Is anyone running a "serverless" setup for webpage/PHP/database on AWS or Azure?

[Matt 26z]

I know that you can serve static pages/images/video without maintaining a server by using their "bucket" services.

If I understand this correctly, you can use their compute services to run the PHP. Then use their database.

What I'm not fully clear on is if this stuff can all be tied together to totally eliminate maintaining a server (OS, webserver, etc). It seems like this is possible.

[sandman!]

It’s possible but with any real usage it will cost you 2-3x the cost of just renting vms or dedicated servers.

[wankawonk]

It’s possible but with any real usage it will cost you 2-3x the cost of just renting vms or dedicated servers.

it's probably closer to 20x-30x the price...

AWS is *expensive*

[Ferus]

it's probably closer to 20x-30x the price...

AWS is *expensive*

No.. Only if you move your legacy shit to run serverless.

Decoupling you app first, then run it as functions. You can scale in and out as you wish and stretch across regions as you wish.

All I do now is moving customers to cloud native solutions. Most try themself first and fail hard because they dont know what the technology can do.

I have worked with online-training communities that have 5000 concurrent users streaming video, that cost the same as two small VPS running at vacares.

So if it cost you 5x its because you are doing it wrong.... ALWAYS use certified people (and not entrylevel certs)

[Why]

it's probably closer to 20x-30x the price...

AWS is *expensive*

this is completely false, its no where near 30x expensive, because when you are using serverless you only pay for then time spent computing.

source: i have serverless resources executing 50m+ tasks per day and cut costs by 80%.

[Why]

So if it cost you 5x its because you are doing it wrong.... ALWAYS use certified people (and not entrylevel certs)

spoken like someone with certs whom likes to justify high prices by pointing to a cert.

certs dont mean shit! lets be real, you can study for them without ever actually using any underlying services.

[XSAXS]

it's probably closer to 20x-30x the price...

AWS is *expensive*

Every time I look at AWS this has always been my understanding/conclusion. Wayyy too expensive and FUCKING complicated to wade through their insane terminology, options, pricing, gotchas.

this is completely false, its no where near 30x expensive, because when you are using serverless you only pay for then time spent computing.

source: i have serverless resources executing 50m+ tasks per day and cut costs by 80%.

Very possible (in fact, probable) my conclusions have been poorly informed. But so far, I haven't found a good enough reason to put myself through the torture of mastering AWS, especially when there are other simpler options available.

[sandman!]

A few years ago I spent a few hours pricing out a clients usage to move to aws the clients cost using their own hardware aka colo to aws the cost was 3 k colo plus about 80k of hardware to 15 k aws , anyone that can do math will figure out colo is way cheaper in the long run.

this is completely false, its no where near 30x expensive, because when you are using serverless you only pay for then time spent computing.

source: i have serverless resources executing 50m+ tasks per day and cut costs by 80%.

[Ferus]

spoken like someone with certs whom likes to justify high prices by pointing to a cert.

certs dont mean shit! lets be real, you can study for them without ever actually using any underlying services.

You could never get a professional cert from AWS just by reading... Only the Associate level, and that is exactly my point

My prices might be higher, but Im faster, make less mistankes, you get aws funding and aws discounts you cant get without a aws partner, and if i fuck up, i have direct access to 3rd level support.. Global..

But its your call... I dont care. I save my customers a Lot of money and have spend thousands of hours training, and love working on New projects, so I dont need your approval

[wankawonk]

make less mistankes

[Ferus]

A few years ago I spent a few hours pricing out a clients usage to move to aws the clients cost using their own hardware aka colo to aws the cost was 3 k colo plus about 80k of hardware to 15 k aws , anyone that can do math will figure out colo is way cheaper in the long run.

It's because you are comparing 1:1 and that is how most customers end up burning maybe 5-10x what they really should. They think "I have 5 servers with 8 cores and 16GB ram each, I need the same in AWS"
... if you do that, you will get a shit solution. NEVER use Cloud as IaaS or Colo (exept maybe for DR or Burst-to-cloud) but that is not what we are talking about here.

If I get a customer that want to do a lift and shift, and I cant convince them to at least put refactoring the solution on the roadmap, then I wont work with the customer. I can not afford to have my name tied to a solution, that dont leverage cloud-native solutions.

For 7 months I worked for the world's fourth largest freight forwarding company, building a HCI solution for onboarding all their legacy applications, we are talking two digit $million savings a year from DAY1 - all Opex and no vendor/tech lock-in. I know what I'm talking about.

I have another example. One of the largest regional pharma companies wanted to go full dev-ops on their infrastructure, and move all to either OpenStack or OpenShift. I was asked to review their plan. Both involving two colo-providers, darkfiber interconnect and half a rack of Lenovo servers + ToR and Mgt switches + complete leaf and spine network + routers.

The rough calculations I did for them showed, that even without training, you could run the same load on Azure or AWS for 7 years, before break-even. Have you seen many that uses the same servers for 7 years (successfull people that is) - but what they forgot was they had to replace more than 50% of their staff, because OpenStack is not something you just learn overnight (And OpenShift is somthing you never learn unless you work at REHL) - so they ended up sticking to their VMware colo solution, because they dont want to fire people.

Im not saying traditional hosting is bad - its just not the same. There are many cases where a traditional colo or fully-managed solution is the best, both when it comes cost and business-strategy.


Cloud is expensive if you use it wrong.. same is Colo and anything else

[Ferus]

You dont think; experience = less mistakes?
Let me put it another way they; The mistakes most is about to make, I have already made it at least once and learned from it. So that is why you are saving time and money.

Try and do it yourself from the start, you will have to redo +75% of the mistakes everyone else have

[grumpy]

https://news.ycombinator.com/item?id=7172060

[NatalieMojoHost]

We have a serverless offering too. However, for the vast majority of clients, it is not the right option. By and large big parts of the costs for adult sites are storage, bandwidth, and databases.
The cost of modern storage is pretty much the cost: no matter whom you buy from they're buying hard drives which have become highly commoditized, so the cost of storage is about the same everywhere.
When it comes to bandwidth we really are 20x, 30x, even 40x less expensive than Amazon here at Mojo.
When it comes to the cost for running databases, of course in a perfect world the code should be optimized to minimize the database server usage. Unfortunately, many of the things people want their databases to do are computationally expensive (especially in relation to member sites), and many of the software that they're running isn't very well optimized anyway (that's why we work hard with the biggest CMS providers in the adult space to help them in improving their product).
Amazon databases tend to be verrry expensive. It's true that if you can make your problem fit exactly into their server less workload, you may save some money compared to your existing setup. But most of those savings will be due to writing good, new software that takes advantage of the state of the art, and not due to anything related to serverless itself. In fact, in most cases for $X spent on making a project serverless, the same amount of programming work on dedicated servers will still probably be cheaper.
Serverless tends to make the most sense when you have really high click volume for a particular resource and that resource needs to be customized for each user and you can customize it with only a small amount of data. One example where this works well is creating a custom version of an image which contains overlaid text specific to the user and you're doing it 10s of millions of times a month. That's a great use case for serverless. There's some others, but by and large serverless is unlikely to be the right fit for most adult industry use cases.

[kjs]

Shitty bloated code is *expensive* to host anywhere.