Performance Analysis

Our main design goal is to reduce channel access delay while maintaining reasonable overheads. Like in most DAMA protocols, the actual data transmission (in TA and DD slot) is collision-free. All collisions occur during the request phase. We focus on strategies that can efficiently request access without imposing severe overheads on the system. We analyze the performance of our ARCMA protocol by examining the additional feature that we introduced.

By implementing the adaptive RA channel, we improved channel utilization by exploiting unused TA slots. While avoiding the waste of valuable transmission time, this strategy also reduces contention in the RA channel. In the network with active MS, empty request queues are caused by heavy traffic where the RA channel is saturated with transmission requests (causing collisions). Therefore, it makes sense to relieve contention in the RA channel by switching to multipleRA mode. The following downlink channel has to be similarly converted to multiple ACK slots; therefore, no downlink data packets are transmitted. This imposes a single slot delay to the broadcasting of packets (downlink) to the mobiles. This is a small overhead compared to the otherwise idle uplink and downlink channel.

In our implementation, all requests are sent immediately when the channel first switches to multipleRA mode. In this mode, the number of mobile requests in each RA channel is reduced. In normal mode, five MSs have to share a single RA channel. In multipleRA mode, the maximum number of MSs per RA channel is two. Since the probability of packets arriving is assumed to be the same for all MSs, the probability of a collision, in which two or more requests are made in the same time slot, is greater in an RA channel that handles more MSs. Therefore, the probability of collisions in multipleRA mode is also reduced. The number of MSs handled by each channel is reduced by the number of RA minislots R. If the multipleRA mode does not produce any successful requests, MSs retransmit their requests (still in multipleRA mode) using their original qr. This allows each mobile to utilize the BEB algorithm, but in a channel with lesser contention. This enhances our protocol's efficiency by reducing channel access delay (due to collisions).

Table 4.1 shows an example illustrating the number of mobiles handled by each RA channel in normal and multipleRA mode.

CBR traffic is specially handled because of its periodic characteristic that produces benefits in two areas. First, CBR traffic is transmitted with minimum delay as a result of the request-free access and transmission priority. This feature is essential since CBR traffic is delay-sensitive. Second, since no requests are needed for CBR packets (except for initial setup), contention in the RA channel is reduced. Depending on the number of mobiles with CBR traffic, this scheme can produce significant improvement to the overall system performance.

In the initial transmission request for the CBR traffic, the mobile has to send additional data representing the service type (i.e., CBR or not). Since only a single bit is required, this addition does not pose any significant overhead to the overall transmission packet.

Table 4.1 An example illustrating the number of mobiles handled by each RA channel in normal and multipleRA mode

Mobile ID Normal mode MultipleRA mode

Table 4.1 An example illustrating the number of mobiles handled by each RA channel in normal and multipleRA mode

Mobile ID Normal mode MultipleRA mode

Channel Number

Number of mobiles

Channel Number

Number of mobiles

(single channel)

per RA channel

(R = 3)

per RA channel

1

1

5

1

2

2

1

5

2

2

3

1

5

3

1

4

1

5

1

2

5

1

5

2

2

Note: Number of active mobiles, M = 5; Number of converted RA channels, R = 3.

Note: Number of active mobiles, M = 5; Number of converted RA channels, R = 3.

Slotted ALOHA was selected as the random access protocol in the RA channel. The BEB algorithm was used to provide stability to the protocol. Such schemes reduce access delay by reducing consecutive collision in the RA channel. In addition, the multipleRA mode provides an additional layer of control for reducing collisions. In situations in which the random access protocol is unable to produce a successful request, the adaptive channel access strategy coupled with the BEB algorithm significantly reduces the collision probability in the request channel.

We summarize the relevant features of ARCMA protocol.

• Efficient channel utilization: Schemes such as the adaptive RA channel, the special handling of CBR traffic, and the piggyback strategy significantly improve channel utilization.

• Slot-by-slot transmission: MS receives ACK to transmit request almost immediately on a slot-by-slot basis. When collision occurs, MSs are quickly aware of their failed request and may retransmit in the next time slot. For a protocol that transmits on a frame-by-frame (by periods) basis, the requesting MS has to wait until the next frame before receiving any acknowledgment. A frame usually has the length (in bits) of multiple time slots. This causes delay that can be critical in a delay-sensitive service. In addition, there can be empty slots within that frame that could have been used for retransmission.

• Transparency to AAL: To reduce the integration complexity between wired and wireless networks, a protocol must provide seamless inter-networking such that the ATM Adaptation-Layer (AAL) is not involved. ARCMA protocol is essentially self-contained within its own network layer. The strategy does not involve the AAL.

• Small RA packet: In ARCMA implementation, we use a single byte (256 mobiles) request in the RA slot. Therefore, the RA slot is just a fraction of an ATM packet (53 bytes). Collision in the RA channel only wastes a small amount of the scarce wireless spectrum.

• Preserved packet order: Since all packets are queued in the mobile's buffer and sent sequentially on a slot-by-slot basis, the packet order is preserved. No complex reordering scheme is required at the receiving end.

• Multiple uplink/downlink channels: In our discussion, we assume a single uplink and downlink channel. In actual implementation, there can be multiple uplink and downlink frequencies.

0 0

Post a comment