 given 6 nodes in a string i 0 to i 5 and let us say each node has data for its next neighbor, wants to send some data to its next neighbor. And suppose RTS, CTS and AC takes 1 unit of time each, data and AC or let us say data takes 5 units of time, let us just chain the numbers. So, we have to figure out when does the transfer complete, when does when do all the nodes finish communicating to each other. So, basically if you see one transmission, one frame exchange sequence. So, we always deal in terms of this frame exchange sequence in a to 2 dot 11. So, one frame exchange sequence is what? RTS plus CTS plus data plus AC right. So, that is equal to 1 plus 8 units of time. So, one frame exchange sequence is going to take 8 units of time right. How many can go in parallel? So, let us look at how many can go in parallel round 1. Let us say in the first round which suppose i 0 is transmitting, i 0 to i 1 transmission is going on. Is there anything that can go in parallel with it? i 1 to i 2, i 4 is not transmitting to i 3, i n is sending to i n plus 1. So, can i 2 transmit to i 3? Suppose i 2 is transmitting to i 3, i 0 is transmitting to i 1. Where will happen? Where will be the collision? Suppose i 0 and i 2 are both transmitting simultaneously ok. There will be a collision at i 1 right. So, that is why i 2 cannot transmit at the same time as i 0 even though it is transmitting in the other direction right. How about i 3? Can i 3 transmit to i 4 right. So, the first round what you have is i 0 goes to i 1 and i 3 can transmit to i 4 right. So, that is 8 units. So, there are two parallel transmissions what happens in the next one? Next one you have i 1 transmitting to i 2 and i 4 to i 5 ok. And in the third round you can have who is transmitting? i 2, i 2 to i 3 ok. So, you have 24 units of time as the minimum amount of time in which this data transfer can complete ok. So, all right. So, now going on with this same example that we had ok. So, the key thing that we were talking about is priorities right. So, how do we set up priorities? Now the important thing here is that there are three different types of inter frame spacings right. So, initially before you start transmitting. So, let me first bring up the slide and then we will try to understand why this is so ok. So, if you see that between RTS and CTS ok, do I need any gap? So, I am sending you an RTS you have to send me back a CTS right. Is there any time interval that will come at that point? Timeout it is not a timeout due to switching right. See always remember we are trying to keep the cost down. So, we do not have separate receivers and transmitters. So, you have to finish receiving the RTS and then you have to switch to transmitting the CTS. So, there is going to be a receiver transmitter turn around time correct. What other delays? Let us try to work this out again in a little bit more detail ok. So, RTS, CTS delay is involved ok. So, one is so let us say construction of RTS ok. It will take some time to construct the RTS packet ok. So, this typically we will say ignore because you start counting after the RTS packet is constructed. Then second is transmission of RTS ok. Then this is saying from the network card on to medium right. So, you have created the RTS packet. Now, you have to transmit the RTS packet on to the medium. So, there is a delay involved in being able to put the transmission put the RTS packet from the network card on to the medium right. This is typically what is called your transmission delay right after that propagation delay right right. So, there is a delay involved in propagation of RTS to receiver after that receive reception of RTS ok reception of RTS ok. So, again from medium from phi to from the phi to the MAC right I have to receive the RTS at the receiver correct. Then ok receiver transmitter turn around then construction of CTS right construction of CTS or what is called the MAC processing delay ok. This is the MAC processing delay you have to construct the CTS packet that is your MAC processing delay. So, at least there are these many delays involved in an RTS CTS exchange ok. So, it is not instantaneous correct. So, this is called your short inter frame spacing right. So, basically what this means is after you send the RTS packet there is a certain delay before which you cannot receive the CTS packet ok. So, this is typically of the order of 10 microseconds. So, similarly you can calculate other types of delays in your network. So, these are the different types of short inter frame spacing it is the highest priority for acknowledgement CTS polling response and so on ok. So, this is generally called as 10 microseconds there is also a notion of a slot time we will just come to that in a moment ok. So, let us just see how this looks ok. So, when we are doing RTS CTS you have the sender waiting for a DIFS amount of time ok. There is an RTS packet that goes on the network there is a SIFS amount of time what is this SIFS amount of time all these delays that we just tried to measure then we sense the CTS then again there is an SIFS amount of time and then the data again there is an SIFS amount of time and then the acknowledgement correct. So, this entire thing is your frame exchange sequence ok. So, what is this NAV NAV basically stands for network allocation vector. So, as soon as I hear the RTS the other station is going to set that the network is going to be busy for so much duration of time. The RTS packet comes along with this information of how much data is to be transmitted. So, that is when it is going to set the NAV that is it is just a jargon that is being used is that making sense how the RTS CTS works ok. So, now we understand why this DIFS has to be greater than SIFS time right. If the DIFS time is equal to the SIFS time or less see because nobody ask me the question why 50 microseconds right. So, this class we have been asking a lot of wise right I said DIFS time is 50 microseconds nobody ask me why 50 microseconds why is it 50 microseconds ok. That is a good question right. How did I come up with 10 microseconds as the SIFS time ok. So, it depends upon what does it depend upon? This can be fixed right this can be fixed, but this will depend upon rate distance and also rate right it will also depend upon rate. So, are you doing 11 mbps are you doing 54 mbps right. What is the type of phi that you are using here that is what it will take right. So, that is all these things are going to depend upon the rate of the phi. So, depending upon the type of the phi that you are using for direct sequence spread spectrum ok. And 11 mbps phi you have 10 microseconds. So, this is for. So, if you are going to use frequency hopping spread spectrum and frequency hopping is actually also 11 mbps or if you are using OFDM and 54 mbps then the slot the SIFS time will be different ok. What are the various timings that are involved ok. So, we have seen that SIFS time is 10 microseconds ok. DIFS time has to be more than 10 microseconds. So, if you see here we have defined DFS time as SIFS time plus 2 times the slot time. What is the slot time slot time is just a notional convenience ok. So, there is no notion of real slots in A to 2 dot 11. Just for convenience we are defining it as slot time ok. So, SIFS time is basically 10 microseconds, slot time is sorry slot time is 20 microseconds ok. So, ideally we could have kept DIFS time as SIFS plus 1 slot time ok. Is it not? Why is the slot time defined as 20 microseconds? There is just convenience right that is what I said. Just for convenience we are defining slot time as 20 microseconds. So, given that SIFS is 10 and slot time is 20 microseconds ok. What we do is we cannot have DIFS just as SIFS plus 1 slot time because suppose there is an access point in the medium right. Suppose I have an access point suppose you and I are talking, but there is an access point in the area which is going to also control the medium. So, the access point should have priority over any two random people talking right. Like let us say you are at an airport ok. So, you have a laptop, you friend has a lot laptop and there is an access point also which is installed at the airport right. So, now the access point should get priority over communication in that medium rather than you and your friend right. So, that is the reason why you have what is called the PIFS time in between ok. PIFS stands for PCF Interframe Spacing. So, since you are going to wait for ok, since you are going to wait for this much amount of time ok. Since you are going to wait for a DIFS amount of time, if the access point is going to wait for lesser amount of time then it will be able to grab the medium before you do. So, that is all that is all it means ok. So, in order for this frame exchange sequence to complete you are going to use the minimum amount of time right which is SIFS. After that the access point is the next higher priority station. So, it will use half of this time and then will be any other free agent which can transmit here. So, PIFS time is SIFS time plus slot time, DIFS time is SIFS time plus 2 times the slot time ok. So, that is all that is there to it ok. So, there is also a notion of fragmentation which we are not really going to go into, but basically this is again a straight forward extension of this RTS-CTS mechanism. Now, the question is how long a packet can I transmit on a wireless medium right. The longer packet I transmit what happens? The less the the less the overhead correct with one RTS-CTS I can send the long packet, but the longer the packet the more the chances of error correct. So, depending upon what happens in the medium sometimes what you can do is after the RTS-CTS you can send one fragment, again use the same SIFS acknowledgement mechanism send the other fragment and then again use the same SIFS and acknowledgement mechanism. So, this is called fragmentation in e to 2 dot 11 right. So, when you look at a card ok. So, when you look at configuring either an e to 2 dot 11 access point or card ok. So, you will see 2, 3 numbers there ok. So, you will see 3 numbers there one you will see something called the RTS threshold ok. So, this is just a screen which you will get how many of you have seen this ok. If you have a laptop you just try to configure its wireless interface you will see all these numbers no. So, you will get a table which will say RTS threshold and you have you will have a drop down there where you will set some number there ok. So, let us say 1200 bytes right. Then similarly there will be something called a fragmentation threshold ok. That will be something like 2400 bytes there will be power transmit power. So, if it is an access point you will say ok 100 1 watt or 100 milliwatts you will see channels. So, e to 2 dot 11 has how many channels? 11 channels ok. So, you will see channels 1 to 11 and you have to pick the channel on which you are operating. So, there will be a bunch of such configuration parameters that you will see when you are trying to configure either your access point or your card ok. So, on the access point you will pick. So, generally out of these you have channels 1 6 and 11 are non interfering, non overlapping ok. So, what does that mean? That basically means you can have 3 operators operating in the same area. In the same airport you can have 3 different operators operating without having to worry about any interference caused from one to the other because they are on non overlapping channels completely non overlapping. So, what this basically you have taken 83 megahertz of the spectrum right. So, out of the 83 megahertz spectrum you have said this is channel 1 and then this is channel 2 and then this is channel 3 and so on ok. So, that is the way you are dividing up the channels. So, channel 1 and channel 6 are totally non overlapping ok. Channel 6 and channel 11 are also totally non overlapping. Is that making sense? What we are saying? So, this is the idea of fragmentation. Let us do a small example at this point. Is the fragmentation used to do some kind of reduction in the interference. So, not exactly. So, what fragmentation is used to used for is you set a threshold and you say that if I am going to send such a large packet see the then I am going to break up the packet into smaller chunks. So, that I can utilize you know in case one part of it does not get through then only that part needs to be retransmitted. So, that is the key idea instead of sending one large packet you are going to send smaller packets. So, that your retransmission effort becomes less. So, that is the key reason for doing the transfer. Will it improve the? Ok. So, you tell me. Will it improve the efficiency? No, it is not a clear answer. So, there is a depends clause to it. So, it will improve the efficiency under what circumstances? If there are the medium is loss if the medium is error prone then if I do not use fragmentation I will wind up sending larger packets which will get lost which will have to be retransmitted. So, at that point when I am using fragmentation then it may happen that some of my fragments will get through and only smaller fragments may have to be retransmitted correct. So, in a lossy medium it is likely to improve the efficiency. On the other hand if my medium is loss free then it will reduce it because I have this overhead of you know each fragment is going to be associated with an SIFS and AC you know another SIFS another fragment another AC. So, that is the overhead which is associated with the ok. So, there is something called an EIFS there is also something called an AIFS. Now, we are not going to go into those things here. They come in for the A22.11 E kind of a technique we are not going into too much detail about those interframe spacings. Let us do an example here. So, suppose we have three stations ok. Let us just to understand just to make sure that we understood ok. So, you have S1, S2 and S3 ok. Now, S1 has a packet to transmit at T equal to 0 ok packet of size 500 bytes ok. S2 gets a packet at T plus 120 microseconds of size 1400 bytes and S3 has a packet of size T plus 250 microseconds packet of size again 500 ok. So, we have three stations. Does it matter to whom they are transmitting? Suppose I were to say S1 is transmitting to S3, S2 is transmitting to S1 and so on. Does it matter? Does the destination matter? So, this is again a question out of one of my exams ok. So, in that I have lot of these extraneous detail saying that S1 is sending to S2, S2 is sending to S3 and so on. Does it matter? That is irrelevant because it is a LAN right. So, we are saying it is a LAN. So, it really does not matter who the destination is. It only matters who is going to get a chance to speak right. So, S1 actually S1 is transmitting to S3, S2 is transmitting to S1, S3 is transmitting to S2 ok. That does not matter. Now, given that given that slot time what was slot time? Slot time is 20 microseconds, SIFS time, SIFS time was 10 microseconds right. Given slot time is 20 microseconds, SIFS time is 10 microseconds right. Let us say we have RTS CTS act size as 100 bytes ok. And then we have RTS threshold as 1200 bytes and let us say fragmentation threshold as 2400 bytes ok. And now suppose we say 200 bytes per slot time ok. You can transfer 200 bytes in each slot time. That is just to make the thing easy ok. So, when is the data transfer going to complete ok. Let us not worry about the exact number. Let us try to see how to go about doing this. So, what happens to S1? Send RTS first send RTS is that right? No. First DIFS. First is DIFS. First is DIFS. Then RTS. Then I send RTS. RTS. Is it? See the packet size. See that is why it is important to know what is the packet size. Packet size is 500 bytes. My RTS threshold is 1200 bytes ok. So, for any packet which is smaller than 1200 bytes I am not going to send any RTS CTS. So, this is DIFS plus data plus AC. Is that correct? DIFS plus data plus AC. Yeah. So, there has to be an SIFS. DIFS plus data plus SIFS plus AC right. So, this DIFS is now SIFS plus 2 slot time 50 microseconds. Data is 500 bytes which is going to take 200 bytes per 20 microseconds. So, this is again another 50 microseconds right. This will take another 50 microseconds plus this will take 10 microseconds plus this will take 10 microseconds right. So, 120 microseconds. S1's data transfer completes in 120 microseconds right. Agreed? Okay. What happens to S2? S2 waits for DIFS. Then plus RTS plus CTS SIFS plus CTS SIFS plus CTS SIFS plus CTS SIFS plus CTS SIFS plus CTS SIFS plus data plus CTS SIFS plus AC right. That is what happens for S2 correct. How much does this work out to? Okay. So, this is 50 plus 10 plus 10 plus 10. What is the data? 1400 bytes right. So, this is 140 plus 10 plus 10 correct. That is how much is this? 250 this 250 microseconds. This data transfer takes 250 microseconds on your air okay. So, what happens to S3? S3 starts with DIFS. See look at what has happened. At t equal to 0 S1's packet came 120 microseconds were taken by t equal to 0. At t equal to 120 plus 250 is when the medium has become free okay. The S3's packet has come at t plus 250. So, what happened when S3's packet arrived? It found that the medium was busy. So, what should it be doing? It has to do a back off before it does the DIFS right. So, it has to do back off right. So, even if you assume that it is going to back off for one slot right. So, you will have it going for 20 plus this DIFS data SIFS AC which is 120. Is that making sense? I cannot just say that S3 is the same as S1 although the packet size is the same okay. The data transfer duration is not the same because S1 did not have any back off involved. S3 has a back off involved okay. So, S3 is going to have back off plus all this DIFS business. So, that will be 20 plus the 120 microseconds which we know that this exchange takes 120 microseconds correct okay. So, now you can calculate that this is again another 140 microseconds. So, you add up all these three numbers which is what 260 510 right which is 510 microseconds is when the entire data transfer is going to complete is that making sense? Will S2 have a back off time? No because just when its packet arrives it finds that the medium is free because S1's transfer data transfer took 120 after the AC there is no SIFS after an AC. Once S1's data transfer is complete at the same instant of time S2 packet arrived S2 started sensing the medium right. So, it finds that the medium is free. So, it does not back off right. If S2's packet were to arrive at T plus 119 then it will back off okay. Because it arrives at T plus 120 or let us make it easier T plus 121 it is not going to back off okay. But that is not true for S3 because its packet arrives in between S2's data transfer. So, S3 has to do the back off correct. Okay. How is the back off time calculated? Random right. So, when you ask the question what is the minimum amount of time in which the data transfer has to complete? You have to assume that you know one unit of back off is going to be performed okay. Yeah S3 follows the routine saying that I am going to come. So, see the same thing that happens in the previous slide that is what is happening here. So, S3 is like the station here it got data when it found that the medium was busy. So, when it found that the medium was busy it just picks a back off number. So, this is the station 2 that we are showing here is corresponds to S3 in our example correct. So, this one is going to wait it is going to find that you know somebody else is transmitting it is going to pick a back off number here then it will wait for this DFS amount of time it will wait for its back off to finish and only then it can transmit correct. So, this is the case which we have to compute. So, now all that we are saying is that this back off is a is the minimum that we are picking because we want to know what is the earliest time the data transfer can complete. So, while I am counting down if I find that the medium becomes busy then I have to again stop and go through the process again see that is the case that is happening with S1 in this example. See what happen to S1? S1 is counting down while it is counting down it finds that this guy has already started transmitting. So, it has to freeze its back off counter and then again it has to wait for DFS and then again it has to do the counting down. It is not after back off if during back off. So, I find that somebody else is occupied the medium then I have to wait. See it is always like this the previous slide shows that no see it is always like this. DFS if the medium is busy then you back off otherwise you wait for a DFS count your slots the medium continues to be free which is transmit. There are no 2 times DFS involved is that mean sense? I wait for a DFS amount of time I do the back off counting. If for this entire duration I find the medium is free I transmit that is all. If for the DFS amount of time I find that if I am the first to arrive then my back off value is 0 that is all. You always wait for DFS plus back off. Sometimes the back off value may be 0 if you are the first guy to arrive other times you may have a non 0 value for the back off. So, let us continue let us look at one other mode of 802.11 function. So, this is what is called the DCF mode of functioning. So, DCF means distributed coordination function. So, everybody waits for DFS amount of time and goes ahead. The other mode called the PCF is much easier to understand, but unfortunately it has not been implemented by any vendor. So, all the vendors use DCF only nobody has implemented this PCF mode. The PCF mode is basically saying that there is an access point. So, think of it like this no I can be considered as the access point and all of you can be considered as the nodes. So, I basically poll every node saying that do you have any data to transmit? Do you have any data to transmit? Do you have any data to transmit? That is as simple as that the PCF mode of operation. So, if you look at the slide you find that PCF mode you wait for a PIFS amount of time. Why is this PIFS and not DIFS? Because I have to get priority since I am the access point I have to get priority over somebody else who is waiting to transmit in the distributed mode. So, it waits for SIFS plus one slot time transmit downlink data 1, SIFS and then this guy will transmit the uplink data, SIFS it will transmit the downlink data, this guy will transmit the uplink data. So, that is the way it goes. So, I will send you data and then you can send me back data in the same frame. In the same frame exchange sequence. So, pretty much the same action that takes place. Is there a need for RTS-CTS here? No, there is no need for an RTS-CTS. Is there a need for acknowledgement? Yes, there is always a need for acknowledgement in an unreliable medium. How do I do acknowledgments? See for example, in this we have not put in any acknowledgments. How do I do acknowledgments? I will piggyback it. So, on this data I will put an acknowledgement which will acknowledge this and then this will be data for the next station. So, there is a lot of detail here, but we do not need to go into it because PCF is not really implemented. And if you look at A22.11e, the newer generation of access points which are A22.11e, this PCF has got transformed into something called XCF, hybrid coordination function. So, that is why we will not do too much detail of PCF. So, at the end of the PCF, the access point will send something called the basically to signify that the PCF mode has ended. It will send a packet which will signify that. It is called the CF end. The only way in which you can tell me that to add your name into the polling cycle is using the DCF mode. So, that is why it is always that even in the PCF mode. So, even if you are doing PCF mode, it works like this. So, you will wait for a PIFS amount of time. You will go for what is called the contention free period. So, this is the polling. Then it will go for a contention period. This is your DCF mechanism RTS-CTS mechanism. And this cycle will repeat. After this again it will go for PIFS and so on. So, even in the PCF mode, you find that there is a it is not just contention free. Towards the end of it, you always have some contention period. So, this is for new stations to join and for other conversations. Now, for example, if it is an airport, you do not want the access point to control the medium all the time. So, that two other people who are not belonging to the access point network can never get to speak. So, that is why non PCF traffic. In polling, is there any priority which can be set? Theoretically yes. Practically since nobody has implemented PCF, it does not matter. Theoretically yes, because the access point is the one which decides the schedule. How do I set the priority? I set the priority by giving you a chance twice and giving that person a chance only once. So, the access point can easily do that. It can easily implement priority by saying that I will poll this station twice as often as I poll the other station. Each time you are polled, you can transmit one packet. So, that is the way in which it could be set if there is a schedule in the access point. So, as such in the standard there is nothing mentioned. The standard does not say that this should be the you know it should be around robin or it should be anything. It just says that some scheduling algorithm should be implemented at the PCF for deciding the polling schedule. So, given that let us see how this gateway business works. So, let us say so, what we are now trying to understand this is probably the one of the last points that we need to understand. So, let us say you have a network like this, you have access point 1, you have access point 2 and you have STA 1 and you have STA 2. So, now STA 1 wants to send a packet to STA 2. What are the different segments of the communication? That is what we need to figure out. Send a packet to STA 2. How does that happen? So, the question is how many addresses do I need in my frame? Four addresses I need which four since we are showing four entities, we need four addresses. So, what is the first hop of my journey? STA 1 to AP 1. Second hop is AP 1 to AP 2, third hop is AP 2 to STA 2. Now, given that this is how our four addresses should look like. So, we know that we need four addresses in our system address 1, address 2, address 3, address 4. Can you try to figure out what will go into each of these addresses in each of these cases? So, we have three cases. Now, this is very dry. If I were to just describe this, it will be very boring. So, better to try to work it out. Look at these three cases. What will be the four addresses that go in here? So, what does it say? The first address is the destination address. So, address 1 will be STA 2. Is that right? STA 2, address 1 is the, the next one is the source address, the STA 1. And the third address is to the AP. There is no fourth address here. So, in the first case, what you are basically telling the AP? So, you can think of this as the destination is different from the receiver. The source is different from the transmitter. So, try to keep a distinction between what is the source and what is the transmitter and what is the destination and what is the receiver in your mind. Then, you will be able to figure this out very quickly. What is the source in the first step? STA 1. What is the transmitter in the first step? STA 1. So, the source and the transmitter are the same. What is the destination in the first step? STA 2. What is the receiver in the first step? AP 1. AP 1. So, that is why you have these three addresses working out like this. Similarly, if you go to the third step, what is the destination? STA 2. What is the source? STA 1. STA 1. What is the transmitter? AP 2. So, the second case is the one where you have all the four addresses that are required when this access point is sending it to this access point. So, what happens here? Transmitter is access point 1, right. Receiver is access point 2, but the source is STA 1 and the destination is STA 2, right. Is that making sense? That is why you need four addresses. So, always try to determine when you are trying to do an example of this kind, try to keep the distinction of source versus transmitter and destination versus receiver in your mind. Then it cannot go wrong. So, you may actually the only place you could go wrong is fill in the wrong entry in the wrong column. That is okay, but alright. So, this is STA 2. This is STA 1. This is AP 1. This is AP 2. Okay. So, now you can again use this as an example for setting lots of questions, right. You can give a different scenario. Suppose I put STA 2 here, then what will happen? Suppose I have to go to the internet, then what will be the addresses? So, that is a good way of you know giving examples on this basis. Okay. So, like I said you put the wrong entry in the wrong column. The correct column is defined as far as this table is concerned. What you need to understand is what are the entries? Okay. It does not matter which entry you put in which column. For that you can always look up this information. Now, this guy will tell you that the BSS ID which is the access points ID should go here, right. And then this is the source address is the destination address. So, I just put it so that it looks uniform. So, that it looks as destination, source, receiver, transmitter. Okay. But that is not the order in which the packet addresses need to be constructed. Okay. So, there is a reason for why this is so, but that is not important. Is this format specific to A22.11? Yes. No, this is not the Ethernet format. See, in Ethernet there is no notion of multi hopping. Okay. So, Ethernet is single hop. So, what is the Ethernet frame format? Source address is destination address. Finished. Okay. Significance of single hop is just one connection. So, source is directly connected to the destination. So, when source is directly connected to the destination you call it single hop. Okay. So, when source has to go through an intermediate relay to the destination then it is called multi hop. Okay. All right. So, this is a kind of a summary of A22.11. Not very important, but the slide looks good. That is why I put it there. Okay. So, it tells you what is .11A, .11G, what are the key differences and all that. Okay. So, let us do one exercise which I think is a bit important. Okay. So, what we did, because that is where we have started this whole design of the MAC. Right. So, voice capacity in GSM. Okay. This per cell we found was 125 into 8. Right. It is bounded by 125 into 8 which is less than 1000 users. Right. Using what? 25 megahertz. So, let us do the same exercise for A22.11. Okay. And try to find out what it will have. Okay. So, we know that A22.11 uses 83 megahertz. That does not matter, but we can say that it uses a 54 Mbps radio. Okay. This is the transmit rate. Phi rate. Okay. Let us say it is 54 Mbps or it is 11 Mbps. Okay. You can use either of them. It does not matter. 11 Mbps or 54 Mbps. So, this 83 megahertz is not so significant. Right. But we know that we are transmitting at 11 Mbps or at 54 Mbps. Okay. What is what do we expect as the answer? Okay. So, if you just see the numbers. So, using 25 megahertz and okay and 9.6 Kbps rate. So, using 25 megahertz and 9.6 Kbps, GSM is able to give you about 1000 users per cell. Right. Now, using 83 megahertz and 54 Mbps, how much do you expect A22.11 to give you? More? Obviously. Right. Okay. So, that is why we need to do this calculation to understand why it is not obvious. Try to do the calculation now. Okay. Let me give you a hint. Okay. What you need to do is here there is a voice packet. Okay. So, you will have let us say voice data. This will become data there will be an RTP header. So, let us say this is voice data which is 160 bytes, 64 Kbps. Okay. So, if you are using G711 codec. So, it depends upon the codec that you are using. So, now if you are using G711 codec. Okay. So, you it is going to sample it at 64 Kbps G711 codec sampling. So, you are going to generate this will work out to I think 160 bytes every 10 milliseconds. Okay. It is going to generate 160 bytes every 10 milliseconds. Okay. So, what you will do is you will take this voice data. You will attach an RTP header. Okay. RTP is the protocol in which you have to carry voice. Right. Say all these guys have to attach a RTP header. Okay. Then what happens? You attach a UDP header. Okay. Then you will attach a MAC header. Okay. Then you are going to attach a MAC header. Then you attach a PHY header. Okay. So, so many headers get attached to your 160 byte data packet. Alright. And then you are going to transmit it on to the wireless medium to the access point. Okay. And then the access point has got to transmitted back to you. So, just try to come up with some rough numbers. So, let us say you assume some numbers for this RTP header is 20 bytes, UDP header is 20 bytes. You know, MAC header you can take to be another, what is the MAC header in 802 total. Let us just take 20 bytes as the numbers. Okay. PHY header is again now .11 the PHY header is I think close to 60 bytes or something. Okay. So, if you just take these numbers, what do you find? Alright. Let me do one thing. Let me tell you the answer. I do not have all the exact numbers here, but can you guess? Now that you have some idea of, you know, there are all these headers that are there. Can you guess in what range this number is going to fall? So, if you are just going to do plain calculation, if you just take the headers into account, you will find that this will be around 400. Okay. But then there is one thing which we have missed out, which is not collisions, assuming no collisions. Okay. Assuming that there are no collisions, you still have SIFS plus data plus SIFS plus AC, that overhead. Okay. So, if you take all the overheads into consideration, the total number of voice calls that can be supported in such a system is 22. Okay. See, that is the key difference. So, that is why I wanted to illustrate this. Why it is so important to design the system, keeping the final requirement in mind? This system is not designed for voice. The system is designed for data. System is designed for unlicensed spectrum. It is designed for low cost devices. It is designed for simple implementation. Okay. So, if on such a system, you try to send a voice traffic, okay. The total voice under kind of good conditions, what you can have is about 22 voice calls is what you will see in such a system. Whereas, compare that with the GSM, you know, it is operating on much lesser spectrum. But it is able to cater to much larger number of users primarily because it is designed for that purpose. There are no headers. That one overhead has been taken off and already you get a huge gain. All right. So, do one thing. You try to figure this out. Take some numbers for the RTP, UDP headers and all that. You already know the numbers for the SIFS data. Already know the numbers for back off. Right. So, remember that voice is now a two-way communication. Okay. So, because of that, you will find that it comes down to some number like this. Okay. Also think of what is happening from the access points perspective. Right. So, the access point is the one to whom everybody is trying to talk to. Right. So, even if these guys do not see any contention, at the access point, there is always be always going to be contention for transmitting. Right. So, the access point is contending because it always has a queue of packets which it needs to send. Even though there may be four clients, each client has one packet, but the access point has four packets, one for each client. Understand what I am saying. Okay. So, that is why the access point, the back off keeps on getting more and more severe. Okay. So, in reality, you get a number something like around 20 18 to 20 calls is what you can support in a A2 to dot 11 system voice calls. Okay. Theoretically, does the number of users supported by an access point change from 11 MPPS to 54 MPPS. Okay. Marginally, it will change, but it is not significant because a significant overhead is in terms of your SIFS time, your DIFS time. You always have to wait. Right. So, the access point has sent out one packet. What does it have to do? It has to back off. Even upon see the one key difference between Ethernet and A2 to dot 11 is that even upon successful transmission, you have to back off. Why? If you are unsuccessful, you have to back off that everybody can agree to. Why do you have to back off even if you are successful? Give a chance to others. Correct. Because of that. Right. If I am successful, I know exactly when my transmission ends. And if I start counting DIFS from there, I know that I am going to be the guy who counts down first. Right. So, even if I am successful, I have to back off so that other people who are trying to transmit on the medium can get a chance to transmit. So, think of what the access points life becomes. Right. It has more packets to send. And now after every packet, it has to back off. Okay. So, the system slows down. So, access point becomes the bottleneck when we are doing voice over IP in a Ethernet, sorry, in a Wi-Fi medium. Right. So, voice over IP over Wi-Fi when you do, you have to keep in mind that the call handling capacity of the system is very small. Okay. So, once people discovered this, they went on to say that okay, now how do we improve this? Okay. So, how they improved that was at the access point. So, 802.11 e came into being as a result of this. Right. So, at the access point, they said that you know if I have multiple queues. Okay. So, let us say this is the priority. So, at the access point, so priority is implemented different queues and different waiting times. Okay. So, I will just try to motivate how this thing came into being. Okay. So, this is let us say your voice queue. This is your FTP queue. Okay. The access point is connected to one STA1 which is doing voice. Okay. Another STA2 which is doing an FTP. Right. Okay. So, what we just saw was that the access point becomes the bottleneck because all the packets get queued at the access point. Right. So, now if the voice flow and the FTP flow have to compete with each other, then the life for the voice flow becomes even worse. Right. What will happen? The delays will start increasing. Right. Voice, what is the maximum delay tolerance in a voice system? It is about 150 milliseconds end to end. Right. So, even if you are sitting at the other end of the world, you need the voice to be able to reach there within 150 milliseconds of leaving here. Okay. So, about 150 milliseconds of which you say that you know half of that delay is in the wireless part itself. So, you say that I need to cross the wireless link in about 75 milliseconds. 75 to 80 milliseconds, you need to be able to cross the wireless link. Right. Where is the maximum delay in the queuing? It is not the delay in propagation. So, that is why 11 Mbps, 54 Mbps does not matter so much because that is not the delay. That is anyway happening at light speed. Right. But the queuing delay, the medium access delay, the DFS takes 50 microseconds out of the whole thing. Right. So, that is where the main delays come into play. So, once people observe this, they said, okay, let us try to separate these flows. So, at the access point, so what A 2 2 dot 11 E has is it has various flow categories. Okay. What are called access categories? AC 1, AC 2, AC 3, AC 4. Okay. So, access categories basically means that I am going to classify a flow as a voice flow or as a video flow or as a FTP flow or as a HTTP flow. Okay. Each of these are the four examples of the four categories. Okay. Voice is highest priority and periodic. Right. Voice is periodic. Highest priority and periodic. Video is high priority, but not periodic. Right. It is not constant bit rate. It is variable bit rate. Then you have FTP which is a bursty data. And then finally, you have HTTP which is your best effort flow. Just say that, okay, if I am able to get this through well and good. Okay. So, those are the four access categories which this guy uses. Okay. AC 1 is voice. AC 2 is video. AC 3 is FTP. AC 4 is HTTP. Okay. So, A 2 2 dot 11 E creates four different cues and then it can do some kind of prioritizing. It can say that I am going to serve two packets from the AC 1 cue every time before I serve a packet from the AC 4 cue. Okay. That is one way of doing prioritization. Another way of doing prioritization is that the back of values that I choose for the AC 1 cue will be smaller than the back of values that I choose for the AC 3 cue. Correct. So, these are the ways in which dot 11 E tries to do prioritization. Okay. How does it know that it is a voice packet because everything is IP because you are going to monitor the call setup. Okay. You have to inform that this is a voice packet. Okay. So, I can actually infer by just looking at it. So, there are two ways in order to understand. One is where you explicitly inform that this is being set up under this category. Okay. So, at the time of setting up a flow or at the time of you know connecting to Google, the card on your machine can inform the access point that I am starting a flow which is belonging to such and such category. Okay. So, I could explicitly inform or I could infer it by looking at the packets that are going past. Now, the voice packets are going to be small packets with large headers. Okay. Data packets are going to be large packets with small headers. By looking at that also it is possible to figure it out. Which field will make it known? I will make some change in the MAC header. So, I will make a change in the MAC header because the access point will process up to the MAC level. I will make a change in the MAC header which will say which is the access category for that particular flow. Okay. I mean there are a fair amount of detail. We are not going into all the details here. But it is not hard to understand once you understand the philosophy behind why it is so. What is the difference between access categories AC3 and AC4? AC3 is actually burst data. AC4 is best effort. Okay. So, FTP for example, you may just want now we will meet the same categories later on when we do Ymax. Okay. They also have the same categories, but they are called different names. Okay. So, FTP is basically burst, bursty flow. Even if it is a let us say even GPRS has the same access categories interestingly. Even in GPRS you will find the same four categories. Okay. Burst flow means what? I have to do a lot of allocation for a short duration of time. Correct. So, just when the download is happening that particular node will require a lot of resources and then it will release the resources. Okay. AC4 means it is just best effort. If there is left over space then I will carry the reader. If the request in QAC4 is more than the request in QAC1, even then the priority is decided by the amount of back off that you are going to do. Right. So, the packets in AC4 Q are going to back off more than the packets in AC1 Q. Right. So, that is the whole idea behind prioritization. Okay. So, in fact you can do lots of experiments in this. So, many people have asked me this question. So, let me answer it in common. What kind of assignments can we give using wireless, when we are teaching a course on wireless networks. Okay. So, the type of assignments that we usually give. So, one of them is to do some kind of application programming. Okay. You have this J2ME emulators which you can download. You can say that create a location management application very straightforward thing. Okay. You download the emulator, you give that kind of an assignment. That is one type. Another type of assignments is to use one of the simulators. So, there are three popular network simulators. Okay. The free one is called NS2. Then you have QualNet. Okay. So, QualNet is the academic version is free. Okay. And then there is called one called OpNet. Okay. Now OpNet is a commercial one, but that is very hard to get. But what you can get is an individual users academic version. Okay. So, you may not be able to get one for your entire class, but you can get one. You know, you can ask each student to sign up on the OpNet site and download the individual version. Okay. So, now some of those things you can play around with these things. Like for example, the question that you asked, what will happen if I if the AC4Q is given higher priority over AC1Q or what will happen if there is more traffic in the AC4Q as compared to the AC1Q. Such experiments are very easy to do using OpNet or QualNet. Okay. So, you can give those as assignments. You can say, okay, set this as the traffic, set this as the priorities and generate a graph which shows what is the throughput, what is the delay. Okay. So, those are good ways of giving assignments. Okay. The third type of assignments which we give here are the actual measurement based one. You know, you take an access point, you install an access point somewhere, you give a student a laptop and you say, okay, go and measure what is the signal strength that you receive when you are standing away from it, away by two walls. What happens if it is a concrete wall? What happens if it is a wooden wall? What happens if it is a ceiling, floor, all those kind of things. Can? Ethereal. Ethereal can be used. So, that is all you use. You know, you run ethereal and then you are going to see what is the signal strength. In fact, you do not even need ethereal. Ethereal you need only for trying to see what other traffic is there on the network. Okay. Another kind of experiment which you can do is you can carry out ping tests. Right. You can say that I am able to see the access point. Does not mean that I can communicate with the access point. Right. So, for example, the access point is transmitting it at 1 watt. The card is transmitting at 100 milliwatts. Right. So, the card can receive the access point. Doesn't mean that the access point can receive everything that the card is transmitting. So, even though you are standing at a distance where you can get a decent signal, your voice call may not go through because voice is full duplex. Right. You need a two way connection in order for a voice to go through. So, if at that point you stand and you carry out a ping test, you will see that a fair number of packets get dropped. Okay. The delays are increasing. So, such kind of experiments can be done. Even with just one access point, maybe one laptop or two laptops in your lab, you can give out such assignments. Is there any mobility management in dot 11? In theory, yes. In practice, no. Okay. So, in theory, also a to dot 11 f is the one which is the inter access point protocol. See, once there is a protocol between access points, then you can manage the mobility. Okay. So, dot 11 up to dot 11 b g e and all, they don't have mobility management. Okay. Dot 11 b is sufficient as far as learning dot 11 is concerned as far as giving assignments to students is concerned. Okay. Some of them will also have patches for dot 11 e. If you have that, then it's great. You don't need any of the others really. Okay. Because the difference between b and a is in the physical interface. Okay. Is in the physical layer. Now, I don't know enough to give assignments based on physical layer. So, it doesn't matter to me whether I have a or not in the patch. G and b again, the difference is only in the data rates. But there is no difference in the protocol. So, if you have b and e, they are two different versions of the protocol. That's good enough. If you want to give an experiment on the power management and routing, well, it depends upon the thoroughness with which you want to do the experiment. If it is a course assignment type of an experiment, yes, it is sufficient. Okay. But if it is like, you know, it's your own research project, you are doing some PhD in something, then you may have to do a lot of work in order to use NS2 for doing that. Because many, it makes a lot of approximations. NS2 has a lot of approximations in terms of who can hear whom. Okay. So, depending upon the accuracy with which you want the result, what is the aim of the experiment? If the aim of the experiment is to familiarize the student with the technology and do something and generate some graphs and get some general understanding of it, then it is sufficient. If it is to come to some conclusions using the numbers that you get, then it is not sufficient. If there are two wireless networks in the same area, how are the two differentiated? So, one way is the channel. So, 802.11 has 11 channels. So, one network will operate in channel 1, another network will operate in channel 6, a third network will operate in channel 11. They will be totally non-overlapping. Out of the 83 megahertz, you are going to use subparts of the spectrum. 54 Mbps can still remain as the data rate. See, data rate depends upon modulation techniques. It doesn't depend upon how much of the spectrum you are using. I mean, it does depend upon how much of the spectrum you are using, but main thing is what is the modulation technique that you are using combined with the available spectrum. Where do we configure this? So, if you just look at the access point. So, if you are installing an access point, it will give you a configuration screen. Various access points close by will not be under 1 percent. So, you have to have an agreement with the other guy. So, that is why what happens is in the airport, it is not that anybody can just set up an access point. So, the airport manager will tell you that you use this channel, you use that channel, you use the third channel. So, the operators which are setting up access points have to coordinate with each other, so that they don't interfere. That coordination they will do, because if they interfere, then both their reputations are going to get tarnished. So, think about this question of how did we arrive at 22 as the number of voice calls or at least work out what are the various additions that have to happen? How much is the overhead in a voice call? We will come back and we will try to answer that question.