At the rate they're innovating and growing, I wouldn't be surprised to see the Amazon Web Services operation grow to be larger than the rest of Amazon's business.
Jeff Bezos spoke about this in some depth at startup school in 2008. He even highlighted a journalist's quote suggesting that Amazon was sneaking web services through the back door and it would soon be larger than Amazon itself. I don't know the numbers offhand, but I wouldn't be surprised if they've already surpassed the storefront side of their business.
I believe estimates are that AWS revenues are somewhere in the lower hundreds of millions. Very far from rivaling their tens of billions retail business.
What about the profits? Shipping, stocking and moving around physical books has got to be more expensive than marking up racks of computers and bandwidth.
Real costs associated with building out a datacenter are significant. I'd be surprised if building and operating a warehouse comes anywhere near the cost of building and operating a datacenter.
You've got things like generators, HVAC, PDUs, electrical engineering, network equipment, non-H2O fire suppression systems (not to mention the cost of the racks, systems and power itself) that make a datacenter a pretty hefty investment.
Even if you amortize the cost, it's amazingly expensive.
They're probably following the same strategy they used in retail: undercut everyone and go for long-term market share domination.
I'd bet they're still significantly in the red every year. Eventually they'll get big enough to start skimming a few points of profit off the top. Margins will always be low because that's a huge part of their strategy.
I keep hearing this but when you compare the same machine you get on Slicehost or linode with the Amazon prices, Amazon ends up being the most expensive.
I use Amazon because of it's convenience and brand name, but I've always seen it as having a decent premium.
Amazon.com is also very often not the absolute cheapest option for products when compared purely on price. It's only if you calculate in reliability/shipping/customer support/convenience that they're the completely unbeatable on average.
Anyone can buy $100k worth of servers and $10k/mo in bandwidth and easily beat AWS pricing. That doesn't make it competitive with AWS though.
AWS is providing world-class network/server infrastructure as a commodity, which is decidedly different from the kinds of infrastructure most hosting companies have.
Doesn't this contradict your previous comment? How exactly do they "undercut everyone and go for long-term market share domination"?
I actually think AWS is profitable and while they might not have recuperated their initial datacenter investments, they do cover the accounting depreciation (which is all that matters).
They undercut everyone offering the same product. Rock-bottom hosting that might stay up and might have customer service available sometimes is not the same product.
Maybe I have a hard time understanding this, but the whole discussion was how Amazon undercuts their competitors and runs the business with a small loss because they are looking at the long-term game.
Saying that Amazon sells you customer service seems entirely unrelated. Yes, the might have better customer service and they might have more things (EBS, Route 53, etc) but they certainly do not undercut their competitors price-wise.
If you run out of storage on Slicehost, though, you have to get a bigger machine, with more RAM and bandwidth. Run out on EC2 and you just add a new EBS store. There are major advantages to the completeness of the group of tools AWS includes.
Has anyone done credit card handling (input, submission to payment gateway, and storing for subscription billing and on-file orders) on EC2?
A while back I recall Amazon saying that this was possible. We're looking into the possibility of moving to the cloud, and on first look our PCI guy saw some problems. We've just started experimenting so could easily have overlooked something, but these were the stumbling blocks we saw. It looks like these new features address 2 of these 3:
• PCI requires limitations be based on outbound traffic from the cardholder environment. Amazon only allowed inbound filtering. Now they have outbound filtering, so this may be no longer problematic.
• PCI requires internal machines to be placed on internal private networks using NAT. Amazon did not support NAT. Now they do, so this block may be gone.
• PCI requires that all traffic be monitored with an IDS in the cardholder data environment. It doesn't appear possible to do a central monitoring machine with IDS in EC2.
Yeah, we've seen that. That's what I was thinking of when I said Amazon said it was possible.
However, when it comes to actually doing that things seem a bit less clear. For instance, outbound filtering is a requirement, but Amazon just added that, so how did one satisfy that requirement before today?
That could be what they had in mind. Our PCI guy tends to frown on that, though, because the purpose of restricting outbound access is so that if the machine does get compromised the bad guy can't ship off credit card information. If the outbound restriction is implemented on the machine, the bad guy might simply be able to turn them off.
It's possible that we've got an overly strict PCI guy.
Anyway, this particular issue appears dead now, as these new EC2 features add outbound filtering.
I don't think your PCI guy is overly strict. It's pretty clear that the intentions of the requirements are what you described. What might have worked, though, is to have virtual machines inside the EC2 instances in your VPC, and use this to filter traffic through a separate virtual machine.
Still, it's unnecessarily complicated and as you say, a resolved issue now. :) The new features announced fits PCI needs quite well. I haven't looked into the IDS issue you mentioned in your first post yet, but I hope it's possible to resolve somehow or get around with compensating controls.
(Disclaimer: I'm no PCI DSS expert, just an unlucky engineer trying to make a compliant system.)
Completely agree. The AWS team is one of the very few examples of rapid iteration and improvement from a big company.
I'm as interested in the AWS team as I am in any startup that exists today. I'd love to read about the tech challenges/team make up, etc. Is there any good coverage of this?
Yes, would definitely be interesting to find out how they maintain the quality and release so often. Would be a good data point to see if they use any of the agile processes and any tweaks they have done to make it work for them.
If Bezos's personality is decidedly noncorporate, so are some of his ideas about how to run a large organization. One of Bezos's more memorable behind-the-scenes moments came during an off-site retreat, says Risher. "People were saying that groups needed to communicate more. Jeff got up and said, 'No, communication is terrible!' " The pronouncement shocked his managers. But Bezos pursued his idea of a decentralized, disentangled company where small groups can innovate and test their visions independently of everyone else. He came up with the notion of the "two-pizza team": If you can't feed a team with two pizzas, it's too large. That limits a task force to five to seven people, depending on their appetites.
I first tried Rackspace Cloud, who would send me frequent marketing-style e-mails with stock photos of intelligent-looking office employees, ask me to participate in raffles, and other nonsense that I would quickly filter.
I much prefer seeing new feature announcements from AWS in my inbox! (And on HN.)
I use AWS both for work and for personal (love the free tier micro to play around with on my own time) and I am always impressed with how polished new features feel.
I was hoping that this would finally be a way to have an ELB inside a firewall, but alas VPC doesn't support ELBs yet.
AWS Elastic Beanstalk, Elastic Load Balancing, Amazon
Elastic MapReduce, Amazon Relational Database Service
(Amazon RDS) are not available for use in a VPC at
this time.
