AWS pricing is a bit ridiculous. $100,000 can buy some much hardware even while taking account of the additional costs like electricity and hardware failures.

It's a lot, yes, but you're getting a lot more than just the instance: - You can back up the image with fantastic reliability and recover into multiple AZs automatically with an autoscaling group for practically peanuts. - You can place that instance into your own private VPC complete with security groups (stateful firewall), Network ACLs, and inter/intra AWS routing. - You get strong authentication and RBAC through IAM over the instance and it's environment.

That's what I consider in the "free or nearly free" tier, off hand. The other benefits come with being able to interface seamlessly and quickly (same infrastructure) to the rest of AWS services.

You might do better at finding a piece of hardware that does that (and I'm curious now what a 2TB RAM server goes for) but I think you'd be hard pressed to find a way to start from scratch to deploy that instance and all of the services that come with it for under that price. People with on-prem compute likely have some of that already, but the value here is that you could request an X1 today without ever having been a customer before, and you'll get all of that, and access to more, just with that one instance.

If that's not a good value proposition today, then I'd say wait just a few months. Today probably marks the highest price anyone will ever see for an X1. Given past history, it's just going to go downhill from here.

Also, the common use case of "I don't need this hardware 24/7/365."

If you need to do some analysis or computation on a massive data set once a month or something, it's going to be cheaper to pay $5k/yr (assuming you run for 24 hours a month and don't make use of the spot market) than to purchase and maintain the hardware and infrastructure.

The AWS instance is very predictable, in operation and cost, and that's valuable to business.

If your unique snowflake in your own datacenter (don't forget to factor in the physical space and your datacenter personel into your costs) doesn't work well, it can mean replacement, additional costs and downtime. If the AWS instance has a hickup, terminate and replace (I'm not saying that's going to be trivial either at a 2TB RAM instance size).

A big hunk of hardware of your own also represents significant CapEx and a depreciating asset for business. Spinning this monster up costs you $13 to start testing immediately, and you can walk away from it at any time. That's worth a lot.

"Fail quickly, fail cheaply."

And the staff costs of good sysadmins, and the reliable network, and everything else that comes with AWS?

For some use cases, the pricing of AWS makes sense.

I tend to agree with you on AWS. I swear by it. But in this case I would probably go at it this way...

How many hours can I run the machine before it costs more than building one? Probably a month? (random guess) Will I be running it longer than that?

If yes, build the machine, if no, just rent it from AWS.

I can't even imagine a scenario where I run this 24 hours a day 365 that I wouldn't build out a Hadoop cluster or similar for.

Well, the advantage of AWS is that you don't have to run it 24/7 unless your workload requires it.

Your math is off. At $13.338/hour you're looking at just shy of 10 grand per month ($9763 for a month of 30.5 days.)

It looks like one of these bad boys would set you back about $40000. So you'd break even at 4 months. If you're going to go for the 3 year reserved instance, with $100K upfront, you'd be way better off on capex and opex just buying and colocating the thing (not considering other expenses.)

The bigger you grow the more the balance tips towards do it yourself because your costs for colocation, managing hardware, etc amortize out better with scale. AWS gets a little better with scale but not nearly as quickly.

My math isn't off... it's for 3 hours a day and I state that in my post (perhaps not clearly enough). Your number is for 24. The point was if you're not full utilization it could make sense.

People using AWS are not typically, say, small business people looking to eke out savings however possible so they can take an extra vacation or buy an new car. They are corporate people or people running startups using other people's money. Their success or failure doesn't depend on saving the organization money so they are willing to pay more to have less to worry about themselves personally.

Quite the opposite. If you're a small business in need of compute, then it's quite a value.

I'll give you a case in point. I have a friend who works for a construction company. He has three servers he's cobbled together all at one site, and a NAS for storage. They provide email and document sharing, accounts management, etc. Every new blueprint that comes through needs to be reviewed for accuracy, but they come in as images, so he has another server that just runs tesseract OCR when he gets a pdf in. He asked me if he should get three servers, or one and virtualize the three. He wanted to plan for growth, but didn't know what that would be. the NAS was underutilized.

He's buying no servers. He's setting up Workmail and Workdocs for their users. (Workspaces is an option in the future.) Blueprints are uploaded to S3 as they come in, where the file arrival notification triggers a lambda function that performs an OCR pass on the file and dumps it into a new bucket. Because this function runs on-demand, he avoids paying for a server to sit there, and only pays a few cents for a doc. His other services no longer run on servers he has to manage, so he doesn't worry about patching. He has reliable, off-site backups for the first time and version control, and he only pays for the exact amount of storage he's using as opposed to buying space for his NAS in advance. And he doesn't need to worry about scaling.

That story plays out all the time. It's arguably a bigger savings for smaller people who can't afford the upfront costs than large corporations that already are likely to have massive sunk costs in datacenter space and assets. If you're a small business, the moment you go cloud you not only save money, but you gain capabilities (the biggest being in redundancy and reliability, which are the most important...how many small businesses have a DR/COOP capability on-prem?) which you would never have had otherwise.

> It's arguably a bigger savings for smaller people who can't afford the upfront costs

Well first let me say that I have been doing all of this (as a business person who knows tech and Unix) for quite some time. And what you wrote above spins my head because I don't know much about AWS (but I can easily turn up a server if needed). My point is what you are describing for a small business person requires a tech person knowledgeable with AWS to implement (and keep on top of) which is the "cost" as opposed to perhaps saving money in this particular example for this particular customer. I am no doubting that AWS (which I see you work for in your profile) is good for certain types of uses (such as perhaps the example you are giving). But you still need a knowledgeable tech guy to keep it all working it is not like using webmail vs. running your own email server (imap, smtp etc.)

Additionally once you are on AWS you are pretty much locked into AWS by the way (as you describe) things are setup. I am not convinced that at a time in the future AWS will not change their pricing to take advantage of the lock that they have (even though now they continually drop prices). Or add extra fees or whatever. Being in business so many years I have seen this happen and changing from AWS will be near impossible. (In other words for people on AWS such as you describe there is no easy alternative competition for a system once designed and specific to AWS in the way you detail).

> Additionally once you are on AWS you are pretty much locked into AWS by the way (as you describe) things are setup. I am not convinced that at a time in the future AWS will not change their pricing to take advantage of the lock that they have (even though now they continually drop prices). Or add extra fees or whatever.

This is just FUD. AWS prices are always lowering, there is never a time they have increased. Even if they increased their prices, Google, Rackspace and Microsoft would eat their lunch. There is plenty of competition in the computing space.

> Being in business so many years I have seen this happen and changing from AWS will be near impossible. (In other words for people on AWS such as you describe there is no easy alternative competition for a system once designed and specific to AWS in the way you detail).

If your solutions are so narrowly defined that they won't run on anything else than AWS, then you are doing it wrong. You might run on specific AWS services, but there should be no reason why someone could not recreate their solutions over on Google Compute or other services. Yes, you might have to rework some of your solutions, but it should be very doable.

> This is just FUD. AWS prices are always lowering, there is never a time they have increased. Even if they increased their prices, Google, Rackspace and Microsoft would eat their lunch. There is plenty of competition in the computing space.

The prices are lowered, but AWS is also continuously introducing new more powerful instance types and deprecates older ones.

> If your solutions are so narrowly defined that they won't run on anything else than AWS, then you are doing it wrong. You might run on specific AWS services, but there should be no reason why someone could not recreate their solutions over on Google Compute or other services.

AWS has many solutions that are different than you are used to. For example there is no NFS (well there is EFS which supposed to be similar, but it's still in preview for at least a year now). You're forced to improvise, and use their solutions. Typically S3, which is not exactly 1:1 alternative to NFS, so you'll need to improvise and rework your applications to that model.

If you want to move to something else, you no longer will have these services available.

> Yes, you might have to rework some of your solutions, but it should be very doable.

Exactly, that's what's pointed out. You'll have to rewrite your applications. There's no lock in that can't be solved by rewriting. The problem is that the rewrite might be quite expensive.

>My point is what you are describing for a small business person requires a tech person knowledgeable with AWS to implement

It may seem that way, but in actuality, it really doesn't. In the case I described, their "tech" guy was someone who'd been pushed into the role by necessity. He had the most knowledge out of a body of people who didn't really make it their area of expertise, and became the IT guy as a result. He did a well enough job working on his own. But there's things he's going to miss, and it's a lot to keep up with if you want to run an infrastructure right. As you pointed out, it's a lot different running Software-as-a-service solution like Webmail vs. Infrastructure or Platform as a service.

But that's the nice thing about AWS (and to be fair, Azure and Google do this quite well too) is that there's a solution for your tech level. If he were really knowledgeable and wanted to roll his own mail solution, he can, and stand up instances running sendmail and put in his own MX records into our DNS service and manage all of the bells and whistles. Or, he can use Workmail (or Gmail for business if you want to be agnostic) and point and click his way to getting mail set up.

So there's a lot of range there, and the higher you go up the stack, the more management you hand off to the cloud provider at the (hopefully small) expense of control. But if the provider's offering fulfils all of your requirements, then it's a great deal for small, non-technical, shops who can offload their workloads into a provider who are experts at managing infrastructure, and you, the business owner, can create the occasional account and otherwise focus on business.

There's a great chart on this you might have seen, but it looks something like this: http://1u88jj3r4db2x4txp44yqfj1.wpengine.netdna-cdn.com/wp-c...

In any case, there's a solution for everyone depending on their skillset, and arguably, (in my humble yet biased opinion =P ) AWS has the best spread. If I had one self-criticism about us, it's that it's not always intuitive as to which of our services fall in the SAAS/PAAS/IAAS stack. Education will always be a challenge no matter who the provider is, and being that cloud is still relatively young, it's up to AWS, Microsoft, Google, and everyone else to explain just what "cloud" means and how to start with it.

As for locked in, it’s AWS’ goal to keep you by ensuring you /want/ to use the service, not because you’re forced to be here. There’s no lock in other than transit fees. Which, if you have thousands of terabytes, can be hefty. That’s bandwidth for you. But there’s nothing proprietary. Take your code and data with you. I’ve seen engineers walk customers through the process of getting their stuff out. It happens. Not often, but it does, and we’ll help with that too.

> bring up a smtp and imap server and I can put it on the colo box or do it on AWS.

Is that all the server will be used for? If you just want to run a Linux server 24x7 you will not be taking advantage of any AWS services, so you would probably be better off getting a colo box or VPS to act as your mail server.

If you really want to run this on AWS, you just need to spin up an EC2 instance, and configure the Security Group (think of it like a firewall). There is a wizard which will guide you through the process[0]. Once the instance is running you can (optionally) use Route 53 to set up DNS records for your new instance.

Also, feel free to buy my book :-) It starts with the basics like launching instances, and then moves on to more interesting AWS-specific features [1].

0: http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/launching...

1: http://www.amazon.com/AWS-System-Administration-Practices-Sy...

With regard to my comment (below) I guess what I am looking for is a guide like "hello world" for AWS.

I think this is a great intro into the different components. https://www.airpair.com/aws/posts/building-a-scalable-web-ap...

If you have any other questions, feel free to reach out to our forums, or drop me a line. =)

It really depends on your needs... if I can have my postgres/mysql, or other table storage serviced by another company without having to become intimately familiar with another set of software to administer and/or pay someone to do so it may be a good option.

If you can save a FTE admin, you're saving at least $10K/month, that can cover the hosting and bandwidth for a pretty big site/software/service on AWS and the like.