And it doesn't come with any Storage or Network Transfer. You will soon require a Phd to figure out the cost optimisation on AWS, or simply switch to DO for a much simpler pricing.

I just can't fathom why AWS is so popular given its high mental overhead. Every other VPS provider is like, "Look! It's a server with local storage and memory and CPUs. It comes with this much bandwidth and this much storage. If you pay us you can have it!" AWS: "Hold my beer."

EDIT: I get it, I realize that the value of AWS is in the whole basket of services you can use, not just EC2, but damn do they make it unpleasant. If DigitalOcean gets more enterprise-serious and adds more hosted services, AWS really better watch out.

Most people who buy EC2 instances at scale won't want most of them to have persistent storage attached. Storage is handled by a DB or S3.

Right? I'm surprised at how many of the comments here seem to be about using AWS for hobby-scale stuff... I keep a Digital Ocean VPS for my personal website and OpenVPN AS and stuff like that, but anything I'm deploying for work on AWS is... either totally transient (I just need compute and enough working space to store the code I'm executing) or critically important (local storage is not sufficient, I need redundancy and shared state and lots of 9s). I can't think of many commercial use cases outside of prototypes where you'd want a bundle of compute + storage all on the same hardware.

Its like we stopped the "cattle vs pets" metaphor a bit early and some people are mad that a butcher is selling steak instead of a whole cow.

(And just for the record, my team has pretty minimal AWS usage... we're not one of the cloud kool-aid shops. We buy overflow worker server capacity on EC2 spot instances, but our primary app stack runs completely on colo'd bare metal for our day to day operations. Even on our bare metal though... individual app/worker servers are disposable and persistence is provided elsewhere)

They also do not give you any easy to understand centralized view of where you're spending money.

I spent a week trying to figure out why AWS kept charging me a $1.36 every month. Turns out I had a reserved instance that I forgot about. However damned if any of their "cost explorer" can help you at all.

The only way to deal with it is to open the EC2 console and then switch to each zone one by one until you find one that has any running or reserved instances.

The AWS console is a joke, I have no idea why its so popular.

While I agree that AWS pricing and UI can be a nightmare for individuals to understand, at every mid to large enterprise I've worked at they simply haven't cared - a corporate card is entered in an admin screen somewhere and forgotten about.

I honestly don't think this matters to most serious AWS (ie paying substantial amounts of money) customers as much as it maybe matters to individuals looking for a 5 buck VPS to host a side project on. The small players like Digital Ocean/Linode/etc exist and cater more easily for this business if you want it, and it doesn't seem to be helping them make major inroads against the big three.

Well there's still an advantage to that flexibility. Sure it's fun for a random project to have average value to every specs, but some project require much more RAM than CPU, or much more bandwidth, or the more usual much more storage. If you need any big amount of storage on Digital Ocean, you need to take one of their beefy machine, even if you don't need that much RAM or that much CPU. Sure there is object storage, but that's a whole different storage system and it doesn't play well with a database server.

When you actually need specific specs at a huge scale, it's not much more complex to do it over AWS.

for a database server (or for anything that just needs a big filesystem), DO has block storage, which just looks like an attached hard disk. Works fine with a small VM.

they also have memory-optimized or cpu-optimized VM choices.

Seems like GCP is best for sliding amounts of compute, because you can create a machine with whatever RAM/CPU combination you want (more or less).

I'd argue that buying the beefy machine at DigitalOcean would still cost far less than AWS. And you'd be getting on with your life versus banging your head on AWS for the day, week, or year.

If you think that is confusing, try ordering Microsoft SQL server on Azure which was measured in "DTU"s - data throughput units. Which globbed together disk, CPU, and RAM into one "easy-to-use" metric!

Actually, you have a per-vCore purchasing model now.

Management: "Increase revenue per user!"

Azure: "Introducing DTUs, the simplified, easy-to-use pricing metric for SQL Server!"

They offer Lightsail for exactly this purpose.

I think amazon has reached the tip. So many side projects that lay there abandoned because if im not setting up my own vpc subnets and gateways manually i feel defiled and ashamed.

This is what amazon has successfully done, raised the cognitive sunk cost so fucking high you need to pay for their bs certificqtion i didnt even study but somehow passed even aftwr arriving to the exam 30 min late because the city of vancouver are such fucking master extortionist of parking tickets and somehow everybody has decided to park...i digress

Tldr use do or linode....dont soend decade on aws and realize the greatest devil the trick pulled is that he dont exist well lemmme tell u friends unshackle yourseles fromm aws and use netlify

I know you're being downvoted, but this sentiment toward AWS is exactly what will eventually kill it. Why should someone waste their life away learning the complexities of AWS? It's not a lasting addition to one's life like, say, learning biology or history or how to write. It's needless complexity that won't have any value in 15 years.

Dunno about you, but I have not found many "complexities" in the parts of AWS that people actually use that don't map pretty directly to using other cloud providers or running your own systems. I look at an AWS thing, go "oh, okay," and go about my day.

Learn fundamentals, apply fundamentals. It's pretty simple.

Sounds like a startup opportunity. :) RackSpace for hobby projects?

+1 for Netlify

Dedicated AWS instances for machine learning AI-assisted AWS cost optimizations would be the next bestselling AWS product.

You mean this thing they just introduced yesterday? https://aws.amazon.com/blogs/aws/new-predictive-scaling-for-...

Product configuration can be very efficiently modelled as a CSP. You should use MiniZinc instead

'Our product line pricing is so simple you don't even need a FPGA to figure it out!'

Did Amazon hire the MS licensing folks to work on their subscription prices?

No because at least they give you the prices! MS would always say you need to 'get a quote'...

Azure prices are (notoriously) public, and there is a nice public calculator available at:

https://azure.microsoft.com/en-us/pricing/calculator/

(link may redirect to localized versions)

I have no idea where you'd get the "get a quote" thing, other than corporate customers usually needing (or wanting) some hand holding and guidance given that a) most of them already have applicable discounts as part of current enterprise agreements and b) accurate quotes require more than guesswork at sizing - we take into account usage patterns, whether or not the customer can benefit from reserved instances (which significantly lower costs 1-to-3y down the road), etc.

(I'm a Microsoft FTE working on Azure solutions, and have a personal Azure account which has detailed, per-resource cost breakdowns complete with links to the public pricing pages...)

I replied to someone making reference to MS licensing pricing (like Exchange CALs and such), not Azure pricing. MS has done great work in the cloud with Azure.

Azure prices are public, unless they've funneled you through a CSP (which Microsoft loves doing)--at which point pricing is a black-box nightmare with even worse support than Azure has by default.

Not so fast there, you need the associate certification. If younforget to bring your id and basically flush a cool few hundred american

AWS is a low-level service provider; on top of that you get services like DO or Heroku that abstract the complexity away for you. But that only adds cost. Basically you exchange complexity for cost, or, pay more for having to do and know less.

Digital Ocean has their own data centers, they aren’t built on top of AWS

https://www.quora.com/Does-DigitalOcean-have-its-own-datacen...

I wonder how much (volume) discounts you'd need from AWS to profitably host a digital ocean on top of it...

You’d be competing with Lightsail. It’d be hard to compete with Amazon on their own hardware.

Spot instances are about $3.50 a month, for what it's worth.

I wonder how long the spot price will remain this low. It does fluctuate after all.

Ooh. Thanks for reminding me about spot. That's way better than the on-demand price. (Still 1/4 the cores of the Scaleway offering though.)

The a1.medium 12-month reserved cost is $151 which is about $12.60 a month.

I think the high price is to amortize the Annapurna Labs acquisition and to pay for NRE.

Now they will own the entire vertical stack like Apple, but unlike Apple, Amazon can live on minimal margins.

BTW, apparently I predicted it in 2011 ;)

"Rumors about ARM-based EC2 instances from Amazon this year. Also SSD based EBS. #aws #ec2 #arm #ssd #cloud #amazon

3:14 AM - 27 Feb 2011"

https://twitter.com/nivertech/status/41667359487823872

> I think the high price is to amortize the Annapurna Labs acquisition and to pay for NRE.

I don't think that's a useful way to think about pricing a product or service. The price has to be driven by the market, not your costs, which means ultimately your costs have to be driven by the market as well or you're done for.

I suspect they're pricing this a little high because they don't have a lot of them yet, meanwhile plenty of people will pay a bit extra to get a chance to try this out and do proof-of-concepts on them. As long as the market created by the latter matches the availability due to the former, this will work out fine for Amazon. They can always lower prices as these become more mainstream. Novelty value can be real value.

Neither google nor microsoft have ARM CPU's yet, so there is as-yet no price competition.

Scaleway.

FOUR ThunderX cores for THREE € per month.

AWS offers one core for TWENTY $ and with the typical extra charges… excuse me but what the actual fuck?!?!

Only for situations where ARM is an absolute requirement, which are few and far between.

Linaro has an ARMv8 cloud, too: https://connect.linaro.cloud

Packet.net (bare metal cloud hosting vendor) have ARM instances.

Agree, this is still quite expensive.

It seems like a better value than a t2 instance because it's not a burstable core. The network performance is probably much better on this instance type as well.

I was hoping for something in the $0.10 /month range. An A53 type multicore ARM CPU works out at only $5 or so to buy, and only $3 per year of power to run, so a timeshare on a single core of that machine ought not cost much!

Are the employees free? As well as memory, motherboard, etc?

t3.nano instances already start at about ~$4 (+ ebs and network). Their lightsail alternative pricing model for ec2 instances also has $3.50 and $5 instances [0].

[0]: https://aws.amazon.com/lightsail/pricing/