IoT with serverless computing – THR2117



so first of all thank you by coming my name is Georgia my I'm founder of praise the clubs we are a company from Brazil an IOT company and we are starting some operations here in North America in Calgary Canada we do IOT since the last four years so here today in this talk I want to to take some approach that we have using using more as your functions more server less computing instead of other services that we have in nature environment for IOT management management and IOT data ingestion so do we get started what is server less surveillance is an idea of server abstraction so I I can't handle the things about the servers their operations the deployment and and I have a simple environment that I can just deploy or my code or my workflow when I talk about large caps for example and it's simple it's simple because it is even driven so when we talk about IOT we talk about some events that will be handled for example I ingested data in this come from a channel this come from a specific device so I have to take an action about that or I have to take an action when I pick up an image from a camera and I send it to my blob storage and I have to pick up this image Ronna cognitive services on that or something like this and then I process this information so it's easy to do and it's very scalable I don't don't have to think about all I have to manage another VM I have to manage how I contract the server or know nothing about that it's just use and it is scalable by by nature another thing is when I talk about this model the building model of this is buy time and research usage it's not I don't have to play a palapa form for example as I have in the web jobs I don't know if here in the audience we have anyone that use it and web jobs for example that I had to pay upfront and sometimes I have to scale it manually no now I can just use it and I pay for the the seconds there are you in the memory that I allocate on that so that's the main idea when I talk about serverless some of examples that are I'm talking about is about Escher functions but we have another stuff there is surveillance also this is the the component of the the plataform I'm talking about even read for example that can manage all events and trigger something about a lot of services for example I can handle api's I can handle web hooks they can handle even read IOT hub and can use that as an income and I just route it to another to another point and it's very useful another thing that I can use it's the large gaps large gaps is the most simple that I can have it's a work flow is a dragon drop environment then I can just say oh if I receive a message just process it and send a tweet for example it it should be a the most seen example that we saw when the people takes the first house berry pie for example and said oh I will take a signal from or I want to tweet and turn on a lab for example you can do it but it's very very simple but you can do more complex actions for example you can use your CRM you can use your database you can use your line of business in and integrated using and workflow and using drag-and-drop the the propose that I want to show this is sometimes when we talk about IOT we are talking about digital transformation or about integrate our physical world in this digital world but it already exists sometimes I have all the components all the variable is a red on the on my field so when I have it I can't change it so I would take another sensor for example to measure the temperature when I when I have here in this place here is very very cold but over there know I already have sensors to do this and I already have some systems that can control the air-conditioning what I need I need to just connect the dots just a glue chew to say oh I have a sensor here I have to do the work flow and when I need to do that if I have these tools I can do it cost less because I don't have to develop some things that I in a normal way I developed so I have here some scenarios that I can use several layers applications for example when I trigger a function to run a cognitive API I told this in in the last moment you take a picture of a face for example put it on on your blob storage and I write a code to get a trigger when a file getting's in my blob storage pickup that runs the api of cognitive services and take the the the landmarks of face for example give me the emotion of this picture and send a message to someone send a tweet Sandra and Mao save it to my database integrate it with another another service if I have to integrate it with another service for example an external I can use for this action the even read for example and combine it with the their function here I have an ops automation oh you have a function and you need to run a compliant check when you create a new sequel database for example and you can run it you don't have to pay our workload you don't have to pay a machine to do that you only pay by use you only pay if this guy over here takes two or three seconds it will pay two or three seconds if it pops the the two gigabytes for example bytes of RAM for example and when you run this this function the the billing model is not so complex as as most of people see but you you can go to the azure calculator it's very very simple and another one when you do the third-party integration when you have three part services and you just use it to connect the dots here you have some some examples in the middle of these guys here is they even Creed you have some measure functions on the on the end here but you can just change these actors because you can write a small part of code using Azure functions and do these disconnections if you if you want you have a lot of three guys today they are announcing in in ignite I believed that for two so you have a lot of options here and the benefits it is first less operation less ops I don't have to think about servers I don't have to my deploy is very very simple I have I have some to develop it inside my my visual studio my visual studio code you can run these guys that I told locally you can run an azure function locally to the plot you to debug it in the in your machine and then you can open a tune for for integrated and make some tests it is focused on business logic when you when you talk about business logic is simple to develop no more extends codes I looking just for that feature there is more portion than at the atomic action and reduce the time to market so it's easy to do it's easy to deploy it's cheaper to develop and cheaper to test it do not requires an extensive effort to do so you can do into your real team on the your line of develop and this is a big shot of where you can integrate asier functions the the main idea is asier functions you can put in the middle of any anything that you need to process any data when we talking about IOT for example we can for integrate some data we can develop some alarms set some thresholds on detector anomaly in the data that incoming so we have a lot of options here to to develop this guy in the middle now we I don't know who here is using a ready IOT and who is developing IOT but now we have the IOT edge and on the IOT edge you can develop the same measure function that you use in the azure you can ambit on the IOT edge as a module so I can use the same code and I can use also a deploy over-the-air so it's very very simple and I can have the edge processing the power of edge processing using Azure functions as the same you have today here is on a an idea when I talked about the event read event read is a mirror that I can get some topics from these guys over here so I receive a blob storage I see some some problem in the field from a device using IOT hub I have a problem in my address subscription I have some some problems with storage and I have to to handle any action I can do it here as an ingestion of this guy and then who subscribe it or or where it will route this this problem or this message or this data I can handle this event in the other hand so this is the the main idea when we talk about the the ecosystem of surveillance computing using Azure of course and our our top here today is IOT with that I don't know if everybody here knows what is the IOT but when talk about IOT sometimes you can think it's only harder it's only connectivity it's only software it's only take a system module for example the the MX heat that they are they are giving here in the event to test it's a development board it's only pickup deaths send some data to the cloud and it's not so so easy but IOT today is everywhere everywhere we look we have some sensors we have some actuators we have some software and it and using this to to manage something I talking about facilities are talking about insurance I talking about new cars we have some trackers and a lot of good stuff so IOT is everywhere in with when we talk about IOT is everywhere we have this shot here we have IOT platforms that's the the idea that a simple one a simple architecture if we can say there is a an architecture graphic but here I have an applet form we have some communication from the device to this guy so I'm talking about maybe I have a gateway maybe not maybe I have a device that can pass the data through directory to the to to the cloud and I have some problems over here and I have some problems over here the question is how I can do a solution like this be first of all cheaper because costs is a problem if I have a higher cost in my solution I can put it on the field when I talk about let's develop a device so create a hardware do the production do a lot of thing it's not cheaper it's not easy to do when I talk about now let's let's process the guys over here let's create ambit software create official intelligence put it on the edge it's not simple to do you can do it if you have developers if you have engineers you can do it no problem about that but your time to market is very high that's the problem because if you generate costs so when you when you think on this a snapshot and you add to this the idea of used the server less you can develop it it's very very fast so let's see some real use cases because it's it's beautiful when we talk oh I ot I can connect it and read some sensors I said I I turn on a lamp I have a button I have a button that back something on on on a marketplace but what is the real use so we have some health problems and real problems that we have to solve and on this scenario over here we have a near real-time requirement because we are talking about I have a sensor and an O and if if it loses the signal the patient can die so I can lost a life if I have something that I'll I will put this in a an alzheimer patient that will be on the run – on the mall and I have to track it I have to send an alarm if I lost this so I need intelligence on the edge and all of these cases I'm talking about you can amber this using some new code you can do some alerts you can do a lot of other stuff here in this moderate a smart retail you have a bunch of solutions that you need and some and most of them you have to process near real-time or you have to process the business unit specific case oh I have if you if you receive in tracking a package you need to send an alert but only if it it comes from a country for example you have any specific logic you have to develop it so it's easy when you do this this connection using this these services in those three factories you have the same problem here is more the the processing on the field then there are other ones because you have several restrictions you have several problems so you need to to apply some some new approach and the end of the the idea when we talk about IOT is the smart series we can't talk about IOT and don't say all this DM this is the the idea that we want to archive in the in the next year's so IOT M serverless is a good in a very good idea so we have good benchmarks low cost per pharmacy many many triggers already working and and now I guess this is the really really spotlight that we have to take a look you can take it on on the edge until we finish I have to I want to take a case that we that we are developing over there it one of our our clients that we are developing a prototype now we are running this in some in some stores and we take some information using cameras and processing on the edge and in the cloud here I have some informations that I have H motion and gender this is are using directly the the the cognitive services and when I use the cognitive services I have an approach that if I send all this data that I am capturing here to the cloud I have a higher cost here I have to streaming imagine 30 40 50 people coming to the store and every hour so I have in a day thousands of small image that I have to process what we are doing right now we are using Azure functions we are using the cognitive services and we are using the IOT edge over here and starting a POC to using this guy we are promoting low costs here and we don't have any more problems with connectivity problems with the low band of Internet and the cost is is going down because we are using the cognitive services just for the first time when I already have the landmarks of face when I already have the information that I need I just make it a comparison over here on the edge and now I have the the time that I need and I have the cost that I need and now I can use this product in small shops and not only in the big shops or big mouth that's the the main idea when we talk about using this guy I guess we are on time yes No so here here I have we are on time that's the the idea of this of this presentation I ask you to evaluate this session please and go to the app and do your evaluation I'm here in the aside here to talk about if you if you have any question I believe we don't have more time here but here and we can talk about if you have any questions if you have any case that want to discuss okay thank you guys

Be First to Comment

Leave a Reply

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