Road to re:Invent - What Is Serverless
57 minute read | Last updated 7-Oct-2019 | 
AWS, AWS re:Invent, Livestream
Serverless is hot right now, but are we all talking about the same thing? In this live stream, we dive into what serverless is, why it’s a big win for your business, and how to get started.
Here are the slides that I used during the live stream.
Bad Robot Transcript
Morning, everybody. Good morning everybody. How you doing today? Good after forgive me if my voice sounds rough battling a bit of a head cold. Hopefully my voice is going to hold with me the whole time also just need to do a quick little color check. I noticed that when I did my recordings that my color was off this a handy little color card, and I’m so I know the value is on this card, which means I can line them up with the recording afterwards.
Thank you for joining me. Today. We are live here on LinkedIn Adam is part of this bigger series at leading up to the AWS reinvent, which is quickly coming up to us when we just kill the fan quickly coming up onto our schedules a started December December 2nd at 55 days left this kind of crazy.
Awesome. So today we’re going to be tackling a very cool topic very interesting topic to be talkin about service. What is it? So you noticed on the schedule I have is only one stream this week and that’s because I will be at Circle has cough at for the next couple days.
That’s why the topic today at what is serverless ABS what is in New York City of this time around? This is the probably the fifth time. I think it’s been held somewhere in there fantastic conference a lot of people talking about service and I thought it’d be really great to kind of lay the groundwork for the people know what’s going on.
What serverless actually this is. I think there’s a lot of confusion right now, but it is and then we’re going to we’re going to touch with some Ace of you a specific references have such a high level overview then some of the previous streams where we’ve got had done a deeper dive into each service.
But I think that’s super important thing. We need this kind of basis after what’s going on. So as always after watching this live hit me up in the comments here on LinkedIn. I will be monitoring them regularly. If the Periscope stream actually managed to start up and get going you can comment there.
I will try to attract that and that’s a little bit trickier for me to do as a one-person show here, but I will be looking for that. I don’t think the Periscope stream is working this morning and this is always the challenge with live streaming. You never know. What’s going to what’s going to work.
What’s not too bad, but we are live here on LinkedIn. So please keep the questions coming. That’s what I find how this makes most makes the most interesting this or the most engaging because we’ve got this interactive, right? So again, just keep the questions coming all good with that.
What I’m going to do is I’m actually going to share out my ass lights to start with. Yeah. We’re going to start with some slides but trust me, it’s not bad. It’s not a bad thing. So that’s the hashtag for serverlessconf. This week if you want to follow along will cover that in a second.
So what is serverless? You probably heard about it. If you have if you’ve got an idea put it in the comments. Let me know a lot of people, you know, it’s a great buzz word it is I’ve been bandied about more and more in the last 6 months. Definitely.
It’s really starting to pick up and I think I wrote my first essay on serverless in 2016 the start of 2016 talk about how to secure serverless. I’m specifically and then it’s just been in a building from there. But that’s true of any technology specialist something that kind of breaks down Jimmy existing paradigms existing ways of thinking Has service.
So let’s get the nitty-gritty at out-of-the-way. Obviously, there are servers as part of serverless the idea here and will make things a little bit bigger. So you guys get a little better view. The idea is that you don’t have to worry about them. They’re no longer your concern and Knotts sort of the goal is that this is no longer your concern.
You don’t have to worry about the servers that are part of the service deployment. That’s the vvv the name sucks. But it conveys an idea, right? So for a while there in the community, we were joking that you can just call was Jeff maybe the same thing but names I think are important in the idea serverless as a name conveys what we’re looking for here as it’s one key piece of infrastructure is gone and not I think it’s the biggest takeaway, but realistically there’s three uses of the word serverless as it stands right now.
I forgot to put up my little buzz clock. So let’s put the countdown clock in the corner. So there’s realistically at three uses. The word serverless today near going to see these used interchangeably you’re going to see these used sort of Miss used a lot of the time depending on where that person who’s talking with service is coming from so first one is a noun going to see serverless being used at talking about a functions as a service offering.
So earlier in the series we cover date of his land a little bit. We did a little bit of info in the Lambda Lambda is considered a serverless Computing service. Microsoft has a matching one Azure functions. Google Cloud functions is also a similar thing and there’s some open source ones that are out there.
The idea is that you have code and you give it to the service provider and executes of based on a specific trigger. Normally that’s an event whether that’s a request from like a website request or another. I like a message dropping in a queue something like that. So it’s the first use of the word serverless and that’s I find its most common usage is people are talking about the fact that it’s fake.
Serverless, I’m using serverless to you. No, clean up my database. Yeah, they have one AWS Lambda function and that is a serverless function. To be told you a little under the weather. So you have to bear with me as I go to water more often on a wood to the second use of this is the architectural terms of service as an architecture know this is the sort of I feel most true really bad way to describe it by feel is the most accurate way of describing serverless because mostly you’ll hear people in the community talking with service in this way.
This is what service comp is all about. It’s an architectural pattern that pulls together a bunch of sacks services and normally uses something like AWS Lambda to glue them together to build a unique solution. The reason why this is my preferred definition of serverless why the community kind of focus is on this a lot is because this is where you are really start to see the business value take something like Amazon S3 what you’ve already seen in the series AWS Lambda and then a service like Amazon rekognition, which analyzes different images to tell you like a you know, if it was looking at Screen it would say there is a poster in the background or there’s a whiteboard in the background and there’s a person there’s a microphone and so that kind of thinking that there’s no servers that you’re touching right and from a business value perspective that’s almost pure business value that you’re delivering into your users.
And there’s very little operational overhead is a very little burden on you of service as an architectural pattern that the last one is the most frustrating adjectives. So using serverless to describe the to describe what a service is so you would say that Amazon S3 is a serverless service because you don’t ever have to deal with touching the servers of the underlying infrastructure a bad thing.
The problem is people are than using that adjectives to sort of surveillance watch things. Like we saw was Cloud all of a sudden many Datacenter service in Cloud came out with now private cloud and we’re seeing a lot of people talk about stuff and say it’s When you know, it’s really a Maybe from one perspective, but it’s not necessarily right.
So but still a valid use Lambda is a serverless service, right? You never have to touch servers in Lander University handling infrastructure. So there’s validity to using it as an adjective. So three ways their lot of the time where you’re going to find the most challenging. This is a big Challenge and security and actually what I’m talking about it service, tomorrow are on the main stage and is when you mix up calling Lambda serverless and the architectural pattern serverless, if you say I’m doing surveillance Security in your only ever looking at Lambda vs.
The entire architecture, that’s a problem because you’re leaving a big chunk on a dressed and there’s a big chunk. That’s not getting properly Managed IT properly evaluated. Okay, so just keep that in mind keep that up to speed 3 different different definitions for serverless. Let’s go and look at that functions piece though because the functions piece is absolutely Because that’s really what makes this stuff come alive.
You can have these various serverless services like Amazon S3 like image. I like database recognition. I’m like some the machine learning services, but if you can’t stitch them together and build something cohesive out of it. What’s the point? That’s where Lambda comes into play. So let’s look at serverless compute for the next few minutes here.
Okay, if you have questions, thank you Brian. I think you Johanna appreciate the feedback. I really like this topic because I think it’s one of those really exciting things where we can finally start to get away from building it4it sake we can start to develop deliver pure business value.
So this is some slides they did from a Trend Micro a webinar a couple months ago. I repurpose them here because I think they’re really important when we look at the shared-responsibility model and we look at a virtual machine. So if you ask for a server, so you have to like a Windows boxer Lennox box from AWS and NEC to get an ec2 instance.
You are responsible for everything you see on screen right now. So the operating system Applications and the data guy. So if patch comes out, which there’s some patches come out tomorrow, it’s October Patch Tuesday and for Windows, you’re going to have to deploy those patches. You have to test them.
You’re going to have to make sure that that is up to us stuff and meet your quality specifications are also going to have to harden an operating system have to configure the users and everything you put on top of that. So the example that I always use it every the chuckle out of is that in this model if you want to set your administrator password the password you absolutely can a horrible idea Pro tip never do that obviously and it should be a beginner tips and still broke.
If you’d be surprised what we see out there, but if you want to set horrible admin passwords, you absolutely can if you want to make everyone administrator in a box you absolutely can’t that’s going to end in disaster but it’s 100% your responsibility. So you have that control. That’s how virtual machines at War instances in AWS Cloud.
That’s how they work if we switch to Containers now, right so containers give you an Additional layer of abstraction. So you noticed the bottom part here are your provider’s responsibility. The container run time is a new layer of abstraction that we’ve added in on top of the operating system.
But we’ve taken the operating system away from the user. These are no longer needs to configure Windows or Linux underneath what you now have is your application in your data that sitting in a container that’s your responsibility. So you’ve delegated one piece of the responsibility out to your provider.
That’s a win right? That’s why containers are are picking up that the ironic thing is that with kubernetes people are trying to take this back. If you’re not using a managed kubernetes service, you’re pushing that responsibility for the runtime of the OS to another team in organizational. You’re not moving your business ahead.
Don’t get me started on kubernetes as a rant. It’s going to be bad, but we’re going to leave that for now. We’ll tackle that a whole nother issue when we move into serverless Computing to function so into AWS Lambda specifically we’ve added another layer of abs. Fraction, you’ll see now we have the function execution environment right below our box, but all of these pieces are now our service provider’s responsibility.
So now the OS the container runtime the app and then the functional execution environment are all AWS is responsibility. They take care of that. They’ve issued some really cool paper explaining how they take care of that. But now we just have our data. That’s our only responsibility is our date in this case our data in AWS.
Lambda is Cote. It’s the code that were writing. That’s our data now, but you see how this model progress is nice and easily you start with your virtual machine. You’re responsible for the OS the app the data you move two containers. There’s an Alero of abstraction but you push a layer of responsibility away when you finally move into functions, there’s a new layer of abstraction, but you push two layers of responsibility now onto your service provider and it makes you a lot easier to figure out what’s going on now Mirko just put a great question.
In there. I talked him. It was all dress right now. What’s the common misconceptions related to service today, and implementation mistakes, right? We’re going to cover that right as we work through this right now Miracle fantastic question. So I think a lot of people don’t understand this shifting and why there’s a business of value there but a service app ends up looking something like this.
OK little sample. What we’ve got is some sort of database service. So if they were using like Dynamo bottom left corner for storage, you can say you’re using S3 for cash YouTube using cloudfront for machine learning any one of the the natural language processing services or like image recognition recognition to make it a little bit easier.
Then you got functions as a service. You got a TBI slander that’s blooming all this together. So any time something happens in one of these Services you’re reacting in Land O. Okay, and that means your programming flow is different. It’s now almost by default an event-driven programming so user makes a request that a bunch of stuff.
There’s no service just sitting there idle waiting right or at least they’re not your responsibility. And that’s the real difference here is that all the services are now build on a transactional costs olander is build in u.s. East last time I checked a couple weeks ago was $0.20 for every 1 million executions.
Okay. So imagine being able to run this look at this diagram and we say this is a service where user upload the photo and they get back an analysis of that photo telling them what’s in that photo and you know, if it compares to any other photos that other users to upload it you can do that entirely without running any infrastructure Dynamo.
You just simply create a table. You don’t run anything S3. We’ve already seen you to shove stuff in a bucket the cashing take this gets taken care of by cloudfront. The machine learning is a a literally send an image file get results back and you do that all from Lambert’s you got very little over had very little operational overhead here, which is great and this is why as a slide from Santa Clara Summit in the spring or had on stage showing at all.
These small sampling of the hundreds of thousands of customers that are leveraging Lambda and then gradually moving into Service as architectures. I’m within their environments and I think Marico to answer your question or start addressing your question. One of the biggest misconceptions is that you need to build everything in service or you know, it’s an all-or-nothing proposition and I find where most people start to adopt the service mindset is starting with serverless Computing AWS Lambda and moving to use that to clean up little things with a native Yasir to glue specific events in need of your us.
So we did a stream we can check on the link at I put in on the second comment there in Amazon are we did the answer to Lambda where we connected every time someone dropped an image into a bucket. We did something right and that’s a really easy way to kind of dip your toe in the water and I think a lot of people I think they’re in a date they have that misconception that they’ve got to go all in on server.
It’s a smart idea but you don’t have to start there. You need to Buchan baby step in and then I think also a lot of people have a misconception that a lot of problems can be solved by serverless because they’re worried about lock-in and I think that’s a huge misconception.
If you break down your code and decompose all your applications into a serverless mindset where you have workflows basically user has asked for this or the system needs to do this. You may be locked into your current implementation. But because you’ve mapped everything out and the code running in Lambda is in your running a JavaScript or python or go where some other language you can take that off and go to another cloud provider but a lot of businesses.
I talked a lot organizations are afraid of lock in and there’s no evidence to back that up. And I think that’s a huge misconception and that’s a big problem. So I know that’s part of your question miracle and the second part. I think I’ll dress after this slide. So later on.
There are four primary benefits to serve as architectures Vols no server management, right? So if you go to a service like RTS, it asked you to provision a database server not a serverless servers, right you go to the Dynamo and it just says yes. Give me your data. No server management flexible scaling so you don’t have to worry about a provisioning capacity.
Lambda just goes right there’s account limits and you may hit those and you have to worry about those are some point but landed just works right once is hooked up to trigger things just start going pay for Value again. No idle time for Lando. Your Lambda function is not executing.
You’re not paying for it similarly for a lot of these serverless Services you are paying a reduced rate when idle so ES300 for storage you’re paying a certain amount and then for Access you’re paying another amount and but that is all tied directly to your value and then automated high-availability.
So this thing is vsync scale-out massively with you not having to do anything else and I think that’s a huge benefit or for huge benefits of the service, Arkansas. And one of the common implementation mistakes, I think is that people try to prescale these they try to zoom it up and go.
Okay. Here’s how it’s going to scale. Don’t worry about it. I haven’t the beauty of these designs is that they essentially scale themselves out and that’s a wonderful thing. So give me an example from a cloud Guru they were at a participated in Melbourne Australia about a month ago.
They had to serve Those Days Inn Melbourne Melbourne for the Ozzie’s. I was promised a Canadian why so we know all the letters so disclaimer. I’m an instructor on a cloud Guru so I know him pretty well I’m and obviously the oldest there’s a financial interest there because I’m a video is Doctor for them, but this is public information.
So they had tweeted out how it’s part of their slides a to give the talk today to dinner after how much uses the ark is there an entirely service infrastructure and they deliver video courses. So pretty clear value when a student is watching a video that’s value to the business and their back end of this.
Costs for Lando wear under $50 a month to maintain a platform that serves a million students and API Gateway was about $700 a month cuz there’s a lot of logic and security pieces there and then they’re cloudfront. We going to work out to be about seven thousand a month and it sounds like a lot of money but the interesting thing is when you break down that that spending was that 85% of their it overhead was going to the thing that directly matter to students which was delivering videos to them through cloudfront that is a huge business win.
Because if you look at your business right now, you’re paying a huge amount of stuff in the background that doesn’t actually tied directly to the value that you’re delivering at. We’re trying to deliver at 4 on behalf of the business, right you’re running all this back in infrastructure to get to the point where you can deliver value so the Advantage Services Serverless architecture Tabitha, they make them far more direct line to Value.
So just before I flip over to my browsing to show you some of the stuff in the weeds. Remember the if you want to hear more about circle of stuff this week tomorrow at 2, Tuesday and Wednesday service comp is happening in NYC at service composite hashtag on Twitter.
It is always super social lots of stuff going on there and you probably going to learn a lot of cool stuff. You can follow along on that hashtag right now. Let me sleep over to my browser and I will show you a couple quick things. So let me make this look prettier for y’all.
So it’s always lose my sight. I put the link in the top and we’re capturing everything that we’re doing here for our event is the stream. We’re currently in. But I also just published a talk that I gave it all the summit’s this year and called Advanced security automation.
It would help with the link worked Advanced security automation Made Simple. I’ll double-check that Lincoln a second, but it’s actually what it is is it shows you some surrealist patterns to automate security stuff check that out when you get a sec after I fix it and I wanted to show you what page they’re going to AWS.
Amazon.com / serverless. You’re going to see a quick definition of service. So you’ll see they’re using the architecture definition of service. So we went back to her in a service Computer Service architecture service the adjectives they are talking about architecture. So we need to be us talk service. They talk architecture most people do and then they go through some of the benefits like why are you going to use it and make sure you the AWS serverless platform.
So you see here. They’re calling out specific services that would qualify to use the adjectives serverless to Lambda. Obviously Lambda at Edge. Yes fargate. Container service, but it’s considered service because you have to worry about anything underneath the elastic filesystem dynamodb the API Gateway step functions help you stitch together multiple circles compute things but something like SNS, which has been around forever same with sqs their service cuz you don’t set up a server and soak it process Q-Zar process messages.
You just say I’m going to send you messages you give them back to me or deliver them when I need them, right? That’s the concept behind serverless. So lots of crazy server Services here. You should take your time and kind of meander through this page there. We better for benefits of serverless.
I’m so no server management flexible scaling. I automated for high availability. I know it’s kind of boring sometimes watching these pages. So if you actually come back to I will share it at the right time. So if you come back to the Santa Clara keynote, I’m going to drop this in the chat here on LinkedIn and this is where Vernor starts to explain the differences and Savannah’s the CTO of Amazon.
Cam he does a really good job of setting it up and explaining sort of these advantages of what works for service why service is important and why you should be looking into it. I’m so full foot Flare from the from the service page Olympus has a thing called a serverless application repository.
So most people are going to start with serverless by just using Lambda like we did earlier in this year. You’re going to see him to give you that post to I’m so when you’re seeing Lambda and you’re going to start without as a kind of a dip your toe in the water.
Into the service idea. I can’t talk and type for some reason I do a bunch of stuff and talk but not talking time. And so will there’s also this idea call the service application repository or Tsar ago and they offer the community bility to publish applications. I’m so the others can reuse them others can start with that.
Right? So if you go to star and it’s going to give you an idea of how Works basically the app repository is there you find what you want. You can figure it you deploy it out. That sounds like a lot of work, but it’s ridiculously easy. And actually we’ve got four minutes.
So we’re going to deploy serverless application from the repository and you’ll see how simple it is. So if I click on I don’t like that big menu click on the browser applications button actually going to take me into this view of the repository and you can start to search through this right? So if you’re looking for something in particular like an Alexa skill, you can have the basic start of an Alexis.
Withdrawal serverless, by the way, everything is behind Amazon Alexa. It’s all running in circles infrastructures list of scales and trillions and trillions of executions a month going on the back end and so you can go through here and find some stuff from Partners like gated dog has a log forwarder hear some of the end of his community Heroes and service Heroes of publish stuff in here and just to show you what’s possible.
So I always keep talking about, you know image resizing. Here’s a start of an image resizing service. If you want to spend what up, what we’re going to do though is a good old hello world, why not? Hello world works and has work since the beginning of time. Basically very simple.
This is a python version of hello world that you can look at the license. You can look at the required permissions. I love that. You can also go to the GitHub page for it. So this case you’re pulled into the AWS slabs project model for it is a very very simple hello world, and I think they actually switched away from python into JavaScript.
There may be a python there about 1 somewhere but Is that our pantry in the service application repository if we simply click the orange deploy button? This is going to take us to RI Lambda Management console and we can copy this as a Sam resource, which is what we want to do that’s going to take another repository and put it into our account, but we can look at the template.
So in this case if we look down to the template, is it a cloudformation template that builds the stuff in our account? It’s going to create a server list function. Awesome. It’s going to be running node. That’s fine. And it sets a few things right and starter AWS Lambda function and you permissions.
It requires needs a bounce policy for email. That’s fine. We’re going to click on copy as Sam resources going to take it from the repo put it into our account. notes That’s fine. So, what did I there we go gas. They always moved it. So you can copy the same resources the URL to get it out of the repository.
So if I want to show you guys You love live streaming. Sometimes you say the right thing. Sometimes you’re wrong. So I have a copy as Sam resource is this link? And no it’s even worse gives you the application ID that you can use later. So ignore that. Rewind a minute what we’re going to do scroll down and got to read me file.
That’s fine to read me strong. So it’s not in Python are app name. We don’t need to give an identity primary. We’re going to click the orange the boy button. We got a minute and a half left 15 hours an employee button. It’s deploying it will not take that long.
Identity must have value road to reinvent Tsar test. So given identity name Brammer and we’re going to cook that and that’s just going to fill in the permissions that it needs the back-end. I with that name. So now that should work against screaming live. Sometimes it works perfectly. Sometimes it doesn’t but here now our application is being deployed and it’s taking us to to the main page listening to see our resources to helloworld role is creating a process and this going to take a minute or so a man at the end of it where I have a really simple little hello world application.
It’s going to just respond. Hello world. We click on the test app, very basic example of a service application and but there’s a ton of stuff in the repo and they could really complicated really quickly but in a good way, right? So now we get the screen notifications are app has been deployed if we click the orange test app button.
What we’re going to see is we getting started that’s great resources here. Let’s click down to see Hello World roll being used K2. We have deployments are Sam. So this is created. All I’m looking for right now is the URL at some point is going to tell me there was a URL monitor still here.
It is running. Okay, so let’s go to functions and you’ll see service repo great and now we tested here. I’ll give it a sample value. So you saw this in our Lambda tutorial cook create. Test everything needs a name. And now when we run it 6s details, that’s it.
Just a simple hello world right now. You would think that you don’t even respond hello world, but it’s a very very basic application to start with and if we look at the code, it’s probably going to actually have Hello World in there somewhere. Yeah, it doesn’t even actually say hello world.
That’s a shotty hello world. So I apologize for that and but very simple very straightforward. What we do is we took it a resource out of the Sam or out of the Tsar repository and the deployed it ourselves. I am going to actually make a pull request to fix that later on if I have time and same with the with the read me for it, but the idea here is that there’s a bunch of these serverless applications that you can get started with because of there is a huge amount of benefit here right? I’m just going to go back to my slides before we wrap this up.
Remember the goal here is that you’re pushing more and more away, right? This is the service compute view. We’re taking your responsibility and making an arrow into just the code write your data is your code for service computer and you can start stitching things together and create some pretty complicated application some pretty valuable applications with really little to no overhead.
If you don’t have to run the service if you don’t have to patch all that stuff, you don’t upgrade it when you get to see these start of these for benefits, right? No server management. No, really underlying infrastructure management you that flexible scaling you pay for value and get that automated high-availability.
Remember? I’m at serverlessconf the rest of this week. Is it a hashtag on Twitter? I it is a really active conference on social in here 10 stuff from some really great and smart smart people talking about the latest and greatest serverless. I will put this image up again on the site.
You got it here on the screen. But remember the three uses of the term service to talk about a Tervis Lambda and it’s can Patriots talk with the ark Structure which is what most the time you want to be talking about and then as an adjectives to describe a specific service be aware when you’re in a conversation around service to make sure that you’re using the same type.
I’ve definition cuz if you’re talking about a specific service or compute and somebody’s talking with the architecture, you’re going to get wires crossed and it’s going to be a problem and thank you for bearing with me. Despite the gravelly voice despite the the sort of the misfire on the Hello World stuff.
And as always let me know. I monitor the LinkedIn chat constantly, I get the notifications. If you comment after the fact that would be up on my site and up on YouTube later and follow serverlessconf during the week at tomorrow at Tuesday and Wednesday is going to be some really fantastic at talks given im a nose go up on the on the conference site afterwards as well, but follow along and join the conversation service is super exciting.
It is definitely something we are going to hear about more and more as we are leaving up to AWS reinvent and but I appreciate you guys taking a little bit of time out of your day and following along and I will see you on next week on the stream am going to put that up.
It’s a Canadian Thanksgiving this weekend. And so I’m going to have my turkey day was great. Leaves are all changing at and then I’ll be back on the streams on next week. So I check out for that at thanks again. I hope you have a great week. Will talk to you online real soon.
Take care. Morning, everybody. Good morning everybody. How you doing today? Good after forgive me if my voice sounds rough battling a bit of a head cold. Hopefully my voice is going to hold with me the whole time also just need to do a quick little color check. I noticed that when I did my recordings that my color was off this a handy little color card, and I’m so I know the value is on this card, which means I can line them up with the recording afterwards.
Thank you for joining me. Today. We are live here on LinkedIn Adam is part of this bigger series at leading up to the AWS reinvent, which is quickly coming up to us when we just kill the fan quickly coming up onto our schedules a started December December 2nd at 55 days left this kind of crazy.
Awesome. So today we’re going to be tackling a very cool topic very interesting topic to be talkin about service. What is it? So you noticed on the schedule I have is only one stream this week and that’s because I will be at Circle has cough at for the next couple days.
That’s why the topic today at what is serverless ABS what is in New York City of this time around? This is the probably the fifth time. I think it’s been held somewhere in there fantastic conference a lot of people talking about service and I thought it’d be really great to kind of lay the groundwork for the people know what’s going on.
What serverless actually this is. I think there’s a lot of confusion right now, but it is and then we’re going to we’re going to touch with some Ace of you a specific references have such a high level overview then some of the previous streams where we’ve got had done a deeper dive into each service.
But I think that’s super important thing. We need this kind of basis after what’s going on. So as always after watching this live hit me up in the comments here on LinkedIn. I will be monitoring them regularly. If the Periscope stream actually managed to start up and get going you can comment there.
I will try to attract that and that’s a little bit trickier for me to do as a one-person show here, but I will be looking for that. I don’t think the Periscope stream is working this morning and this is always the challenge with live streaming. You never know. What’s going to what’s going to work.
What’s not too bad, but we are live here on LinkedIn. So please keep the questions coming. That’s what I find how this makes most makes the most interesting this or the most engaging because we’ve got this interactive, right? So again, just keep the questions coming all good with that.
What I’m going to do is I’m actually going to share out my ass lights to start with. Yeah. We’re going to start with some slides but trust me, it’s not bad. It’s not a bad thing. So that’s the hashtag for serverlessconf. This week if you want to follow along will cover that in a second.
So what is serverless? You probably heard about it. If you have if you’ve got an idea put it in the comments. Let me know a lot of people, you know, it’s a great buzz word it is I’ve been bandied about more and more in the last 6 months. Definitely.
It’s really starting to pick up and I think I wrote my first essay on serverless in 2016 the start of 2016 talk about how to secure serverless. I’m specifically and then it’s just been in a building from there. But that’s true of any technology specialist something that kind of breaks down Jimmy existing paradigms existing ways of thinking Has service.
So let’s get the nitty-gritty at out-of-the-way. Obviously, there are servers as part of serverless the idea here and will make things a little bit bigger. So you guys get a little better view. The idea is that you don’t have to worry about them. They’re no longer your concern and Knotts sort of the goal is that this is no longer your concern.
You don’t have to worry about the servers that are part of the service deployment. That’s the vvv the name sucks. But it conveys an idea, right? So for a while there in the community, we were joking that you can just call was Jeff maybe the same thing but names I think are important in the idea serverless as a name conveys what we’re looking for here as it’s one key piece of infrastructure is gone and not I think it’s the biggest takeaway, but realistically there’s three uses of the word serverless as it stands right now.
I forgot to put up my little buzz clock. So let’s put the countdown clock in the corner. So there’s realistically at three uses. The word serverless today near going to see these used interchangeably you’re going to see these used sort of Miss used a lot of the time depending on where that person who’s talking with service is coming from so first one is a noun going to see serverless being used at talking about a functions as a service offering.
So earlier in the series we cover date of his land a little bit. We did a little bit of info in the Lambda Lambda is considered a serverless Computing service. Microsoft has a matching one Azure functions. Google Cloud functions is also a similar thing and there’s some open source ones that are out there.
The idea is that you have code and you give it to the service provider and executes of based on a specific trigger. Normally that’s an event whether that’s a request from like a website request or another. I like a message dropping in a queue something like that. So it’s the first use of the word serverless and that’s I find its most common usage is people are talking about the fact that it’s fake.
Serverless, I’m using serverless to you. No, clean up my database. Yeah, they have one AWS Lambda function and that is a serverless function. To be told you a little under the weather. So you have to bear with me as I go to water more often on a wood to the second use of this is the architectural terms of service as an architecture know this is the sort of I feel most true really bad way to describe it by feel is the most accurate way of describing serverless because mostly you’ll hear people in the community talking with service in this way.
This is what service comp is all about. It’s an architectural pattern that pulls together a bunch of sacks services and normally uses something like AWS Lambda to glue them together to build a unique solution. The reason why this is my preferred definition of serverless why the community kind of focus is on this a lot is because this is where you are really start to see the business value take something like Amazon S3 what you’ve already seen in the series AWS Lambda and then a service like Amazon rekognition, which analyzes different images to tell you like a you know, if it was looking at Screen it would say there is a poster in the background or there’s a whiteboard in the background and there’s a person there’s a microphone and so that kind of thinking that there’s no servers that you’re touching right and from a business value perspective that’s almost pure business value that you’re delivering into your users.
And there’s very little operational overhead is a very little burden on you of service as an architectural pattern that the last one is the most frustrating adjectives. So using serverless to describe the to describe what a service is so you would say that Amazon S3 is a serverless service because you don’t ever have to deal with touching the servers of the underlying infrastructure a bad thing.
The problem is people are than using that adjectives to sort of surveillance watch things. Like we saw was Cloud all of a sudden many Datacenter service in Cloud came out with now private cloud and we’re seeing a lot of people talk about stuff and say it’s When you know, it’s really a Maybe from one perspective, but it’s not necessarily right.
So but still a valid use Lambda is a serverless service, right? You never have to touch servers in Lander University handling infrastructure. So there’s validity to using it as an adjective. So three ways their lot of the time where you’re going to find the most challenging. This is a big Challenge and security and actually what I’m talking about it service, tomorrow are on the main stage and is when you mix up calling Lambda serverless and the architectural pattern serverless, if you say I’m doing surveillance Security in your only ever looking at Lambda vs.
The entire architecture, that’s a problem because you’re leaving a big chunk on a dressed and there’s a big chunk. That’s not getting properly Managed IT properly evaluated. Okay, so just keep that in mind keep that up to speed 3 different different definitions for serverless. Let’s go and look at that functions piece though because the functions piece is absolutely Because that’s really what makes this stuff come alive.
You can have these various serverless services like Amazon S3 like image. I like database recognition. I’m like some the machine learning services, but if you can’t stitch them together and build something cohesive out of it. What’s the point? That’s where Lambda comes into play. So let’s look at serverless compute for the next few minutes here.
Okay, if you have questions, thank you Brian. I think you Johanna appreciate the feedback. I really like this topic because I think it’s one of those really exciting things where we can finally start to get away from building it4it sake we can start to develop deliver pure business value.
So this is some slides they did from a Trend Micro a webinar a couple months ago. I repurpose them here because I think they’re really important when we look at the shared-responsibility model and we look at a virtual machine. So if you ask for a server, so you have to like a Windows boxer Lennox box from AWS and NEC to get an ec2 instance.
You are responsible for everything you see on screen right now. So the operating system Applications and the data guy. So if patch comes out, which there’s some patches come out tomorrow, it’s October Patch Tuesday and for Windows, you’re going to have to deploy those patches. You have to test them.
You’re going to have to make sure that that is up to us stuff and meet your quality specifications are also going to have to harden an operating system have to configure the users and everything you put on top of that. So the example that I always use it every the chuckle out of is that in this model if you want to set your administrator password the password you absolutely can a horrible idea Pro tip never do that obviously and it should be a beginner tips and still broke.
If you’d be surprised what we see out there, but if you want to set horrible admin passwords, you absolutely can if you want to make everyone administrator in a box you absolutely can’t that’s going to end in disaster but it’s 100% your responsibility. So you have that control. That’s how virtual machines at War instances in AWS Cloud.
That’s how they work if we switch to Containers now, right so containers give you an Additional layer of abstraction. So you noticed the bottom part here are your provider’s responsibility. The container run time is a new layer of abstraction that we’ve added in on top of the operating system.
But we’ve taken the operating system away from the user. These are no longer needs to configure Windows or Linux underneath what you now have is your application in your data that sitting in a container that’s your responsibility. So you’ve delegated one piece of the responsibility out to your provider.
That’s a win right? That’s why containers are are picking up that the ironic thing is that with kubernetes people are trying to take this back. If you’re not using a managed kubernetes service, you’re pushing that responsibility for the runtime of the OS to another team in organizational. You’re not moving your business ahead.
Don’t get me started on kubernetes as a rant. It’s going to be bad, but we’re going to leave that for now. We’ll tackle that a whole nother issue when we move into serverless Computing to function so into AWS Lambda specifically we’ve added another layer of abs. Fraction, you’ll see now we have the function execution environment right below our box, but all of these pieces are now our service provider’s responsibility.
So now the OS the container runtime the app and then the functional execution environment are all AWS is responsibility. They take care of that. They’ve issued some really cool paper explaining how they take care of that. But now we just have our data. That’s our only responsibility is our date in this case our data in AWS.
Lambda is Cote. It’s the code that were writing. That’s our data now, but you see how this model progress is nice and easily you start with your virtual machine. You’re responsible for the OS the app the data you move two containers. There’s an Alero of abstraction but you push a layer of responsibility away when you finally move into functions, there’s a new layer of abstraction, but you push two layers of responsibility now onto your service provider and it makes you a lot easier to figure out what’s going on now Mirko just put a great question.
In there. I talked him. It was all dress right now. What’s the common misconceptions related to service today, and implementation mistakes, right? We’re going to cover that right as we work through this right now Miracle fantastic question. So I think a lot of people don’t understand this shifting and why there’s a business of value there but a service app ends up looking something like this.
OK little sample. What we’ve got is some sort of database service. So if they were using like Dynamo bottom left corner for storage, you can say you’re using S3 for cash YouTube using cloudfront for machine learning any one of the the natural language processing services or like image recognition recognition to make it a little bit easier.
Then you got functions as a service. You got a TBI slander that’s blooming all this together. So any time something happens in one of these Services you’re reacting in Land O. Okay, and that means your programming flow is different. It’s now almost by default an event-driven programming so user makes a request that a bunch of stuff.
There’s no service just sitting there idle waiting right or at least they’re not your responsibility. And that’s the real difference here is that all the services are now build on a transactional costs olander is build in u.s. East last time I checked a couple weeks ago was $0.20 for every 1 million executions.
Okay. So imagine being able to run this look at this diagram and we say this is a service where user upload the photo and they get back an analysis of that photo telling them what’s in that photo and you know, if it compares to any other photos that other users to upload it you can do that entirely without running any infrastructure Dynamo.
You just simply create a table. You don’t run anything S3. We’ve already seen you to shove stuff in a bucket the cashing take this gets taken care of by cloudfront. The machine learning is a a literally send an image file get results back and you do that all from Lambert’s you got very little over had very little operational overhead here, which is great and this is why as a slide from Santa Clara Summit in the spring or had on stage showing at all.
These small sampling of the hundreds of thousands of customers that are leveraging Lambda and then gradually moving into Service as architectures. I’m within their environments and I think Marico to answer your question or start addressing your question. One of the biggest misconceptions is that you need to build everything in service or you know, it’s an all-or-nothing proposition and I find where most people start to adopt the service mindset is starting with serverless Computing AWS Lambda and moving to use that to clean up little things with a native Yasir to glue specific events in need of your us.
So we did a stream we can check on the link at I put in on the second comment there in Amazon are we did the answer to Lambda where we connected every time someone dropped an image into a bucket. We did something right and that’s a really easy way to kind of dip your toe in the water and I think a lot of people I think they’re in a date they have that misconception that they’ve got to go all in on server.
It’s a smart idea but you don’t have to start there. You need to Buchan baby step in and then I think also a lot of people have a misconception that a lot of problems can be solved by serverless because they’re worried about lock-in and I think that’s a huge misconception.
If you break down your code and decompose all your applications into a serverless mindset where you have workflows basically user has asked for this or the system needs to do this. You may be locked into your current implementation. But because you’ve mapped everything out and the code running in Lambda is in your running a JavaScript or python or go where some other language you can take that off and go to another cloud provider but a lot of businesses.
I talked a lot organizations are afraid of lock in and there’s no evidence to back that up. And I think that’s a huge misconception and that’s a big problem. So I know that’s part of your question miracle and the second part. I think I’ll dress after this slide. So later on.
There are four primary benefits to serve as architectures Vols no server management, right? So if you go to a service like RTS, it asked you to provision a database server not a serverless servers, right you go to the Dynamo and it just says yes. Give me your data. No server management flexible scaling so you don’t have to worry about a provisioning capacity.
Lambda just goes right there’s account limits and you may hit those and you have to worry about those are some point but landed just works right once is hooked up to trigger things just start going pay for Value again. No idle time for Lando. Your Lambda function is not executing.
You’re not paying for it similarly for a lot of these serverless Services you are paying a reduced rate when idle so ES300 for storage you’re paying a certain amount and then for Access you’re paying another amount and but that is all tied directly to your value and then automated high-availability.
So this thing is vsync scale-out massively with you not having to do anything else and I think that’s a huge benefit or for huge benefits of the service, Arkansas. And one of the common implementation mistakes, I think is that people try to prescale these they try to zoom it up and go.
Okay. Here’s how it’s going to scale. Don’t worry about it. I haven’t the beauty of these designs is that they essentially scale themselves out and that’s a wonderful thing. So give me an example from a cloud Guru they were at a participated in Melbourne Australia about a month ago.
They had to serve Those Days Inn Melbourne Melbourne for the Ozzie’s. I was promised a Canadian why so we know all the letters so disclaimer. I’m an instructor on a cloud Guru so I know him pretty well I’m and obviously the oldest there’s a financial interest there because I’m a video is Doctor for them, but this is public information.
So they had tweeted out how it’s part of their slides a to give the talk today to dinner after how much uses the ark is there an entirely service infrastructure and they deliver video courses. So pretty clear value when a student is watching a video that’s value to the business and their back end of this.
Costs for Lando wear under $50 a month to maintain a platform that serves a million students and API Gateway was about $700 a month cuz there’s a lot of logic and security pieces there and then they’re cloudfront. We going to work out to be about seven thousand a month and it sounds like a lot of money but the interesting thing is when you break down that that spending was that 85% of their it overhead was going to the thing that directly matter to students which was delivering videos to them through cloudfront that is a huge business win.
Because if you look at your business right now, you’re paying a huge amount of stuff in the background that doesn’t actually tied directly to the value that you’re delivering at. We’re trying to deliver at 4 on behalf of the business, right you’re running all this back in infrastructure to get to the point where you can deliver value so the Advantage Services Serverless architecture Tabitha, they make them far more direct line to Value.
So just before I flip over to my browsing to show you some of the stuff in the weeds. Remember the if you want to hear more about circle of stuff this week tomorrow at 2, Tuesday and Wednesday service comp is happening in NYC at service composite hashtag on Twitter.
It is always super social lots of stuff going on there and you probably going to learn a lot of cool stuff. You can follow along on that hashtag right now. Let me sleep over to my browser and I will show you a couple quick things. So let me make this look prettier for y’all.
So it’s always lose my sight. I put the link in the top and we’re capturing everything that we’re doing here for our event is the stream. We’re currently in. But I also just published a talk that I gave it all the summit’s this year and called Advanced security automation.
It would help with the link worked Advanced security automation Made Simple. I’ll double-check that Lincoln a second, but it’s actually what it is is it shows you some surrealist patterns to automate security stuff check that out when you get a sec after I fix it and I wanted to show you what page they’re going to AWS.
Amazon.com / serverless. You’re going to see a quick definition of service. So you’ll see they’re using the architecture definition of service. So we went back to her in a service Computer Service architecture service the adjectives they are talking about architecture. So we need to be us talk service. They talk architecture most people do and then they go through some of the benefits like why are you going to use it and make sure you the AWS serverless platform.
So you see here. They’re calling out specific services that would qualify to use the adjectives serverless to Lambda. Obviously Lambda at Edge. Yes fargate. Container service, but it’s considered service because you have to worry about anything underneath the elastic filesystem dynamodb the API Gateway step functions help you stitch together multiple circles compute things but something like SNS, which has been around forever same with sqs their service cuz you don’t set up a server and soak it process Q-Zar process messages.
You just say I’m going to send you messages you give them back to me or deliver them when I need them, right? That’s the concept behind serverless. So lots of crazy server Services here. You should take your time and kind of meander through this page there. We better for benefits of serverless.
I’m so no server management flexible scaling. I automated for high availability. I know it’s kind of boring sometimes watching these pages. So if you actually come back to I will share it at the right time. So if you come back to the Santa Clara keynote, I’m going to drop this in the chat here on LinkedIn and this is where Vernor starts to explain the differences and Savannah’s the CTO of Amazon.
Cam he does a really good job of setting it up and explaining sort of these advantages of what works for service why service is important and why you should be looking into it. I’m so full foot Flare from the from the service page Olympus has a thing called a serverless application repository.
So most people are going to start with serverless by just using Lambda like we did earlier in this year. You’re going to see him to give you that post to I’m so when you’re seeing Lambda and you’re going to start without as a kind of a dip your toe in the water.
Into the service idea. I can’t talk and type for some reason I do a bunch of stuff and talk but not talking time. And so will there’s also this idea call the service application repository or Tsar ago and they offer the community bility to publish applications. I’m so the others can reuse them others can start with that.
Right? So if you go to star and it’s going to give you an idea of how Works basically the app repository is there you find what you want. You can figure it you deploy it out. That sounds like a lot of work, but it’s ridiculously easy. And actually we’ve got four minutes.
So we’re going to deploy serverless application from the repository and you’ll see how simple it is. So if I click on I don’t like that big menu click on the browser applications button actually going to take me into this view of the repository and you can start to search through this right? So if you’re looking for something in particular like an Alexa skill, you can have the basic start of an Alexis.
Withdrawal serverless, by the way, everything is behind Amazon Alexa. It’s all running in circles infrastructures list of scales and trillions and trillions of executions a month going on the back end and so you can go through here and find some stuff from Partners like gated dog has a log forwarder hear some of the end of his community Heroes and service Heroes of publish stuff in here and just to show you what’s possible.
So I always keep talking about, you know image resizing. Here’s a start of an image resizing service. If you want to spend what up, what we’re going to do though is a good old hello world, why not? Hello world works and has work since the beginning of time. Basically very simple.
This is a python version of hello world that you can look at the license. You can look at the required permissions. I love that. You can also go to the GitHub page for it. So this case you’re pulled into the AWS slabs project model for it is a very very simple hello world, and I think they actually switched away from python into JavaScript.
There may be a python there about 1 somewhere but Is that our pantry in the service application repository if we simply click the orange deploy button? This is going to take us to RI Lambda Management console and we can copy this as a Sam resource, which is what we want to do that’s going to take another repository and put it into our account, but we can look at the template.
So in this case if we look down to the template, is it a cloudformation template that builds the stuff in our account? It’s going to create a server list function. Awesome. It’s going to be running node. That’s fine. And it sets a few things right and starter AWS Lambda function and you permissions.
It requires needs a bounce policy for email. That’s fine. We’re going to click on copy as Sam resources going to take it from the repo put it into our account. notes That’s fine. So, what did I there we go gas. They always moved it. So you can copy the same resources the URL to get it out of the repository.
So if I want to show you guys You love live streaming. Sometimes you say the right thing. Sometimes you’re wrong. So I have a copy as Sam resource is this link? And no it’s even worse gives you the application ID that you can use later. So ignore that. Rewind a minute what we’re going to do scroll down and got to read me file.
That’s fine to read me strong. So it’s not in Python are app name. We don’t need to give an identity primary. We’re going to click the orange the boy button. We got a minute and a half left 15 hours an employee button. It’s deploying it will not take that long.
Identity must have value road to reinvent Tsar test. So given identity name Brammer and we’re going to cook that and that’s just going to fill in the permissions that it needs the back-end. I with that name. So now that should work against screaming live. Sometimes it works perfectly. Sometimes it doesn’t but here now our application is being deployed and it’s taking us to to the main page listening to see our resources to helloworld role is creating a process and this going to take a minute or so a man at the end of it where I have a really simple little hello world application.
It’s going to just respond. Hello world. We click on the test app, very basic example of a service application and but there’s a ton of stuff in the repo and they could really complicated really quickly but in a good way, right? So now we get the screen notifications are app has been deployed if we click the orange test app button.
What we’re going to see is we getting started that’s great resources here. Let’s click down to see Hello World roll being used K2. We have deployments are Sam. So this is created. All I’m looking for right now is the URL at some point is going to tell me there was a URL monitor still here.
It is running. Okay, so let’s go to functions and you’ll see service repo great and now we tested here. I’ll give it a sample value. So you saw this in our Lambda tutorial cook create. Test everything needs a name. And now when we run it 6s details, that’s it.
Just a simple hello world right now. You would think that you don’t even respond hello world, but it’s a very very basic application to start with and if we look at the code, it’s probably going to actually have Hello World in there somewhere. Yeah, it doesn’t even actually say hello world.
That’s a shotty hello world. So I apologize for that and but very simple very straightforward. What we do is we took it a resource out of the Sam or out of the Tsar repository and the deployed it ourselves. I am going to actually make a pull request to fix that later on if I have time and same with the with the read me for it, but the idea here is that there’s a bunch of these serverless applications that you can get started with because of there is a huge amount of benefit here right? I’m just going to go back to my slides before we wrap this up.
Remember the goal here is that you’re pushing more and more away, right? This is the service compute view. We’re taking your responsibility and making an arrow into just the code write your data is your code for service computer and you can start stitching things together and create some pretty complicated application some pretty valuable applications with really little to no overhead.
If you don’t have to run the service if you don’t have to patch all that stuff, you don’t upgrade it when you get to see these start of these for benefits, right? No server management. No, really underlying infrastructure management you that flexible scaling you pay for value and get that automated high-availability.
Remember? I’m at serverlessconf the rest of this week. Is it a hashtag on Twitter? I it is a really active conference on social in here 10 stuff from some really great and smart smart people talking about the latest and greatest serverless. I will put this image up again on the site.
You got it here on the screen. But remember the three uses of the term service to talk about a Tervis Lambda and it’s can Patriots talk with the ark Structure which is what most the time you want to be talking about and then as an adjectives to describe a specific service be aware when you’re in a conversation around service to make sure that you’re using the same type.
I’ve definition cuz if you’re talking about a specific service or compute and somebody’s talking with the architecture, you’re going to get wires crossed and it’s going to be a problem and thank you for bearing with me. Despite the gravelly voice despite the the sort of the misfire on the Hello World stuff.
And as always let me know. I monitor the LinkedIn chat constantly, I get the notifications. If you comment after the fact that would be up on my site and up on YouTube later and follow serverlessconf during the week at tomorrow at Tuesday and Wednesday is going to be some really fantastic at talks given im a nose go up on the on the conference site afterwards as well, but follow along and join the conversation service is super exciting.
It is definitely something we are going to hear about more and more as we are leaving up to AWS reinvent and but I appreciate you guys taking a little bit of time out of your day and following along and I will see you on next week on the stream am going to put that up.
It’s a Canadian Thanksgiving this weekend. And so I’m going to have my turkey day was great. Leaves are all changing at and then I’ll be back on the streams on next week. So I check out for that at thanks again. I hope you have a great week. Will talk to you online real soon.
Take care.