FOG COMPUTING- I



in a previous lecture we discussed about cloud computing and its importance in Internet of Things so cloud is very important because you see that Internet of Things IOT devices sensors RFID devices and so many different types of devices they send so much of data and finally those data have to be handled and that is the reason that cloud came into picture that all these data will be sent to the cloud for further processing and so on and so forth now the main problem with cloud in Internet of Things environment is that latency so what is meant by latency I will explain to you before we go formally about discussing the concept of fog is that let us say that we have a cloud and we have different IOT devices deployed now what we have discussed when we were talking about cloud computing in the context of Internet of Things is that each of these devices they sent all this data to cloud for further processing and storage for further processing and storage now that's a problem that's a problem because you see these peyote environments this IOT environments number one are constrained in different ways with respect to bandwidth with respect to processing with respect to memory with respect to energy and so on and so forth now this processing can be handled with the help of cloud but what about the bandwidth what about the energy consumption because what is going to happen in this sort of scenario of use of cloud in the IOT context is lot of data is going to float all around over the network lot of data through the internet are going to be sent to the cloud and that will unnecessarily consume the bandwidth that will also consume unnecessarily the limited energy that is resident in all these devices and so on so we do not want to do that because communication consumes most of the energy so we do not want unnecessary communications to take place and even if we do we have to limit even if you have to communicate and that is required because in a network basically IOT is basically a network so you network communication is required but how do we handle it efficiently this is what we are going to discuss in this particular lecture so can we do something which is better than cloud so this is where fog comes into picture fog was introduced by Cisco and so it was sort of like an idea about how to Bing the cloud facilities close to the IOT devices because as we saw that you know we have to have you know all these data sent to the cloud that will you know that will not only take the bandwidth-limited bandwidth that is there in this kind of environment but also that is going to take a lot of time so that is going to take a lot of time so in this particular case the time that is required will be the time from when that event is sensed that piece of data is sent to the cloud so this one let us say this is t1 this is t2 then the time for processing t3 and finally that response will be sent back right so t4 so the response will be sent back for maybe actuation or something like that so this basically becomes T 1 plus T 2 plus T 3 plus T 4 so these this is the total time that it takes until when the receiving device gets a signal about what to do and by such a time in most of the real-life applications of IOT by this kind of time the you know most of the you know the events most of the unwanted events would take place for example if it is a surveillance application maybe the intruder by this time because there is so much of latency that is involved by this time the intruder might have already entered into the territory or if it is a medical emergency scene scenario by this kind of time the time it takes to send it to the cloud cloud processing it getting a response back etcetera etcetera so you know the real-time illness is going to be lost and bye-bye because of this particular issue what is going to happen is if it is a medical emergency situation the patient might die right so what is required is can we reduce the latency can we reduce the latency and this is what we are trying to do in from computing so as I said that form was basically proposed by Cisco and the whole idea is can we bring the cloud facilities the attractiveness of cloud closer to the IOT device layer and the whole idea is to solve the problems that are by cloud computing for use of IOT for data processing so this is the whole objective of fog computing and the whole idea is also to reduce the delay that is incurred in sending the data from the sensor device to the cloud from the cloud getting a response back and actuate in the particular device so can we reduce this particular time so this is the whole idea the whole premise under which the for computing works so conceptually what we have we used to have in IOT and cloud you know this is the device layer where all these IOT devices the physical devices operate this is the cloud where all the data are sent for processing and storage so what fog is saying is it is going to be sort of like a middleware or a middle layer rather where some of the computation some of the processing some of the storage at least transient storage is going to happen before the data that is sensed by these devices are sent to the cloud before it is sent to the cloud can we do some intermediate processing intermediate storage for you know quicker decision-making this is the whole idea behind the use of fog computing so let us go back to some statistics some of which we you know we also went through at the beginning of the course so we know that nowadays because of all these different sensors etcetera etcetera there is lot of data that is floating all around so it is estimated that by around 2020 40% of the world's data will come from sensors and 90% of the world's data will be generated only during the period of last two years okay so in ninety percent of the world's data were generated only during the period of last two years so you know it is also estimated that everyday about 2.5 quintillion bytes of data is produced and the total expenditure on IOT devices will be about one point seven trillion dollars 20:20 so given all these different statistics we now have to think about an architecture of Internet of Things wherein we can use these all these different devices in a scalable manner such that the processing happens with large number of devices in a quicker manner and quicker and in an efficient manner the total number of connected devices sorry the total number of connected vehicles worldwide will be about 250 million by 2020 as per estimates and there will be more than 30 million IOT devices again as per estimate so the amount of data that is going to be generated by these IOT devices is obviously quite huge so how do we handle this kind of huge data I mean one way is basically to use all these big data analytics and so on but even before that can we reduce the processing time of this data can we do something from the network point of view and that is where we have to take help of form so what we need from computing because cloud has certain deficiencies you know it is insufficient to handle the requirements of IOT so there are issues with volume of data that is produced by you by the IOT devices the latency that means the time that it takes for a sensed data to go to the cloud and then come back that duration is the latency and the bandwidth bandwidth means that how much you know how much data is going to and how much the channel is going to be occupied because of this communication of all these data from the IOT devices so the computing architecture looks like this so we have all these IOT devices we have these IOT devices and we have the cloud so as I said traditionally we have to use these IOT devices to sense the physical phenomena occurring around them send the data to the cloud and get a get an action or comment back so as we can see over here that this is the traditional cloud model and so why do we need for computing because we want to reduce this particular time and typically these cloud servers might be physically located even continents away means you know it is difficult to have them in different cities and so on but if you continents away also so this physical limitation also introduces large latency in communication so in terms of the data volume it is estimated that by 2020 about 50 billion devices will be online and presently billions of devices produce exabytes of data every day and this is this big data right so so many so much of data is going to be produced every day not only every day every second every minute so much of data unusual volumes of data are going to be produced because of the introduction of Internet of Things exabyte exabyte means what 10 to the power 18 so you know we have gigabyte terabyte beta byte zettabyte exabyte right so really we used to use traditionally with gigabytes of data maximum but now it is with you know Internet of Things and so on it is very common to have terabytes petabytes exabytes of data that is produced every day and to be handed to handle this kind of data volume so the device density is also increasing every day so the current cloud model is unable to process this amount of data so the private firms factories plane companies they all produce huge volumes of data every day so if you look at this particular figure this will be clear so we are not talking about data produced by a single firm we are talking about data that is produced by several civil firms you know and the different devices that are used in those forms the IOT devices the embedded systems devices that are basically used in those forms of civil you know huge volumes of data are produced every day by a single firm and definitely large number of firms also produce large volumes of data factories the same kind of thing factories also produced large volumes of data European companies airplanes themselves have lot of different types of sensors they also produce large number of large volumes of data so all these data would have to be sent to the cloud for further storage and processing so the current cloud model that we have already gone through in the previous lectures on cloud computing cannot basically store all these data so this data that is produced the raw form of data has to be filtered before the data is sent to the cloud so this has to be processed filtered before it is distant for storing and processing in the cloud in terms of latency lot of time being taken by a data packet for a roundtrip and this is what I was explaining to you with the diagram that I showed at the outset so an important aspect for handling time-sensitive data is basically to handle this issue of latency because if it is time-sensitive it is real-time data so time is important and that is the reason why latency has to be handled with special interest if the edge devices send time-sensitive data to the cloud for analysis and wait for the cloud to give a proper action then it can be it can lead to many unwanted results so while handling time-sensitive data a million ii can make a huge difference so look at this particular figure over here so ambulances then different buildings and different other you know different other devices and so on cars and so on so they they basically generate data which are time sensitive in nature so they basically generate data which are time sensitive in nature and so that is the reason why they have to be processed ready fast to be able to use the data in a meaningful manner so they have to be played process pretty fast so ambulance for example will generate some data so the time that it takes the data for the data to go from here to the cloud and come back this can be represented with this you know this can be shown in the form of this kind of equation so written c equal to the time it takes for the data to go from the device that means the IOT device to the cloud plus the time for data analysis plus the time it takes for the data to travel from the cloud to the device so latency will be increased and when the action reaches the device accident may have already occurred if it is an emergency situation or a connected vehicle situation so this is the reason why fog computing is very important so in terms of the bandwidth the bit rate of data sorry in terms of the bandwidth bandwidth is calculated as the bit rate of data during tons transmission so if all the data are generated by IOT devices and those data that are generated by these devices are sent to the cloud for storage and analysis then the traffic generated by these devices will be saying simply gigantic so these IOT devices are going to consume almost all the bandwidth because of this and handling this kind of traffic will be simply a very hard task so billions of devices consuming bandwidth if all the devices become online even ipv6 or IP based technologies will not be able to handle the facility to all the to provide the facility to all the devices and the data may be may be confidential which the firm's do not want to share online so they say these are the different problems one is the the privacy of the data this is of concern to the firms the second is that you know dealing with these kind of problems with IP based technologies like ipv6 is a problem and also the issue of having millions of devices consuming bandwidth so you know how do we handle them together in a synergistic manner reduced latency of data appropriate action at the time right time prevents major accidents such as machine failure and so on so a minute delay while taking an axle makes a huge difference and this is what I was explaining to you during the medical emergency scenario a person might die a person might completely die if you know the decision-making takes lot of lot of time compared to the time for sensing so it has to be the time for reason making that means processing storage etcetera should be conformant with the time for sensing so the time it takes for sensors almost after something is sensed immediately they are after it has to be disseminated and corresponding action also has to be taken in real time the security IOT data must be secured and protected from the intruders data are required to be monitored 24/7 and appropriate actions should be taken before the attack causes major harm to the network and this is what I was explaining to you and the operational reliability the data that is generated from the IOT devices are used to solve real time problems the problems of integrity and availability of data must be guaranteed and one ability and tempering of the data can be hazardous processing of data at the respective suitable places data can be divided into three types based on the sensitivity one is time sensitive data second is the less time sensitive data and third is data which are not time sensitive at all so this kind of filtering has to happen with respect to the sensitivity of data so extremely time sensitive data should be analyzed very near to the to the data source and data of which are not time sensitive will be analyzed in the cloud so time sensitive data closer to the devices non time sensitive data send it to the cloud monitored data across large geographical areas the location of the connected IOT devices can be spread across a large geographical area example the monitoring the railway track of a country or a state the devices are exposed to the harsh environmental conditions additionally as well so when should we use fog if the data should be analyzed within a fraction a minut fraction of a second if there is huge number of devices in the network if the devices are separated by a large geographical distance or if the devices are needed to be subjected to extreme conditions so with this we come to an end of first part of lecture on fog computing in this lecture we have understood what fog is the genesis of for computing and also about how for computing can help in building Internet of Things systems we have also in the process gone through some limitations of the use of peyote

7 Comments

  1. Nangialay Nangial said:

    very nice sir it is very clear

    May 22, 2019
    Reply
  2. Raja Ram said:

    Easy Understandable……simple points and easy to remind…thank you sir

    May 22, 2019
    Reply
  3. Shaibu Yahaya said:

    Great explanations. Thanks.

    May 22, 2019
    Reply
  4. adil rahat said:

    thanks

    May 22, 2019
    Reply
  5. Асем Махамбетова said:

    So greate description. I love it. Many thanks

    May 22, 2019
    Reply
  6. João Cardoso said:

    thanks for a great lecture

    May 22, 2019
    Reply
  7. ROJALIN JENA said:

    sir,thanks a lot for ur cooperation

    May 22, 2019
    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *