The Time Frame

The MASCARA protocol is built around the concept of MASCARA Time Frame (TF). A Time Frame is a variable length timing structure during which ATM data traffic and / or MASCARA control information flows are exchanged through the wireless link. The WAND architecture has adopted a centralised approach, where the Access Point decides which ATM connections will be served within the time frame, and the Mobile Terminals accept this allocation. The AP can allocate bandwidth over the wireless link dynamically. Dynamically means that when an ATM connection gets a number of slots in one time frame, it is not sure that it will get the same number of slots in the next time frame. In fact it might not be served in the next time frames. An ATM connection is allocated slots according to the traffic requirements it has at every specific moment. The efficient performance of this latter requires the real-time needs of all (uplink and downlink) connections. As downlink traffic we define the traffic that flows from the ATM network via the AP to the MTs, while uplink traffic is defined as the traffic that flows from the MTs to the ATM network via the AP. Downlink needs are immediately derived from the arriving downlink ATM cells. Uplink needs are either expressed through reservation requests that are piggybacked in the data MPDUs or derived from ATM service and QoS parameters. This technique offers the advantage of avoiding collision, when accessing the wireless communication medium. To deal with the part of traffic that cannot be anticipated by the AP, a contention-based method must be nevertheless considered. In WAND the contention period is based on the slotted-Aloha technique. As contention period suffers from limited channel efficiency (due to collision resolution schemes), the MASCARA protocol limits the use of the contention-based method to the strict minimum. Regarding the downlink traffic,

CEC Deliverable No: AC085/INT/ACT/DS/P/035/b1

Title of Deliverable: Wireless ATM MAC – Final Report

contention can be avoided as the scheduling responsibility is located to the AP; for the uplink traffic, contention is only used to issue reservation requests in order to receive (in a subsequent Time Frame) some reserved bandwidth, or to transmit control information.

The name given to the MASCARA protocol developed in WAND reflects the combination of both reservation-based and contention-based access to the wireless transmission medium: Mobile Access Scheme based on Contention And Reservation for ATM (MASCARA). Each Time Frame is divided into a downlink broadcast period, a downlink traffic period (based), an uplink traffic period (also reservation-based) and an uplink contention-based period as outlined in the following Figure 5.

Variable length Time Frame

Variable Boundary

Variable Boundary Variable

Boundary

Reservation based traffic Radio

turn-around

Broadcast

FH Period Down Period Up Period Contention Period

From MT to AP From AP to MT

Contention based traffic

Time

Figure 5 MASCARA Time Frame General Structure

Each of the three periods has a variable length, depending on the instantaneous traffic to be carried on the wireless channel. For the two periods operating in reservation mode, uplink and downlink, it is possible that they collapse to empty periods when no traffic is present. For the contention period, a minimum size is kept to allow any new MT to signal its presence by sending a dedicated control packet. The Time Frame is always beginning with the Frame Header (FH) period, which is used to broadcast, from the AP to the MTs, a descriptor of the current Time Frame. As the size of the Time Frame can change, it is necessary that each MT can learn when each period begins and how long it lasts. Each Time Frame period comprises a variable number of Time Slots.

The physical layer overload of the wireless link is considerably larger than that of wired media. Hence, efficient data transmission can only be achieved if the length of transmitted data packets is not too small. On the other hand, the high BER, characterising the wireless media ask for not-too-large data packets to keep the packet error rate below maximum acceptable values. In the ATM world, the information granularity corresponds to an ATM cell, which is 53 bytes long. This piece of data is considered short compared, for instance, with conventional LAN MASCARA frames (such as IEEE 802.3 or 802.5), so that it would be inefficient to send each individual ATM cell on the air as a single MPDU. Therefore the MASCARA protocol defines the concept of ‘cell train’ which is a sequence of ATM cells sent as the payload of a MPDU. More precisely, each MASCARA Protocol Data Unit (MPDU) comprises a MPDU header, followed by a MPDU payload containing ATM cells generated by or destined to the same MT. In terms of duration, the time required by the physical layer

CEC Deliverable No: AC085/INT/ACT/DS/P/035/b1

Title of Deliverable: Wireless ATM MAC – Final Report

to initiate a MPDU transmission (referred to as physical header) plus the time needed to send the MPDU header is equal to the transmission time of a single ATM cell, i.e. to one slot. The details of the MPDU structure will be given in a subsequent section.

Time 4th MPDU

3rd MPDU 2nd MPDU

1st MPDU

MPDU body: Cell trains of 3, 4, 2 and 1 ATM cells PHY & MPDU headers

Time Slot

Figure 6 Packing of “cell train” within MPDU

TIM E FRAM E STRUCTURE

m oveable

n tim e slots 1 tim e slot

W D LC ed C ELL W DLCoverhead ATM Cell

CONT. Period UP Period

DOW N Period FH Period

Variable length

M PD U PER IO D TIM E FR A M E

m oveable

M PDU n ...

M PDU 2 M PDU 1

M PDU payload: cell train...

M PDU Hdr PHY Hdr

A TM C ELL ATM cell Hdr ATM Cell Payload: unchanged

Figure 7 MASCARA TDMA Structure Summary

The time frame of the end-system WAND MASCARA might be variable or fixed. Its default scheme is variable, allowing the MASCARA to set it to fixed length if needed.

Concerning the ATM performance, it appears that variable frame length could be better than fixed frame length. A scheduler using a variable time frame could utilise frame length freedom to accommodate better the ATM traffic on the wireless link, increasing or decreasing the frame length according to the ATM traffic contract needs.

CEC Deliverable No: AC085/INT/ACT/DS/P/035/b1

Title of Deliverable: Wireless ATM MAC – Final Report

A fixed time frame could restrict the scheduling algorithm to accommodate ATM cells in a ‘sub-optimal’ way.

Variable time frames make the scheduling process more complex, albeit potentially more efficient, since the scheduler has an extra degree of freedom for allocating slots in the next time frame, the time frame length. The frame length is found after the slot map is ready. On the other hand, when the time frame is fixed, the scheduler will do its best to accommodate the traffic within a certain amount of time slots, optimising the allocation for the specific time frame.

As far as system performance is concerned, a fixed time frame would be easier to work with than a variable one. When a MT wants to handover, it should listen to the other APs; if the MT knows when the new AP will initiate its next time frame, the handover procedure could be faster.

In case of Dynamic Channel Allocation (DCA), channel sharing among several APs would be easier using fixed time frames. When a central point controls the channel sharing, it need not take into account the different lengths of the time frames. When the APs need to content for acquiring a channel, the DCA performance would be better if each AP knew the length of the time frames of the other APs. Then the APs would have an idea on when the channel would be available for acquisition.

When a MT looses, for some reason, the frame header, it should search for the next frame header to re-establish its connections with the rest of the network. When the time frame is fixed, the MT knows with more accuracy when to search again for the beginning of the next time frame. Meanwhile it can enter a micro-sleep mode to save some power. When the time frame length is variable, the MT has no indication on when the next time frame will start. In that case, it should scan continuously the wireless link to retrieve the new frame header.

Concluding the issue on variable vs. fixed time frame length, variable time frame approach appears to be better for ATM traffic performance when the AP area cell is at equilibrium (no handovers are performed, DCA is not needed). Several DLC implementation issues and system complexity though, make the fixed time frame length approach also attractive.

Time Frame Fixed Variable

ATM Performance ✓

MAC protocol Complexity ✓

Handover ✓

DCA ✓

AP-MT Communication ✓

CEC Deliverable No: AC085/INT/ACT/DS/P/035/b1

Title of Deliverable: Wireless ATM MAC – Final Report