The Serverless Hype Explained!
HTML-код
- Опубликовано: 26 сен 2020
- Why are "serverless" cloud products so popular right now? What makes them different from using traditional server infrastructure? In this video, I'll go over all the pros and cons of going for a serverless architecture. On it, you can run your website or any web application.
🌍 Social
Twitter: / savjee
Facebook: / savjee
Blog: savjee.be
❤️ Become a Simply Explained member: / @simplyexplained
💌 Newsletter: newsletter.savjee.be (no more than once a month)
📚 Sources used to make this video:
savjee.be/videos/simply-expla... 
Наука
![DDG & Blueface - Top ft. Swae Lee [Music Video]](http://i.ytimg.com/vi/p2wIDXar9ns/mqdefault.jpg)
Peer-to-Peer is the true serverless!
What is peer to peer?
@@abhirb12 have you used limeware or ares or any torrent? Thats peer to peer
@@juanmanuel7305 Also every cryptocurrency to ever exist.
@@abhirb12 peer to peer means that every member in the network can fetch and serve data equally. It's a form of decentralized networking and thus no single point of failure
@@wolvenmoonstone8138 : something like git?
a DDOS attack on a serverless hosted web site might cost you more than the downtime of a crashed server
Exactly. I was baffled to see he didn't mention that. Also, even if you're the ultimate security expert, one of your functions might not scale well, you could end up stuck at an infinite loop, etc. I mean, sure, you can (sometimes) put caps, but that beats the purpose of serverless.
It's the perfect solution only if every single member of your team is infallible.
configure cloudflare
AWS is withstanding huge DDoS attacks all the time, and they are better at it than you.
You can limit the number of concurrent invocations if you're worried about this. And I'm sure providers like AWS and Google are better at preventing ddos attacks
Sooo true
2:19 - I like how he uses a cassette tape to show storage... I'm also growing old fast.
Maybe he grew up using a trash-80. :)
Cassette's are still widely used today for data backups weirdly enough. They are dirt cheap and store a LOT of data
This doesn't sounds like a server less system. It's sort of dynamic load balancing. A peer to peer connection would be server less.
So it's basically still in a server just not next to you. "Cloud"
it just means you dont managed and setup the server. i also dont understand why it's "cloud", it's mumbo jumbo words
@@archmad cloud is just another server where you can run your the same things like hosting but on a aws server
You’re right, it’s the biggest pile of waffle I’ve come across and been around since the 90’s. It’s basically a business moving from in-house to external. This has been going on since the birth of the net.
I love the way how you compare the relative things ...then it just makes so much sense
Thanks for making these awesome videos
Thanks a lot, dude. Finally, someone explain serverless as clear as possible, so even I understood the concept.
Cloud are still servers and often managed by another group of people. All the issue with servers are still there just that someone said they manage them by taking your money. And I found their reliability and service quality are not higher than servers manage by my own people yet cost more.
To my understanding, what you're describing in this video is mainly the concept of virtual sever.
Server-less is more like AWS lambda: no server OS management, no network / firewall management, no container management, no load balancing management.
I could be wrong though.
You are not wrong.
No, virtual servers can, and most probably will, exist in-house. Virtualisation makes no difference where it’s set up, it’s simply a way of setting up. Serverless is a term which doesn’t really make any sense, it’s still run by servers, just no longer in the building.
Serverless products & services are a great way for centralization and this is a great way for ... censorship. Maybe something like IPFS (OrbitDB)?
The more dependant we become to "cloud" providers, the more control over the information we gave them, the less freedom we either as users or developers have.
You could say that of the cloud in general. It does indeed centralize a lot of internet services. It's crazy to think that when AWS goes down, so does a large portion of websites on the internet. www.theverge.com/2017/3/2/14792442/amazon-s3-outage-cause-typo-internet-server
@@juanmanuel7305 time to develop your own OS using your own language and run it with your own software on your own hardware. Better create your own ISP company as well.
@@brad7957 what? im not even a purist, just a thought, and i do host my own things on my own bare metal servers, even though i work as a cloudops, i dont really like cloud providers
I love how you explain complex topics. Thank you very much!
Thanks for the video, hadn't thought about the vendor lock-in issue , tricky situation indeed!
The wireless/serverless example is really good
Nice explanation, I think serverless applications will have its place on the ecosystems but on a minor level. For example; less frequent, less performance dependent and single use operations inside a larger system may be a better option for serverless. We will probably see more frequent use cases on upcoming years, but the server and serverless distinction is currently way too blurry in my opinion.
Really well explained! Thank you for this video
Thanks so much for making this amazing video! this is well explained
Serverless really had me confused till I saw this video. Thanks
Lovely explanation, bravo!
very well explained. thanks for this content
What an amazing explanation .. SUPERB
OMG, this explanation is soooo clear
very well explained! Thanks
You are an excellent teacher
Amazing video, thanks 👍.
I don't consider AWS to be serverless because you need to configure your instances and you have to update the software(os packages, wordpress, mysql etc) running your site. You also have control of how the scaling is performed. Platforms like Vercel and Netlify are truly serverless as you simply deploy your code and the rest is taken care of. There's no instance to configure or any initial resources you need to allocate.
AWS has many regular but also many serverless products. Check out S3, DynamoDB, Lambda, CloudFront, Aurora, SQS, Kinesis, ... All of these don't allow you to fiddle with the software. They scale up & down without you intervening.
@@simplyexplained Agreed. Thanks for the clarification
Great explanation!
I think it can also be misleading to think that moving an on-premise infrastructure to the cloud saves money. For example, we greatly depend on fast disk I/O (high IOPS) in what we do and if we were to provision the same storage capacity with the highest possible IOPS, it would cost an arm and a leg from the usual providers. I think it makes sense for non-niche environments.
I think the video should also address some of the security issues (pros and cons) with hosting your digital assets "in the cloud."
Nice explanation thanks
I guess server'less', like wire'less' just means we the end user doesn't have to face the actual server/wire when using it.
This sounds more like moving your stuf to the cloud in a SaaS, IaaS or PaaS system. Because those are having the same characteristics of what u are saying. You should have made a video what the difference is between cloud-models and serverles
well explained!!!
I love these kinds of videos!
Glad you like them!
@@simplyexplained They are seriously awesome. You are one of my favourite youtubers!
Really helpful!!
very well explained
you didn't mention there is a "cold start" when the request hit your lambda function, which is a deal breaker for almost all business out there.
True, but cold starts are only a problem when your application receives no traffic for extended periods of time. In case of Lambda: no cold starts if your function is frequently hit (say every few minutes)
Estimation of traffic to your servers almost impossible. So it cannot be ascertained whether there's overprovisioning - waste of costly resources or underprovisioning - crashing of system
How can they charge you for cpu using? Technically,it’s quite a lot of work.
Excellent
Very useful. And please do the kubernetes serious.
Thanks
I dont get it, do i still have control over the OS that my service/site is hosted on? What if i want to SSH in and install some package that is required for a specific process that my service needs? I feel like with serverless im giving away a lot of control and the power to develop and customize processes for whatever i need them for, instead i can only do things that are 1) allowed by the cloud, and 2) compatible with the cloud. Or am i getting this wrong?
It really depends on where on the technology stack you purchase the service. IaaS (Infrastructure as a Service) will get you access to a virtual machine like AWS EC2, where you can install custom packages and give you more freedom. FaaS (Function as a Service) like AWS Lambda is at the other end of the spectrum where you only code the function input and output and all (almost all) of the infrastructure details are abstracted away and you are completely locked in to the "power and control" of AWS infrastructure.
You are correct: with serverless you're giving away a lot of control. If you take compute products as an example. AWS Lambda (or others) only allow you to upload code, specify how much memory you want and that's pretty much it. Installing additional software on the system is not possible. Everything must be a part of your application.
It's very limiting, but also very liberating if you can adapt.
I really don't understand "Serverless" like... yeah.. all of these "functions" are executed on a server somewhere...
Isn't server just a software? And the corresponding hardware just decides how powerful or reliable its gonna be. If you want, you can use Raspberry PI as a web server and serve your website from it. 🙂
You can look at it in both ways. A server is a piece of hardware (special computer, weird form factor, ...) but, like you mentioned, it's also a term for software that is capable of offering services to others on a network. A server can be both.
You don't have to worry about them? Wrong! You still need to manage/config/plan/monitor them, right?
Right. You don't have to worry about scaling (most of the time) and about underlying infrastructure. Some management is still required.
Sorry, but how can you have a database/website/file host without a server somewhere? The 'cloud' is just a bunch of non-local servers.
Is it justified to call a service *Serverless* if it is still working on some server under the hood. It's just a fancy name for PaaS based services.
Serverless simply means you don't own the server the functions are executing on. YOU are serverless, not your application.
You're thinking of p2p not serverless, only p2p can theoretically scale infinitely to meet the traffic; the keyword there being "theoretical" as in practice p2p scales terribly if not implemented homogeneously so each node contributes it's fair share , which mobile devices cause massive amounts of asymmetry, collapsing and fragmenting the network once it reaches a particular size.
All that meaning this distinction is meaningless.
Do Kubernetes and Docker please.
I'm gonna start marketing errorless code. We will create the errors for you.
Haha, fun idea! Reminded me of "Chaos Monkey". A tool from Netflix that kind of does that ;) github.com/Netflix/chaosmonkey
“The serverless service provider adds more servers” - doesn’t sound particularly ‘serverless’ to me.
What you are explaining here is managed vs unmanaged servers.
As long as you connect to a specific address (domain or ip) it’s a server.
The only true serverless approach is P2P i.e. Peer to Peer.
This is what happens when somebody makes a video before they know what they're talking about. Unfortunately these days that seems to be about 99% of the time. You are conflating serverless with cloud computing and mixing up the benefits and the descriptions as if they were all one thing.
It's always easy to tell when someone doesn't actually know the definition of the thing they are trying to tell you about. THEY NEVER ACTUALLY DEFINE THE THING. All they ever do is talk about the features and the benefits without ever telling you what the thing is or how the thing actually works.
With cloud computing you still have to be concerned about a server. That server may not be your own physical hardware. And that server might be virtualized or inside of a container. Your cloud service provider will spin up multiple containers if you need them. But you still have to have configured those containers that are virtualized servers...
With serverless computing you can forget about even worrying about those containers. You never have to configure a server and you do not have to worry about how many of them are spun up or not. You do not design a little mini server to handle a little microservice and then put that in a container and then ask your cloud provider to spin up multiple copies of that as necessary. With serverless computing all you do is write a function. You give that function to the serverless service provider and let them worry about whether to put that on its own separate server or not. Rather than spinning up containers with their own abstracted away operating systems they simply allocate memory and CPU time to run that function and send you back the result. You never even had to think about a server or configure a server or containerize a server. That is the difference between cloud computing and serverless computing. Yes, all that is still on a server. But you never had to spend a moment of thought as to how that server was configured. You don't care, as long as your function returns the correct result.
What boggles my mind (AKA "frustrates the hell out of me") is how much research I have had to do to figure out the difference because there are so many people like you who are in a hurry to jump in and say something, ANYTHING, so they can get the likes and views before anyone else, that they get it completely wrong.
Great post! This explained a lot! :D
So we don't have to manage those Nginx servers and it's configuration, do we?!?
Yep!
Cloud is great until you get the surprise $$$ Bill from the provider
The unpredictability of the "serverless" provider's policies and the unpredictability of the costs of using their service makes this an absolute no-go for our business. There are way too many nightmare stories of people checking their account in the morning to only realize that their provider has charged them for multiple tens of thousands of dollars due to an unpredictable event or, even worse, a simple programming mistake.
От души братан! Thanks!
A marketing spin on online services.
the detail of the B in the money bags instead of $ 👌🏽👌🏽👌🏽👌🏽👌🏽
Yeah this is - 20 years ago?
Can smartphones & desktops be used as servers combined, there are millions of them, it's like running blockchains.
There's already a file storage service that allows desktop owners to share their unused disk space for some small bit of compensation. Smartphones are not a good fit coz they are optimized to save battery consumption.
@@myxsys What is the name of the service.
@@palabinash Sorry, I can't remember the name. I came across it last year
@@palabinash "sia coin"
Aren't we really talking about IaaS vs Saas?
This is just software as a service, as an end user you might don't have to worry about servers, but someone else will.
serverless messaging is really for low-capacity apps. Those caps suck for anything more than 100 msg/second
So, serverless is like a taxi. You are in a car but we just don't call it a car because there is a sign on top saying taxi.
To me serverless is just marketing b.s.
It only sounds like marketing BS when idiots like the guy who made this video conflate cloud computing with serverless and run out and make a video with lots of animations just to fool people and make them think they know what they're talking about.
the advantages of big tech serverless are: its cheap and easy. The disadvantages: the know all of your traffic/business, so your completely the pig in their barn. And when the time is right the slaughter you like parlor. So you save 20-30% of server costs for 3-6 years and then lose 100% of your business.
This is more "pay as you go" than "serverless" Cryptocurrency, Torrents, and Open Distributed File Systems aka peer-to-peer are "serverless".
Technically even peer to peer is server to client once resources are requested.
3 to 5 years life span... and then on AWS you have a cpu from 2008
Yoooo!!!
Just sounds like someone give you access to a server to me. Also if your hosting a small website you can get a small VPS that can handle thousands of visitors a day for like $5/m. This seems absolutely pointless to me.
Could've used the term cloud,its easier to understand 😕
"The cloud" and "Serverless" aren't the same thing. You could for instance rent a server in the "cloud" but that wouldn't be very elastic. It wouldn't scale automatically, not patch itself against security issues, it won't charge you only for your usage, etc...
@@simplyexplained Google was probably high af when they named their "serverless" platform "Google Cloud Platform"
Hi..I have an environment variable 'stage' and I want to assign another env variable 'isVariableProducion' value as true if the 'stage' value is prod..how do i do that ?
Managed services.
AWS: "Nice website you got there. Would be a shame if someone... CLOSES IT"
Can't every hosting provider do that?
@@simplyexplained Absolutely. But It's funny cause of what happened to Parler. But like stated, there exists abstractions. Parler neglected to use one, and instead chose to trust AWS. That was a mistake they paid dearly for.
why deploy serverless when you can make a cluster of raspberrys and deploy there. It's a lot more affordable, serverless should be for development. and if you are worried about ddos, you can rate limit your app, also you can code a ssh jail honeypot and attach it to the network.
What are you talking about, why should it only be for deployment?
@@slapmyfunkybass some serverless enviorements charge by the number of request handled, in combination with the size of those requests, the impact of a ddos is on your wallet, even with rate limiters implemented in your app.
"Serverless" in this context is marketing jargon... it's simply pay for our servers so you don't have to build and run your own.
TLDR: Cloud like AWS = Serverless
Surprise: So called "serverless" services still use servers.
True serverless is running things on your local computer or via a peer-to-peer connection.
This is one of those terms which are complete waffle. Serverless basically means walking into a business, picking up their server and moving it externally. There’s nothing serverless about it, it’s still maintained, can still crash, still has limited resources. Serverless is basically a marketing term to persuade the business to pay externally instead of internal IT staff.
I don't have a butler so I am serverless
"This is called overprovisioning." That's just your fault. You should be solving chess endgames when your server has no requests.
Or mine some crypto coins 😅
instead of using this fishy server less word , it would be better to be honest and use some clear terms like managed and unmanaged ,
yeah idk i have a small website running on a literal raspberry pi
i can serve about 500ish concurrent users (not concurrent requests!) no problems, and i have NO optimizations in place and no caching set up (except that i use no framework and host my pictures elsewhere)
if you have a very simple website with no large user base and you are ok withup to 3 minutes of downtime a day, just pay the 50 bucks startup costs and get a raspi, an extra ethernet cable, a usb drive, an sd card and a case (and maybe some cooling idk) and take the 2-3 ish hours to configure it all then you are up and running for basically no costs. Raspi takes very little electricity and the internet is just your home internet. Better your website slowing down than paying hundreds of dollars after getting ddosed (or having a blogpost go viral)
I say "po-tay-to", you say "po-tah-to".... There is nothing without a server on the internet. IoT, cloud - whatever you say...they all require server. The "Serverless" is just a word to trick you to use the cloud server. Instead of deploying and managing your own server you are just using the "Already Deployed" cloud server. 🙄🙄🙄 And "Serverless" is just a feature to attract customers. It only benefits you some cost reduction with sacrificing some features. That's all. Nothing else.
wireless is exactly that, what's he smoking?
Serverless products are the New way to implement authoritarianism through capitalism. Thus capitalism IS real authoritarianism and communism is real liberty
Explains "Serverless" but not "The Serverless Hype"
Please help me
Only the Amish are truly serverless.
A lot of bull shit.
Well this is probably the most disappointing video I've seen sofar. I am a total nitwit when it's about servers/networking. Watching video's is how I get lectured. I finally got a video I totally understand, I never heard about serverless ...something.. This video explained it; I felt really intelligent and got exited. I want this too... But then in the end it tells me just don't use it because they are all different and .. well I allready forgot... I like your video's but please: next time start with the BAD news and then tell the GOOD news ... (life is allready soo complicated for me and a lot of other people)
I understand. But in this case it would be very difficult for me to talk about the downsides of serverless, without having first explained what it's about. Also: the downsides aren't that extreme. I know a lot of companies that use serverless backends and have been extremely happy with them. I've designed a few myself, and being aware of the limitations is very important.