Prev: Regarding State Flow....
Next: Matlab fails to start-up
From: Shamsad on 19 Jan 2010 01:25
hello every one

Can anybody tell me pls how to calculate throughput in multihop network in mac layer.
best regards
sharmeen
From: Walter Roberson on 23 Jan 2010 01:32
Shamsad wrote:

> Can anybody tell me pls how to calculate throughput in multihop network
> in mac layer.

I haven't checked around to see what 'csma/ca' is, but generally in
CSMA, Carrier Sense Multiple Access, throughput depends upon the
collision rate, the frame length, the carrier detect algorithm, the
back-off algorithm, the network diameter, the inter-frame gap, the baud
rate, the bits per baud, the error constellation correction code, the
probability of noise, the noise characteristics, and probably other
elements as well. When you add in multiple hops, you have to add in
aspects such as the latency of processing of each hop, and whether the
hops are pure store-and-forward or are cut-through when conditions allow.

On the whole, considering the number of factors, it is usually easier to
do monte-carlo simulation of the throughput than it is to come up with
an explicit symbolic formula parameterized for all of these factors.

In my experience, one thing that I have seen people get wrong many times
is in assuming that the noise is uncorrelated with the probability of
each bit being corrupted being independent of the probability of the
next bit. In the real world causes of noise, that is almost never true.
Different error correction codes can be used for different kinds of
noise, but unless you are somehow adaptive about which code you use
based upon the amount of noise you are detecting, then you are probably
making a poor trade-off, either too much correction for good conditions
or too little correction for bad conditions. If this is all for
wireless, then in order to calculate throughput, you need to take into
account that the back-off algorithms involve either slowing down the
transmission speeds (802.11b, 802.11g) or dynamically determining which
frequencies are getting through best and using those (802.11n, which is
remarkably similar to the way the old Telebit Trailblazer modems worked...)
From: Shamsad on 23 Jan 2010 02:52
Walter Roberson <roberson(a)hushmail.com> wrote in message <hje55g$cfm$1(a)canopus.cc.umanitoba.ca>...
> Shamsad wrote:
>
> > Can anybody tell me pls how to calculate throughput in multihop network
> > in mac layer.
>
> I haven't checked around to see what 'csma/ca' is, but generally in
> CSMA, Carrier Sense Multiple Access, throughput depends upon the
> collision rate, the frame length, the carrier detect algorithm, the
> back-off algorithm, the network diameter, the inter-frame gap, the baud
> rate, the bits per baud, the error constellation correction code, the
> probability of noise, the noise characteristics, and probably other
> elements as well. When you add in multiple hops, you have to add in
> aspects such as the latency of processing of each hop, and whether the
> hops are pure store-and-forward or are cut-through when conditions allow.
>
> On the whole, considering the number of factors, it is usually easier to
> do monte-carlo simulation of the throughput than it is to come up with
> an explicit symbolic formula parameterized for all of these factors.
>
> In my experience, one thing that I have seen people get wrong many times
> is in assuming that the noise is uncorrelated with the probability of
> each bit being corrupted being independent of the probability of the
> next bit. In the real world causes of noise, that is almost never true.
> Different error correction codes can be used for different kinds of
> noise, but unless you are somehow adaptive about which code you use
> based upon the amount of noise you are detecting, then you are probably
> making a poor trade-off, either too much correction for good conditions
> or too little correction for bad conditions. If this is all for
> wireless, then in order to calculate throughput, you need to take into
> account that the back-off algorithms involve either slowing down the
> transmission speeds (802.11b, 802.11g) or dynamically determining which
> frequencies are getting through best and using those (802.11n, which is
> remarkably similar to the way the old Telebit Trailblazer modems worked...)

Thank u so much for your reply. I am using ieee 802.11a standard. I need to calculate throughput in chain topology wireless multi-hop network. In the mac layer i am using CSMA (without RTS/CTS).
For example there is a 4 hop chain topology networks shown bellow.

Source------node1-----node2----node3----destination

source node transmitting data packet throuh this network to the destination. here all node have buffer buffer capacity is 10.Here i just consider only propagation loss. Theoretically we know throughput=1/n(number of hop). I need to calculate the throughput at destination node.
From: Walter Roberson on 23 Jan 2010 15:00
Shamsad wrote:

> Thank u so much for your reply. I am using ieee 802.11a standard. I need
> to calculate throughput in chain topology wireless multi-hop network. In
> the mac layer i am using CSMA (without RTS/CTS). For example there is a
> 4 hop chain topology networks shown bellow.
>
> Source------node1-----node2----node3----destination
>
> source node transmitting data packet throuh this network to the
> destination. here all node have buffer buffer capacity is 10.Here i just
> consider only propagation loss. Theoretically we know
> throughput=1/n(number of hop). I need to calculate the throughput at
> destination node.

As long as there is room on each node for one buffer, the number of
buffers is irrelevant, unless you are assuming collisions and back-off
algorithms and so on, all of which you have implicitly said are
irrelevant by saying that you "just consider only propagation loss".

Throughput is NOT 1/n in any model I can think of at the moment. Latency
isn't either.

Efficiency of the link, in the absence of noise and collisions, would be
calculated by totaling the time not spent transmitting and dividing it
by the total time to reach from the source to destination and
subtracting the result from 1. The total time not spent transmitting, in
a pure store and forward network, would involve a per-node latency at
least equal to the packet reception time. If we assume that the time
required for the radio waves to reach between nodes is equal for all of
the nodes and is designated by T, and the packet reception time is R,
then the total time to transmit is going to be n*(T+R) of which n*R is
unproductive; the efficiency would be (n*T)/(n*(T+R)); the n's would
cancel out, leaving you with T/(T+R) as the efficiency fraction.

The packet reception time would depend upon the signal transmit time of
the packet together with an inefficiency dependent upon the processing
speed of the nodes -- there isn't a node in existence which processes
packets *instantaneously*.

(Sorry, I do not have time at present to examine the question in more
detail.)
From: Makam E Mahmud on 27 Jan 2010 11:57
Can any one please give me a IEEE 802.11g simulink model file.

Thank you
 | 
Pages: 1
Prev: Regarding State Flow....
Next: Matlab fails to start-up