You'd be surprised at what a good job the individual pieces of Amazon's empire do at operating as individual units as far as competition is concerned. It would be VERY bad business to "screw" an AWS customer because they competed with some other part of the organization as a whole. Given that Amazon operates in so many spaces and could potentially expand to any space, this behavior would effectively make AWS unappealing for a significant portion of customers.
I'd like to believe this, but then I can't buy an Apple TV through Amazon because of some petty dispute over Prime Video (which I really don't care about). So if Amazon the store can pull products to try to protect prime video, then how can I be sure that other units won't?
Well, they claimed they removed non-Amazon devices because they didn't support Prime Video. However, for stuff like Chromecast, the burden lies on the app developer (which would be Amazon) to implement Google's Cast SDKs to provide support.
Because of this, Amazon cannot distribute Google Play Services on their own Android devices and therefor cannot use their own Android devices to cast to a Chromecast. They could enable Prime Video on other Android devices but not their own. They feel this will cause confusion and therefor have decided not to write a Prime Video app for Chromecast. Saying "oh, Amazon just doesn't want to write a Prime Video app for Chromecast", while true, doesn't truly explain the complexities of the situation.
The business decision to remove Chromecasts from the network steps from the above desire to prevent confusion. I personally don't agree with it but I understand it.
That's not the problem. The problem is if Netflix is paying Amazon, then their cost is infrastructure + profit, whereas Amazon only has infrastructure costs.
Fundamentally it means Amazon can be more competitive, and Netflix are hoping they can stay ahead on the user experience curve.
This might be true, but only true until a change in management. Given everything I've seen and read about Netflix, I suspect they have contingency plans and should be able to move to a different infrastructure provider in a few months.
Granted this is true in the case of Amazon, but some conglomerates do not act this way.
Telecom companies regularly try to make rules to ensure those that compete with cash cow voice or cable-TV services have to do so on unfavourable grounds.
Considering the bills I've seen for AWS at even modestly-sized companies, I would not be surprised if Netflix spends more than $80M a year on AWS. In fact, considering the amount of bandwidth they use, I'd be surprised if they didn't.
They spent $1.3B on "General/Administrative Expenses" last year - think more than 6% of that was infrastructure?
Quite possibly, but also keep in mind that there's no way Netflix pays anywhere remotely near AWS published prices. I'd be surprised if they pay even 1/5th of the published prices.
Especially on bandwidth, anyone can buy bandwidth at well below 20% of AWS prices elsewhere. Netflix shouldn't be paying even 10% of published AWS bandwidth prices given their volumes.
Most of Netflix's bandwidth goes through their Open Connect CDN, which isn't hosted on Amazon's infrastructure but is rather a traditional setup of colocated servers in strategically positioned data centers.
Previous to this arrangement Netflix used 3rd party CDNs like Akamai which again did not sit on Amazon's infrastructure.
That said I'm not agreeing or disagreeing with any assessment of how much Amazon costs are.
Amazon Prime reminds me of how the telecommunications industry sell services. They have customers signed up to a bundle of services like: landline, mobile, cable/satellite TV and broadband.
If the customers don't want TV they are put on a free tier rather than being cancelled, so they are still officially a TV subscriber and can boost the viewing figures for the service.
It's useful to be vague if you are in a dying industry or your competitors are better established.
I think they meant Netflix pays Amazon more per year for infrastructure than Amazon makes on Prime instant. It's hard to say how much Prime streaming makes per year since it's also part of the 2 day shipping.
I guess it depends how much vendor lock-in they have with AWS. If they are just using their own virtual machine images that can run on any cloud infrastructure, then they could quickly re-locate to Rackspace or Azure if they needed to.
On the other hand, if they rely on lots of proprietary AWS tools, then that seems like a dangerous place to be.
I would also feel very uneasy, but I think it is because of my national origin. I truly find these things of American capitalism very admirable. I am sure it will work for them without a problem. Just look at Samsung and Apple: killing each other in court over patents, without affecting their trade of electronic components.
Do you have the same reaction to Apple buying processors from Samsung or Intel? Google buying operating systems from Microsoft or Apple? In each of those situations they've decided that level of risk is outweighed by the cost of trying to do everything in-house. Data centers are an expensive business to be in and Amazon can amortize their costs over a much larger base than Netflix.
The thing to remember is that businesses routinely do this kind of thing and rely on contracts and markets to manage their risks. Even assuming that they don't have guarantees in the agreement giving Netflix those famously low rates, if Amazon chose sabotage they'd be endangering their much larger cloud business and that would hurt them for much longer than the time it would take Netflix to migrate to Google, Azure, etc. Even if you're completely unethical and a ruthless competitor, it's hard to argue that Amazon wouldn't benefit far more from making Prime Video a more interesting product, which conveniently also has no potential downsides.
Well, I suppose I don't have the same reaction to those other examples mainly because Amazon/Bezos have shown such a clear desire to capture/dominate market verticals and thus I'd never want to put myself at his mercy in a business situation. I also don't think those examples are perfect analogies for the Netflix-AWS dependency, though I certainly understand your point that businesses routinely rely upon the services of other businesses to get things done and that sabotage would be a bad look for Amazon.
I'm not arguing that Amazon wouldn't benefit from making Prime Video a more interesting product. But finding ways to hamper your biggest competitor, who is fully dependent on your infrastructure, is definitely a way to make Prime Video more interesting. Perhaps I'm being paranoid. I guess I just find Bezos difficult to trust.
A decade or so ago, most people would have said similar things about Microsoft, but I get your point in general and certainly would only trust Bezos to be a tenacious competitor.
I think the big counterbalance now is competition – raising prices on AWS would be a boon for e.g. Azure – but I definitely wouldn't want to bet against abuse if one of Microsoft, Google, Rackspace, etc. dropped out. I'm sure Netflix has pretty good forecasts for how much it'd cost to move operations elsewhere as long as a good option exists.
There is a risk. It needs to be weighed against the benefit.
Assume that Netflix has done the analysis and prefers AWS over other provides. The benefit can be framed like this - if Netflix does not use AWS, then Amazon Video has a strategic advantage in being the only one with access to the best platform. By using AWS, they remove that competitive advantage from Amazon Video.
It's super risky. Amazon could decide to slow things down for Netflix, or introduce some infrastructure issues that they can't track down.
Maybe amazon should be broken up. Break out their online video offerings into a separate company. This would prevent any conflict of interest or blackmail
AWS and Heroku aren't really competitors. Firstly, since AWS is IaaS, whereas Heroku is PaaS, they target different users. More importantly, Heroku is built on top of AWS, so Amazon gets paid for all the resources used by Heroku apps.
Sort of. They're still targeted at different users, though.
Heroku is for individual developers to directly develop their app in terms of.
AWS, meanwhile, is mostly for enterprises: it's a "virtual datacenter" for your (much-reduced) ops team to manage in place of a real datacenter. In that sense, Elastic Beanstalk is not a service that AWS provides developers—it's a service that AWS provides your ops team to provide to in turn provide to your developers.
Wasn't familiar with OpenConnect. It is the program where Netflix puts a caching server in an ISP's datacenter so that the traffic is more "local" and cheaper for all involved.
OpenConnect is the CDN that serves the actual Netflix video, it's Based on FreeBSD 10 and Nginx. Netflix puts them in ISP's and internet peering locations across the internet.
A lot of folks responded to this by talking about how Netflix asks ISPs to put caches in the ISP locations.
That is true but it's not the point.
Netflix runs a large CDN on their own infrastructure, and rents space in various datacenters.
They "peer" from these locations to distribute their traffic, and they also buy transit to distribute their traffic. They also send traffic to ISP caches from these locations.
So they are running their own data centers, in many locations.
Is Netflix at the top end of some curve? My understanding of moving from cloud -> datacenter is, you know, there is some line where it's cheaper to just lease your hardware than pay for cloud infrastructure. At that size you tend to have DBAs, dedicated infrastructure people, etc., to make scaling up go.
But now has Netflix come through that line where the cost savings is outweighed by the uptime/availability etc.?
It's important to keep this in mind: For most companies, their infrastructure is the second largest expense after their employees. For Netflix, the cost of both pales in comparison to the content licensing (which in 2016 they have said is 6 billion). Cost isn't a huge concern, agility is far more important.
That being said, the all in cost for Netflix on AWS is still cheaper than trying to run nine datacenters around the world, which is what they would have to do to replace their usage of AWS.
Often the calculations of managing your own hardware is pretty naive. Folks are correct that they can hire a couple of infrastructure folks to build out a single datacentre, but often there datacentres aren't as big nor as resilient as what they'd get from AWS, at least not at the scale where the luck factor starts to fall to statistics.
Much, much less. If you spend more than a couple of millions a year on AWS you're being robbed if you haven't negotiated a much steeper discount than 50%. I happen to know of companies paying ~25% of published AWS prices. Especially on bandwidth, the published AWS prices are almost criminal - I'd be surprised if Netflix pays even 10% of published prices.
Netflix is in the video distribution business. Their usage is hugely variable over the course of a day. AWS is in the cloud computing business. They are good at selling unused server time. If netflix ran their own datacenters, they would have to pay for infrastructure to sit unused for most of the day to meet their peak demand. With AWS someone else is using and paying for that infrastructure in off hours. The cost-savings for netflix and synergy for AWS make the deal make sense.
You have that backwards. Everything runs in AWS except serving raw video bits from very stripped down boxes in data centers around the world. Those boxes in the CDN do literally nothing but serve video files via https very quickly.
I'm 100% sure it is https (I worked there). But if you don't believe me you can sniff your network traffic the next time you play a movie on Netflix. :)
I thought everything but the video streaming was done on AWS. Like the Website, the recommendation engine, video encoding, etc, So I think thats a lot more then business rules. The FreeBSD CDN just serves the video.[1]
Interesting bits about the microsrvices approach. Other companies like Microsoft still need to wake up and understand the value of small teams, microservices deployed in isolation, real devops etc.
I heard about Orleans but how much is it in use internally at MS? Lot of services are not even on Azure, yet? Ever? It's well known that new fancy modern approaches take years to be understood and absorbed int the company strategy, think Git, Agile, elastic services, REST, mobile... Don't get me wrong I'd love to see MS going back to be an innovator, and Orleans might be an oppotunity. How long has it been cooking, 2 3 years? Where's the disruptive plan to make it the new paradigm?
What did it cost to transfer Netflix to AWS and what are the running costs charged to Netflix. I recall reading somewhere that the first attempt was less than successful, and Amazon had to virtually rebuild the service from scratch.
