Serverless everything, agree?

If not, why?

Serverless restaurants?

No.

Serverless git hosting?
Serverless backups?
Serverless media server?
Serverless PCI compliant storage and processing?

“Cloud” solutions are not always better. They offer better (read: easier, cheaper) scaling and reliability, and less control. Increased privacy concerns and legal exposure, too.

On premise ftw

I would say most things

@Root would probably agree with the others but what do you mean by 'serverless backups'?
In my experience having a serverless database or other persisted storage in a serverless architecture makes backups trivial.

@Root
Isn't serverless git hosting just using something like GitHub?
And you can be PCI compliant by using something like Stripe.
Serverless doesn't mean you have to do all the workload in your own infrastructure. It's about concentrating on your own thing and the business value that brings and farming the rest out.

@dan-pud I mean don’t trust your backups solely to some cloud provider. Keep a copy (ideally two) you have physical access to as well. If the provider closes, cancels your account, or loses your data, you’re screwed. So for backups, always have at least these three:

• Cloud (your daily; trivial to access)
• On-site (outage/cancellation insurance)
• Off-site (fire/flood/fry insurance)

Ideally these should be read only and encrypted as well.

@Root I don't get it, we can't lose on the cloud

We have our code in github

@dan-pud To paraphrase your Stripe comment: “you can be PCI compliant by not needing to be PCI compliant.”

Imagine a scenario where you need to be PCI compliant.

@simpledev “we can’t lose on the cloud”

The cloud is just someone else’s computer, and accidents are a thing. So are outages, dead drives, and malware. Also incompetence, changing legal landscapes, etc.

If you trust your data to a third party, and that third party closes its doors, you lose your data. If the data center it’s in catches fire, you lose your data. If your data gets encrypted by malware, you lose your data. If someone maliciously deletes your data, either at your company or theirs, you lose your data. Etc. Etc. Etc.

I read, though cannot find with a 30 second search, an article detailing a company that was forced to close its doors because they kept their code and production data on one service, and that service closed unexpectedly. With no other backups, they suddenly had no product.

Single points of failure are bad.

@Root
Stripe is PCI compliant and as user of the service you are simply passing the responsibility for the majority of the work.
It's more 'if they are then I am'.
I agree there are certain scenarios where you'd probably not to want to use stripe or others and do it yourself but this obviously comes with a lot more work and resource. If I can I would take the serverless approach and focus that resource on adding more features to my product and I think this makes sense for the majority of other companies.

Lol they said serverless. What we really need are carless cars.

Serverless = Servers around the globe you can't control
It should be called "server less-control"

I'm 3 minutes ahead of you buddy, I do full client-side only projects for "fun" 🥴

@LotsOfCaffeine "Serverless on premise" sounds like a fun idea 🤣

@PonySlaystation "we host it ourselves we just don't know on what machine it runs..."

@nitwhiz you mean good public transportation? I agree.

I feel like severless is half assed decentralized internet. We could be moving most things to P2P solutions and using the to coordinate consensus instead of using shared systems to hold everything.

Most people have a mobile phone with more computational power than a web server pod. Why are we not using that imbalance better?

@irene I agree and I don’t.

On one hand, I hate inefficiencies and waste. On the other: it’s my device! It should only do what I tell it to. The same goes for everyone else’s devices, too.

Now, if it were a distributed system like seti@home, that’s optional, paid by the cpu minute, and sandboxed to prevent privacy issues, absolutely!

@Root @jfgilmore I’m not talking about doing a bunch of idle things on your phone while you aren’t using it.

Think about how much work we offload to a server because we can’t trust the user browser. At that instant the whole client exists on the phone, the client has data needed to calculate a response. The work we do on a server is mundane but it becomes expensive because of consensus checking; we check the user authentication, authorization, use a third party cert, check the values that a client submitted against values that the server had, then we do the calculation and reply with success. Alternately if a few other randomly selected users that are running the same client do the calculation, we can compare to see if they get the same result and know if a client is doing anything janky.

So I’m not talking about crypto mining when your phone is off. I’m talking about doing arithmetic while you are interacting with the app so that the main node is rarely if ever needed.

Sever less server for hosting a server less server without server less server then yes

In addition to the good points @root has made, there's some practical limits which mean it just simply doesn't work in all scenarios. What if I might need more than 1k concurrent functions and my provider limits it to that? What if the jobs might take longer than 15 minutes and I have a similar limitation to that execution time?

Even if we *wanted* to use serverless for everything, the major cloud providers all have similar limits meaning that it simply wouldn't work for our setting. It'd also be cost prohibitive - spinning up a lambda once in a blue moon is infinitely cheaper than running a full redundant cluster 24/7, but it's ridiculously expensive when you have a constant high traffic load.

@AlmondSauce no it's not, many giant news site use lambda

@simpledev That doesn't mean it's not ridiculously expensive. I've done the analysis and crunched the numbers.

Chances are the big news sites got in there early with special deals (happens all the time with AWS) that meant they have lower costs and higher limits - and they likely simply don't care too much about cost because they care far more about being able to scale up instantly, without spinup time, for spikey traffic loads. If that's not a realistic scenario for your use case, then that advantage isn't there.

@AlmondSauce oh I see. Now it makes sense. But lamda is good for startup too, before it gets traction, you need to calculate the cost of hiring devops if we use the traditional way. So it's not serverlees versus Ec2, it's serverlees versus the entire devops team

@simpledev EC2 and lambda aren't the only 2 compute options, even on AWS. We use neither.

@AlmondSauce wow, what you're using?

@simpledev ECS with Fargate instances at the mo.

"Serverless" is pretty misused by now. It should stand for "without servers" - so no cloud either as the cloud is just other people's servers...

Then, that word would actually mean something:
Do it local or do it peer to peer!

That would be a completely different web - quite censor resistant but probably pretty slow and unreliable...

@Oktokolo Yeah, its got to the point where I don't even use the term if I can help it. Just means something different to everyone these days, and it's a crappy marketing term anyway.

@Oktokolo hmmmm what about calling it sharded?

@melezorus34
Is "shard" the new marketing term for "cluster"?

@Oktokolo aren't the functions made very short so the lambda thing becomes fast? I thought of sharding the backend like that and came up with the word.

In some applications it refers to horizontal scaling too, i think, so it can create even more confusion and chaos nyeheheheee

Out of curiosity, how many projects do you have hosted on serverless platforms and how much do you pay monthly?

Because i rent a vps for like $10 that can handle most things including real time multiplayer servers

@yellow-dog I pay 0 and use heroku lols

@yellow-dog few projects, idle or low traffic. Paid £0

@melezorus34 @simpledev yeah so like, neither of you uses serverless for any high traffic project lol

@yellow-dog yea don't remind me T_T

@melezorus34 i work at a company that uses a cloud provider very heavily for all of the software, and we pay insane money, we could roll our own server farms in multiple locations for the price of maybe 6 months of bills

@simpledev News sites are pretty lightweight though. Most of the video and picture content is on a CDN and the page structure can be cached in the browser. Advertising is hosted through a third party. Most of the traffic will be directed to only a few articles.

News sites are basically a big blog where the heavy content is done elsewhere.

@irene I agree with you, news sites are heavy... but only on viewers' device.

so you mean off load to cloud with scaling on premise.

Serveless is good when you have spike traffic requirements, otherwise it is better to host/self host.