07.12.2012 Views

On the Effect of Radio Channel Propagation Models - Oulu

On the Effect of Radio Channel Propagation Models - Oulu

On the Effect of Radio Channel Propagation Models - Oulu

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

lows or exceeds <strong>the</strong> FSL curve. This kind <strong>of</strong> behavior can<br />

be also extracted from route breakage figure (Figure 3).<br />

This phenomenon can be explained quite easily, if we recall<br />

<strong>the</strong> power attenuation behavior. The attenuation under<br />

FPL model is so strong that it effectively behaves nearly<br />

like a CP model, when <strong>the</strong> network traffic load is very low,<br />

since <strong>the</strong> interference power far away from <strong>the</strong> transmitter<br />

is almost negligible. However, as <strong>the</strong> network traffic is<br />

increased, <strong>the</strong> total interference power in <strong>the</strong> network is<br />

also increased, and as a consequence, it makes <strong>the</strong> CP approximation<br />

to fail with increasing traffic.<br />

Important metric from <strong>the</strong> application QoS perspective is<br />

<strong>the</strong> average end-to-end delay, which is presented in Figure<br />

5. Typical network behavior is seen: <strong>the</strong> delay stays constant<br />

at low traffic loads, but increases heavily when <strong>the</strong><br />

network becomes near to <strong>the</strong> limits <strong>of</strong> it's capability to<br />

handle traffic. Pushing even more traffic causes buffers to<br />

fill up, which finally leads to buffer overflows (<strong>the</strong> delay<br />

rise settles). We note that <strong>the</strong> used propagation model does<br />

not seem to have great impact to average delay, at least not<br />

at <strong>the</strong> low traffic load. At heavy traffic, <strong>the</strong>re are differences<br />

between <strong>the</strong> points, where <strong>the</strong> network starts to become<br />

unstable. CP-HP gives <strong>the</strong> lowest delay and FPL follows<br />

it at low traffic load. But, as in <strong>the</strong> case <strong>of</strong> packet loss<br />

(and route breakages), <strong>the</strong> delay <strong>of</strong> FPL starts to approach<br />

FSL as <strong>the</strong> traffic load is increased.<br />

>c O<br />

>. 0,0<br />

10 I.. ..,<br />

0,01<br />

Offered data traffic<br />

Figure 5. Average data packet delay (BC-MAC).<br />

B. IEEE 802.11 and Comparison to BC-MAC<br />

At <strong>the</strong> very high traffic loads, <strong>the</strong> performance differences<br />

between different cases can be clearly seen in throughput<br />

(Figure 6). In this figure, <strong>the</strong> FSL and CP-HP curves <strong>of</strong> <strong>the</strong><br />

802.11 and BC-MAC are shown. An interesting observation<br />

can be made when considering 802.11 curves. <strong>On</strong> <strong>the</strong><br />

contrary to BC-MAC results, 802.11 FSL gives better performance<br />

than 802.11 CP-HP! This is quite unexpected,<br />

but can be explained: The very nature <strong>of</strong> this result originates<br />

most likely form <strong>the</strong> well-known hidden-terminal<br />

problem. CSMA is based on power detection and a single<br />

transmission can be just detected at a distance <strong>of</strong> 250 m.<br />

10<br />

6 <strong>of</strong> 7<br />

However, when <strong>the</strong>re are several transmissions simultaneously<br />

at <strong>the</strong> different portions <strong>of</strong> <strong>the</strong> network, <strong>the</strong> cumulative<br />

power <strong>of</strong> <strong>the</strong>se transmissions can cause signals<br />

to be detected from far<strong>the</strong>r than 250 m. Therefore, this<br />

basically alleviates hidden-terminal problem as compared<br />

to <strong>the</strong> CP-HP case, in which <strong>the</strong> CSMA range is<br />

always 250 m, and shows a better throughput is a result.<br />

The same phenomenon also exists with BC-MAC, but<br />

<strong>the</strong> effect is only minor, since node-specific spreading<br />

codes and processing gain handle <strong>the</strong> situation.<br />

0.7<br />

0.6<br />

0.5<br />

D<br />

Ideal random channel access<br />

--802.11 (FSL)<br />

--802.11 (CP-HP)<br />

0.4 -R<br />

_sc'<br />

=3 . BC-MAC (FSL)<br />

s<br />

0.3<br />

BC-MAC (CP-HP)<br />

0.2<br />

0.1<br />

O 1-<br />

0.001 0.01 0.1 1<br />

Offered data traffic<br />

Figure 6. Average throughput (BC-MAC vs. 802.11).<br />

Offered data traffic<br />

1 .vrL nF+nn uu T<br />

0o0 )Q1<br />

,..<br />

0001 00<br />

. _..<br />

1 -1<br />

1 .OE-01 t<br />

, 1.0E-02<br />

-~eou<br />

1.0E-03<br />

1 .OE-04<br />

...- -.,.-802.11(FSL)<br />

-- - 80211(CPHP)<br />

. 1BCMAC 8021 (FSL)<br />

BC-MAC (CP-HP)<br />

1.OE-05<br />

Figure 7. Average packet loss ratio (BC-MAC vs.<br />

802.11).<br />

This interesting phenomenon affects only at <strong>the</strong> high<br />

traffic load as seen from <strong>the</strong> packet loss behavior (Figure<br />

7). 802.11 CP-HP clearly dominates <strong>the</strong> performance<br />

over <strong>the</strong> 802.11 FSL case at low traffic load. But, as <strong>the</strong><br />

traffic load is increased, <strong>the</strong>re will be more and more<br />

simultaneous transmissions at <strong>the</strong> different parts <strong>of</strong> <strong>the</strong><br />

network, and hence, <strong>the</strong> phenomenon is enhanced. From<br />

about <strong>the</strong> traffic load <strong>of</strong> 0.08 forward, <strong>the</strong> performance<br />

under FSL is better than under CP-HP. As a consequence,<br />

commonly used CP-HP model clearly fails,<br />

since it gives too optimistic results at <strong>the</strong> low traffic load<br />

and too pessimistic at <strong>the</strong> high traffic load.<br />

10

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!