Mbps WLANs.
• We assume that all MSs are connected to the network at the same data rate. Later, we will relax this assumption and consider multi-rate environments. However, multi-rate is considered only in the static sense. By this, we mean each MS can connect to the WLAN at different data rates. However, once an MS is connected, the data rate does not change.
• MSs can be located anywhere within the IBSS. However, they remain at the particular location during the course of simulation.
• The traffic from each MS is sent to random destinations.
• All flows are of constant bit rate. The constant bit rate traffic is to clearly demonstrate the variation of throughput and delay (or the lack of) created by the mechanism. We will relax this assumption in Chapter 5.
• The frame size is fixed at 1000 bytes. Later, we will relax this assumption and consider variable length packets.
• Buffer size is set to 256,000 bits, IF S = DIF S, κ = 5, ω = 5.
During the course of performance evaluation, the following metrics and their measure- ments are used:
• Throughput: Throughput requirement and quantum rate refers to the offered load of input traffic. The experienced throughput denotes the rate of successful data transmis- sion over time.
• Delay: The HoQ delay denotes the waiting time since a packet becomes head-of-queue until the reception of acknowledgement. The HoQ delay comprises the durations of packet transmission, packet retransmission (if any), acknowledgement of transmissions, and waiting times. The queuing delay is the duration for which the packet is waiting in the queue before it becomes the head of the queue. The MAC delay refers to the period between the time when a packet is received into the queue and the time when an acknowledgement packet is received to confirm a successful transmission.
• Jitter: We consider jitter only in the MAC layer. This metric measures the difference between the previous HoQ delay and the current HoQ delay.
• Collision Rate: Collision occurs when the backoff timer of two or more MSs reach zero at the same time. Collision rate is calculated from the number of collisions per second.
• Aggregate Throughput: The aggregate throughput is the sum of throughput from all MSs
• Backoff Interval: The backoff interval is recorded from the actual selected values of BI for each packet.
In the next section, we will present an overview of the results and follow with sensitivity analysis for important parameters of DRAFT+D, i.e., κ, ω, θ, N1, N2, β and Ξ.