KpiAQdXx

1on those two chapters whoa whoa discuss our details as much as possible but that's not merely primary focus of computer network those issues those link layer physical layer issues have been actually always solved there are not really serious challenge left maybe some people do not agree with me for those who still think there is a lot of research issues left in physical layers yes they are allowed each left but by pointers and is from computer science viewpoint that is not as important as the previous changes okay so that's the bottom line by doing any questions I will ask those whether you understand those really important aspect in the exam like for example how how scalability is achieved in that's actually one of the most important issue in chapter 4 chapter 5 flexibility what sort of features can be supported in a certain mechanism like routing protocol that's also important of course can you support on your very fixed functionality or flexibility so that can be achieved in a lot of different environments okay so let me talk about the last protocol SNMP a therapy is a protocol designed for network management and what is the meaning of network management those meanings are written in the box in this slide so it includes how to deploy networks which means how to deploy routers and beings and how do you coordinate hardware software and human elements that's pretty much what they are saying how to manage the will help manage those routers and beings and the management means monitoring the situation like which link is more loaded than other links which link has any problem so test it would be similar to a monitor configure which router should be connected to which other routers all these issues these are generally called network management the purpose of network management is providing connectivity of network of course if network is disconnected no purpose no reason to have those network and good performance the definition of performance would include like faster actively smaller delay or maybe bigger bandwidth higher throughput those kind of goal to achieve those goal is basically the top management and so there are manager actively managing other nodes and there are devices that are subject to the management managing the device and there must be some mechanism for the manager to obtain information from any device and the most recent mechanism for the manager to configure changing the functional managed device so that is the protocol we will need some protocol and SNMP is one of the control protocol between manager and managing device recover the Sdn architecture there is controller and controlled switches so actually open Pro is one of such protocol management protocol and a therapy is another kind of management protocol SNMP has been around much longer than opened role so pretty much all the network devices to support SNMP protocol where open floor of protocol plays on your certain group of routers support all true that's why in stn architecture as an interface between controller and switch open floor is one option and a certain piece of adoption because most routers actually support as NFS world check this management information base this is kind of really small scale database so each managed device has the elements that can be managing or you can think it as a configurable private for example the Wi-Fi router that you use at home those community device they also support a therapy and by using SNMP you can actually change or read those configurable parameters Kamui Wi-Fi router allows change of certain parameters and let's say some other like this Wi-Fi this big project I'm not exactly sure that this view projector has SNMP clients there and basically supporting SNMP or not but some big products to support SNMP and then what do you think what kind of parameters would be supported as a computable parameter maybe his IP address maybe beam strength brightness or maybe a beam angle whatever Wi-Fi router case what do they support to function as a Wi-Fi router they must allow you to configure the private address space for example all the private addresses managed by my Wi-Fi router is a ten-point 10.0.0.0 to like 10.0 like 255 for example so those are kind of configurable parameter of each device so what features are configurable that depends on device time and fenders another example like Wi-Fi router you choose name of your Wi-Fi network officially called SSID that's a lot of kind of computer parameter bimproject unless it has Wi-Fi capability SSID is definitely not an element of MIT so MIT is kind of settle element set of parameters that is computable computer means rightful changes some parameters are just written there readable projector and read and write commands so that's the concept of management information base so as a manager what MIT what sort of manageable parameters are supported by each many device that's the basic concept once you understand this concern all the remained is in principle from this point you can easily design a sedative horrible that may not exactly same as the current standard but it will also work so let's see how this SNMP protocol is actually designed the terminology agent is used in SNMP so a any program that runs on mint device and react with the request from manager Pascal smh really doesn't have any meaning at this name or you may call it quiet so SNMP is the protocol between the managing entity and this managing devices agent and what you do is managing entity may ask value over certain features which means readable privateeer in my vision for a certain device or it may ask to change certain parameters the third type the third possible function is this managed device may detect something some synchronous events like suddenly I have problem and then it may send unsolicited report basically sending error report those also possible that's exactly the three types of communications supported as an appeal so there are two ways convey to complain MIB information again MIB is just a list of parameters a readable or writable in each device once approaches recast the response managing device may ask to a managed device to approach a certain till another way is trapped mode this is voluntary report problem many device when certain thing happens the many device reports the occurrence of the event to mention entity a typical example is like link failure this remote eating and now hugely important this paste is just more detailed a specification of what we just discussed this is how the messy format of SNMP protocol is composed and so that's the end of chapter 5 so chapter 5 we learned control plane or network layer so private functional control plane is routing so we learned several kinds of routing protocols and then we learned the concept of the new approach and then we learned the legacy control protocols like ICMP dr. the type of control protocol that exists before Sdn so ICMP SNMP these are not the only controller progress but these are sort of the most popular and widely used low-cost monitoring control layer Network layer control plane protocols so that is what we have discussed in Chapter five and then the next device link layer from physical layer discussion and they also did not make that distinction and they actually actually contains whole suspects link layer and physically in a single chapter they needed a link layer and it's hard to clearly separate link layer functionality from physical layer functionality the reason is that both typically implemented in hardware they are in intercepts the goal of this chapter six is understanding the basic principles behind the link layer service so what kind of us features what was the functionality of link layer basically that's what we are going to learn and instantiation and implementation so variously clear technology what it means is in today's internet malice that there is not just one link layer technology being used there are bunch of different link layer technologies combined with each other that's where the difference between routers and switches comes switch is a real name device that connects same type of link length let's switch for example we call it an Internet switch that means it connects multiple evil things router we don't call Ethernet router that doesn't make any sense router is fundamentally a relaying device that is capable of connecting heterogeneous link layer technologies one week maybe Internet another it may be Wi-Fi another link may be some optical fiber so that's one of the distinction between a switch and router actually the most important difference anyway there are many link layer technologies and in the in this class there is no reason to understand or discuss this all kinds of link layer technologies today the textbook focuses on two for now much wider area network technology what is the link layer technology for when originally really originally originally means like 40 years ago or even 30 years ago all the when technology of Internet was telephone networks because telephone network was already there so to design an Internet to build a little what you need to do is just creating local metal and connecting it to existing telephone networks past 30 years ago today always today there are a lot of menu while the area network service provider so they build their network by deploying so service over Internet is called UID called Skype lviv and actually all the empty telephone code they all VIP anyway so I explain the difference between technology web technology whenever I have time or chance but mostly focus focus on them this is love structure of this chapter so receive some fundamental principles first two sections and then this is this is more like physical issue and this is a issue relevant to IP network the IP protocol remember ID pour over the network layer this is the layer right on top of link layer so that's going to be discussed here and not totally irrelevant but somehow important the issue they should be put into somewhere in the testable those are listed here commonalities loads links we all know what that means and another poor knowledge that way I am we actually have used his friend so remember PDU protocol data unit is an output of a certain protocol layer so our data or IP layer is called packing in other words P do I belayer is packing and what's the result of link layer how do you call that video display frame packing segment and open x message is the video application layer so that is the meaning of a frame and packages also called Datagram so basically frame encapsulate state under Datagram encapsulates TCP segment or UDP data link layer is responsible of transferring the spray mode Datagram from one node to another node which are physically adjacent each pair adjacent means there is no router in between so link layer is the issue between two points two dots where there is no router in between that is the creation that's why I said an inkling is relatively easy there is no scalability problem there is a little bit but not serious scalability problem why because methods number one point-to-point link which means there is a dedicated physical medium like copper wire between you and your party between two nodes if there's a dedicated link and there's zero scalability issue that link is only used by you very simple case what are those cases we'll see and another case which is a bit more popular is there is a shared video between multiple dots so this medium this transmission each of these videos the transmission age of this video is basically broadcasting so if I send something my name is Sibylle others can link has this nature to you the signal will be dispersed to already know the direction those kind of Route broadcasting style link is another point when link versus broadcasting link these are two fundamentally different and why in this case it's not easy to isolate completely isolate your transmission from others in other words in wireless communication case it's very hard to prevent others to listen or receive data toward you sometimes it's good and sometimes of course it's bad you don't want others to hear you are maybe some secret messages or comes so keep in mind so link there are two types of link points going to get kindling this is almost always used in the communication between routers so routers usually have dedicated optical fiber between two most of our most terminals and usually you use broadcasting type of link although Wireless and this today is basically broadcast and this broadcasting creates some of the complications and discussed the link layer is different from that of course what kind of services are provided by link layers I mean not only real link layer physical implementation of link layer to support all of this but this is sort of hypothetically this kind of services would be desired number one framing out this is of course it must be provided by all link layer spraining means you add header and optionally trailer so that in caption you can capture what kind of information is included in being a your heather wilson and a big hint is he with you there is a concept of mac address medium access control medium access control stance the Mac is the full expansion is the medium access control what is the meaning of medium access control who access mediums at this point that is video access control so it means point point dedicated link there is no issue access control you are the only one who are loved who is allowed to use that medium so very simple it's just you rather way you like you can send test medium could access control in any hated post point link oh please share the link in other words if this has broadcasting nature and then you have this big issue to decide who can transmit who can use this medium at this point because base meat broadcasting is shared only one can use one node can use so medium access control is investors really heart and soul of link layer for broadcasting time so there are multiple nodes who share the same link and then there must be a mechanism to identify each node of course can we use IP address for this purpose yes but that make things very complicated because IP address is not really stable identifier you get DHCP is pretty complex so what the solution adopted by Internet is you have another level of eye by which is this link level itid that is called math address so each node or more accurate to each network interface actually has to kind of identify IP address and MAC address is there other identifiers yes Youkilis name so there are actually three types of level identifiers your DNS name IP address MAC address think is there any correlation is there any regularity of this mapping first of all Tina's name and IP address is there any meaningful relationship though so we have to use database testiness system data service how about MAC address and IP address this is tricky the answer is there is no correlation in my people at four so MAC address is determined by better when your data interface card is manufactured and each panda each manufacturer has their unique ID and they use they put serial number in addition to their manufacture ID as a results medium access control Mac good address is unique throughout the world the uniqueness is guaranteed and this number has nothing to do with your IP address because your IP address depends on where you use your device if you use the advice in your network or using devices or your network created us then to your IP address will be totally different so there's no way that there is any correlation ship between IP addresses and MAC address what do you do if there is no function close to home the function that come up with other identifiers you have Tuesday debase we have to keep that mapping information a table so very similar mechanism as DNS is used in the mapping between web address and IP address but it's much easier problem because in this case you are talking about global internet here talk about just computers in your life as much easier problem I said the mapping between web address and IP address is arbitrary in ninety thousand four and what about in ninety thousand six it can be arbitrary but I've avoided for support some extra feature so that there can be a static relationship between MAC address and IP address why what is that for if the mechanism the mapping can be obtained without relying on tape or database you still any advantage test natural question they are accepted an advantage this mapping between Tina's name and I'll calories this is a headache every time when we want to access a webpage we have to go through Taylor's extra overhead same thing wherever we go to sent a packet we actually have to go through this mapping between Meg addressed an IP address so you said pig feet over it at this point why we need this mapping every time we send by Peggy Wilson this is one of the most important aspect of linklater the mapping between IP address and MAC address how they are used if your does that you you need to understand the mechanism how mattress is involved in IP packet transmission okay so this address and what other services would be nice to have and big layer reliably if your big carrot is reliable delivery very nice some thick layers supported reliable delivery some link layers do not support it reliable delivery if some links do not support a reliable delivery and then what happens if a packet is lost over that link visit me or application layer so that is the difference between the incredible reliability process and the reliability disappear response to entertain the reliability and the reliability the reliable delivery over a certain name is linked rather like ogee empty based upon two driver delivery white by a supposed to drive with anybody reliable delivery are we talking about 100% reliable there's no such thing as 100% but there are certain condition reliable what else through cultural we all know what broken trees basically same concept the reliable frozen tree in TCP error detection obvious error correction audience happy to plots and food degrees this is all the important concern so so far like network layer IP layer we assumed that two nodes can send packet to each other simultaneously that's what we assumed so far but what's the meaning of simultaneous simultaneous transmission linklater IP protocol think that it it sends a packet well in the following condition the boundary condition happens IP layer which sends a packet what is the following condition you ask link layer after you send this one if you call a function called a send call from link layer and then you think you send it but really the packet is transmitted you don't know IP layer that's not any oh I believe you so when I feel I I sent the packet that means I just asked the link layer to send this packet that packet may be stored in the link layer Poppa forever for still like today's viewpoint I set it easy creo it makes you just more confusing okay anyway so when we say IP layer is full duplex full two plasmas both center of the to communicating peers can send their data any time as full duplex in both direction can you send can both a and B when a and B are connected and alb can they both send package to each other exactly simultaneously physically and depends on the capability of your link for example if you have two separate copper wire possible because there's no questions waiting like Wi-Fi Wi-Fi is not full duplex communication supporting link only one node can transmit at a certain moment so why is by nature as help you please how about empty and these full duplex so in all depends on link layers that's what it means whether a link is half duplex or full duplex the last type of this communication sometimes simplex simplex means one directional communication on you one-way communication possible Terra technology that only allows simplex communication so noticed where this link layer is implemented I already said it is pretty much hardware so this is like a network interface card your network interface card is connected to the system was typically and inside this network interface card there are couple of chips and link layer features are implemented there and how this is connected to system software operating system that is called device driver this is particular device may be made by like three comp or made by Samsung I don't know whether Samsung makes even a card whatever and then each operating system each operating system implement a driver software to control interface card so for each metal interface each type of network interface card you need corresponding device driver that's why whenever you install on new you cut or new hardware in your desktop you are asked to install device driver device driver is a really really simple cocktail it appears compares what is very simple because all it does is just matching the requirement from operating system and support the capability of your device so your device as a certain design when you write some value in a certain memory address say older more typically the network interface card has a certain memory space so when you write a certain value in a sultan sultan no dress and i do a certain thing maybe like turn on my network interface card i don't want to go into the details on this operating system issue here so basically for that reason it is combination of hardware software firmware the precipitant formula in hardware came not to really primary topic of our class so the top three layers are obviously implemented in the computer side on the operating system and physical layers they are all implemented in hardware and link layers most of them are and some of them are implemented in in software and device driver is kind of that portion of the Clio this is how it works so it's a computer CPU memory let's go get the pace car CPU memory cathode is the physical they are connected by a link so when that top layer in the CPU the OS cannot ask the link layer to send the Datagram that will create a friend and will send to you help the punch and your counterpart will T calculate the frame and extract Datagram and send it back to the network layer off the right hand side computer and you eventually delivered to the applications any constants so far this is not typical it's just pretty simple concept you framing and sending and the main difficulty comes in those medium access control we photo queue medium access control let me briefly explain the error detection issue this is actually discussed in digital logic design class or discrete mathematics class control science introduction class so I don't really feel the serious necessity of discussing details of this issue so let me quickly go through so how do you detect L you have to add some extra information about the data that you set and if there is any alteration if any change in the main body of your data by comparing the external that you you add it you can detect depending on your error detection skill capability you may be able to tonight one bit error or two with errors or make multiple bit errors one of the easiest mechanism for our technicians parity checking I'm going to skip this this is used in memory the most basic form of error checking and we learnt checksum in audio charger so checksum can detect errors in a sequence of about B's straight for the extension of a penalty mechanism this one is a newer cyclic redundancy check or CRC for short this has been used by in many technologies like Ethernet and Wi-Fi still used it's not really the most advanced error detection mechanism today much more powerful and fancier error detection method is actually use Wi-Fi you're not using CRC anymore is much what much pair of mechanism LT they are not using CLC it's much complex much more efficient effective error detection design and thank you so please the the error detection or error correction hope this skill design itself creates a huge academic field which is open x called information clearly information here is not ill with not only dealing with the error detection but that's huge part of information theory in the computer science curriculum we are not teaching informations theory in depth so let me quickly go through it is relatively complex and relatively powerful of error detection mechanism why do I say is relatively powder see this line it can detect old post errors less than R plus 1 bits what it means is if this is your digital real data and you get all beads this is extra information for error detection and if you get these alby's somehow you can detect any else less than R plus 1 bit slow so if this is like tiny bits you can detect any first care of consecutive errors up to twenty one tiny bits none was there all cuz that's easier to detect first error is hard to take the previous mechanism like checksum you don't have this kind of capability so in essence this is much powerful much more powerful than checksum so how can this be achieved first of all what you do is you choose a pattern G this is sort of a design decision this is my G this is a CRC and I want to use a certain G G is our plus one bit and then so T is keep this is the data that you understand and G is already decided this is constant for given D ng you compute the other and you add that up and then this is possible so question is how do you compute R and that's the question and this is how how is to compute it so G is given and you make ter this pattern exactly divisible by G that's how we really so if this pattern is exactly divisible by Chi no health if they are not exactly divisible there's Adam but how do we how do we compute out from these relations okay this relation is basically same as here you shipped parity by ugly space here and all these are zero so you saw without and this is what you care and that value that you send is exactly divisible by see that's the meeting only need to do this get this out because P is even G is given very simple and how how do you do it you convert this wait is this what you get just moving the thumbs so what you do is you divide TIG basically shifted D by Qi Tian ji is given this is you are so pretty simple it can be done relatively simply and at the receiver side just check whether that is divisible by G the parent that you receive is divisible by gene is detail and not really get important has been around like 30 years and pretty much in all the technologies this is implemented as so deport feature and you know young computer science scientists understanding this old CRC in real detail level I don't think is that important very low small chance that I may ask you to compute out like in the exam okay this is T this is G what is your that's not computer network exam that may be digital logic design exam or discrete mathematics course exam cus does not computer network positive for those who could not come to this makeup session I am really sorry for if they spend too much time on you know studying this this binary division problem okay the next hour section is multiple access protocol this is the core as I said the main competitive the link layer and this for wireline land is in the medium access control and we are going to talk about that issue here by doing today the Taurus session would be three hours long so let's have I wanna help our and 10 minute breaks and then maybe one and 20 minutes so second half since I said there are two types of links points point link and broadcast bingo share the link the example of point-to-point link is between routers that's what I said as an example another example is this dialog access because in Korea nobody use fire of access anymore say good sound very strange to all of you kiona access is basically using teleport line to access internet if you have no other medium that connects your house to internet you have to use existing line and all of it is right is purple light and this is called this use this use a device called modem anyone ever heard of modern when I was your age I had to use this model the speed was 10k meters and after few years later it became 50 KB is fabulous anyway that is spiral of access and this is point pointing because your house is connected to telephone office by a dedicated copper wire there's the designer telephone network there is no medium access control you you're the only one who are using the copper wire really really expensive design now going to an end the PPP's the name of program may hope link layer protocol used for this kind of points poisoning so there is no medium access control issue in P P P P P protocol and you're not going to discuss the P P point point link case in this chapter we are going to focus on broadcasting what are the examples of this broadcasting internet Wi-Fi and this is the the h FC fiber that means fiber H means - whatever that means basically the typical method that you are using today in home internet access satellite communication points up oh this is not a popular method available in some larger countries lot o M is sparsely populated countries installing expensive optical fiber is too much that case satellite communication is is one way and this is also shared medium completely shared medium but that's through to a software and I don't want to make you guys too much into this issue of a wireless communication at this point we'll have enough chance to discuss those issues in chapter 7 so multiple access protocol multiple access means there are multiple nodes who want to access this video share the video so there must be some mechanism to control to coordinate the excess of this shared medium that is video access protocol orbiter access control protocol protocol and pacing they chose Mac protocol is if there are more than two simultaneous transmissions and then post transmission will fail because the signals will be mixed so this is what you want to avoid those situation two signals mixed with each other and both resulting erroneous communication that is called collision you may have this question when two signals colliding each other both will always pain what if people one signal is much stronger than the Odyssey for example you speak with your friends next to you if you speak really loudly and then even if your boys collide with some others voice you can still succeed in communication so collision does not necessarily mean error what I don't want to go into that territory because that's another a deeper level issue for now collision how can we avoid collision there are many easy ways if there is a central coordinator who decides you talk this time you are next and then there's still chance of collision is there such a skill yes MP 3G 4G most of the several technologies basically using this method how about Wi-Fi Wi-Fi doesn't have those central coordinator then you may think why ap base station does not do them in pink layers like Wi-Fi support distributed algorithm not centralized because centralized operative has a cost and this domestic origin has some advantages so it's a design decision okay this big crater technology how we go to to each one to use centralized approach over this widget approach and this to be distributed approach has certain needs some great benefits so many local area network link layer technologies adopt distributed algorithm and even it is one of them and Wi-Fi is another and there are many others one more issue even if you have like distributed algorithm or centralized you need some messages to this issue we exist it's done inside the same bed for data communication means you have a separate Wi-Fi in bed to see money why certain technology technology choose out of ancillary there's good reason for now it's enough to understand that what sort of design spaces are available for a link layer designing for example if you make a company start a company and decided to implement a really fancy new link layer and sound into to others and then you have to make decision do I want to use this for point spoiling or broadcasting and I won't make it to distributed way or centralized raised embed the signaling I'll go and see if you don't know the design space you cannot make good product it's always always understanding space problem space that's the phone number to always be able to petitioning be able to position him a certain technique in this space you must be able to position in existing technology existing skills in this space and then you can see where our solutions where is the area that people do not have solution yet very simple you design the solution for that space things are not that easy because everyone else is doing the same thing and if there is some empty area there must be reason whether there is no demand for the kind of technology or it may be too expensive many reasons but if you don't know the space and positioning a certain technology you have zero chance just pure luck you make something and just by luck it is too good and nobody has done it before highly unlikely okay so what is the ideal multi access protocol and how and we have like broadcasting channel that's the basic condition necessary condition for the needle medium access control and this maximum radius R because the inka pieces are weakest what is the desired feature this is Latin words meaning the desired things so what is desired if there's only one node who won't transmit you should be able to enjoy the maximum capacity maximum rate that's one of the requirement to if there are n nodes who won't send and then they must be equally divided the available capacity on average unnecessarily it's strictly so this is another or desired or feature hopefully it's fully decentralized so that there is no single special note that coordinates the transmissions because if there is a centralized coordinator and then have potentially single take a point this would be even pal there is no clock synchronization if your skill doesn't require any clock synchronization that will be better and hopefully your design is very simple so now you I want you to evaluate each existing today's technology how well they support they fit this desired features number one this is no patron this is hardest to achieve if there are multiple nodes and you divide them equally by the number of contenders this means there is no transmission opportunity loss in other words if there is any collision and then your R will be wasted your total capacity wasted because of the collision collision means you have to be transmitted or and miss that transmission is lost so this really hard touch you although the hard thing to achieve is a spot synchronization without synchronization a synchronous system is always more difficult and synchronous system so generally speaking this is relatively easy to achieve and many techniques many link layers have problem in number two and many technologies also seek radiation okay so the existing typically divided into these three classes lower channel auditioning type of technology what it means is you divide the resources into pieces and you assign this piece to the note that want to transmission typical each other petitioning requires a centralized coordinator do you have to I said typically that necessarily but typically it requires a central coordinator who keeps track of which slot is used by hope so then you can assign unused slot unused resource piece to the new user who who who who want to okay random access style is more popular method in local area networks so what it does is chatter is not divided and collision can occur because there is no centralized coordinator and the really important aspect is this one how do you recover from collisions when collision happens yes that's already happened so you lose that transmission opportunity but you somehow adjust your medium access control behavior so that next time cohesion occurs does not occur all that occurs with smaller probability taking puns this is kind of similar to channel partitioning but typically requiring the central coordinator so that the older participating nodes take tons take turns means have transmission opportunity so fairly at first to use them and there is a rule that if someone send then might one is next so the water that each node can transmit is predetermined that's taking one approach so let's start with channel partitioning number one method is PDA this is really really familiar method because this method is used in CPU scheduling so only one job can be done can be solved at a certain moment so you divide entire time into time slots and each time slot is given to us your job is transmission request so this one times one has a fixed length and depending on how do you coordinate like for example if you keep just one time thread for each user there then that user if it doesn't have any data to send and then that time slot would be wasted but there is no transmission over that time slot so undo Stanzler can go either tiddly maybe extend it a little bit so that this unused times that may be used for some other purposes test possible variation but in general no entire channel capacity is divided in time dimension and allocated to a user that's the meaning oh that's the basic idea of TDMA and obviously you require clock synchronization everybody must know whether the pants lad starts and in this case committed would never come but you have high overhead for coordination very similar method FDMA but with the same as the logical same as the TDMA the difference is now the division is implemented in frequency may not handle a second video chance support if you link and suppose the transmission of multiple frequencies frequency means our points it use a certain frequency our I use a different frequency all the TV stations they all use different frequencies so basically the the radio wave or the sound wave or optical way they all have this concept of frequency it can be applied to a wire line copper wire or both certain copper wire you can set the multiple frequencies depending on the thickness of copper wire how many frequencies or the lowest threw it at the lowest frequency and the the highest frequency that region is depends on the size of that deliverable frequency is determined by the thickness of the copper wire and same thing for optical fiber depending on how thick or patear the company the fiber is the radio frequencies that is deliverable over that we can determine so this question table is usually used for TV or cable networks and over this cable there are like hundreds of TV channels and they are all transmitted in the Pomo FDMA and very simple same mechanism can be applied to a communication Internet communication so actually the the internet access over cable TV this method is used fiber optical communication between routers this method is used this is why they are used in wireless communication of course random access protocol I said random access does not require a priori quotation which means there is no coordination before the transmission everybody I oppose their own medium access control there is no centralized coordinator that's random access randomly access naturally collision occurs and there are varietal random access video message control protocols and they keep how to deal with this collision there's all of this in the following slides so that start will slotted Aloha what is Allah this is hello in Hawaiian and why this protocol has his name is because it is designed by University of Aloha people Universal Hawaii people and there is original alone and there is better more sophisticated extended although I test slotted Aloha so many texts will introduce a lower cost but this although somehow decided to introduce slotted Aloha first and to learn track it down but first the design assumption or the design a rule of alloys slug terrorize like this all friends have same size the prince means that the transmission the Union the PDO of link layer they what they all have same size all the time is divided into equal slots equal size like kicking me they are all divided but this is not feeding me it's just fine is divided and everybody knows everyone has time same time club ok secret those are synchronized so they know that when is the allowed starting point the transmission starting point everybody knows them they don't know whether who will send frame at that time slot they don't know but they know that if someone or send data they are transmission with that that moment so that timing information is given so nodes can start transmission only at the beginning of the slots and when more than one notes are transmitted in the same and everybody occurs what do you do so you sent you p.m. data you send any time slots you want if you're not you there will be no collision fine if there's collision and then you retransmit that frame but not right next not at right next sluts because if their collision at one slots you will have packet loss the one who collides with you will also have fabulous if you and him both try transmission in next following sluts your tube it's a really really bad idea how can you avoid them if you communicate with the collider and you negotiate them I will stand next time so you will send like three times later that would be nice but that requires over it natural solution is random tossing so I pick a random number from one to five so my that is to the other one hopefully he's not waste not - there's possibly - he also choose to but the chance is not very high particularly if your random number span is bigger like 1000 then where's the chance of - Lord choose same value out of 1,000 really well so that's the idea you choose the subsequent times run with a certain probability this is a different type of implementation so basically same idea this probability P you want to make large when contenders there there are small number of contenders you want to make this piece small if there are many other contenders how do you know how many contenders are in your network not easy but there's one way if you keep experiencing packet loss if you keep experiencing collision then what that means it means there are more competitors so in that way you can actually bother to be this is a slide that illustrates how our swati to our lower works some advantage is if you are the only user you will fully utilize the entire capacity so satisfy the desired feature number 1 in previous slide what is the overhead in other words what if there are multiple users and can you divide the network capacity by the number of users without any loss no because you have collision you waste something easy fail the fairness is another aspect of the number 2 requirement it requires clock synchronization it's on distributed no central not centralized so in essence is not bad one issue is this so how bad how much papa was lost I was capacity loss do you have in this mechanism Swati download so I'll open em the Toronado capacity divided by the number of users without any loss test is your ideal your Holy Grail so hundred percent utilization for example maybe not the worst what is the utilization of all maximum efficiency or a little lower 37 percent which means this is mathematical result how you compute this mathematical result way through this this is pretty simple probabilistic it can really a simple background on probability theory you can easily follow this so 30 percent is the expected average utility loss slotted Aloha which means you know you wastes over 60 percent of network capacity is this a good system there's 40 percent is what about detailing if you use CDMA and then there will be 0 collision but you need some control over oil so let's take control always 10 percent of the network capacity and then if you use killing me you get 90 percent of efficiency or 90 percent of utilization so key theory is much better in terms of this resource efficiency now why not use PDA why do you slut it all or it has some advantages like well there is no centralized coordinator you basically need almost no knowledge but you pay the price do we use what little lower today what do you think there are better because like theory is this there is any chance that this one Allah with 40% efficiency survival yes a used everywhere like LT is not used for main transmission purpose but this is what you use when you start something when you start something when you turn on your phone your smartphone you have no idea first of all you have to contact the base station to tell me your presence I just turned on I want communication how can you tell them base station does not know your presence page has no chance to allocate time slot for you because they don't know if he does know your presence yourself you all right all over yourself through the base station so any time if there is no prior information come on sink rate so this is easy this was not required cultivation why because in station or satellite periodically broadcast clock information so you can easily synchronize your clock to the system without camping without any help from others so clock synchronization is relatively simple relatively simple requirement in wireless communication in wire line communication different story when Wireless come because this is relatively easy society the law is in that sense even if it's not really efficient in terms of throughput or utilization nature's is still used now let's compare this with original pure Aloha pure Louis but slightly there is no clock synchronization so transmission can start at any point like this