This is a really amazing and clever idea on Amazon's part: Taking a system or technology that you've already built and are maintaining, and make it accessible to others as a platform. Then continue using that platform as if you were a/the best customer.
They did it with AWS, they've done it with their retail business, and now they're probably going to do it with AmazonFresh/WholeFoods. It makes a lot of sense, and it offers them a lot more flexibility to use the platforms they build in other ways too.
This is a persistent myth that needs to die, AWS was not built to host their websites, and indeed did not do so for many years after it was established.
AWS wasnt their "spare capacity" nor was it them exposing their internal systems to the world so they could sell them off, It was developed from day one as a public facing service.
I don't think what you are saying contradicts the gp. All the expertise that went into developing tools to host their own website probably went into developing AWS. Just because they maintained two separate infrastructures doesn't mean that it's not an example of turning an internal tool into a product.
Your misunderstanding that comment. It was built "in the same datacenters by the same IT engineering team" but it was always it's own thing. "AWS services were quickly adopted by many teams" after it was built and a commercial product.
Sure, the intent was Amazon IT would eventually move to that system. But, they could let other companies work out the issues first vs. risking their main website. Just consider the risk for them if AWS goes down and take their own site outline while they are trying to sell high up time. That's simply a risk they had no reason to take.
PS: Remember, well executed traditional large scale infrastructure has higher up-time than AWS because AWS adds complexity.
AWS was built to host Amazon. Sure, it was a public facing service first, but AFAICT Amazon.com was always the primary target customer, and probably still is. AWS was built based on the needs and experience for hosting Amazon.com.
The experience of hosting and scaling Amazon may have informed some early AWS thinking but AWS was not built specifically to host Amazon and Amazon was never the primary target customer. This is evidenced by nothing more so than by the fact that despite Amazon's prescriptive initiative to move to AWS being well over 5 years in[1], much of retail & digital still runs on Oracle, dedicated HW, and bespoke internal services.
To make the statement today that Amazon "probably still is" the primary target customer betrays a lack of understanding of the scope of AWS and IaaS in general.
[1] It might be even 8+ years. I don't recall when it officially started so using 5+ as a very conservative estimate.
Eli Goldratt's The Goal popularized two main ideas. Theory of constraints and transition from production to services (subscriptions, or both).
Many, many have tried to transition to services. Amazon figured it out.
I was totally invested in Sun's grid computing vision, so I was very slow to recognize AWS. We already had VMs on servers and I couldn't wrap my head around virtualizing everything. It just seemed like so much monkey motion (too much work).
Even so, AWS is slowly, painfully recreating grid computing. By bundling up those services. But unlike Sun, AWS found an incremental path, with many, many ways to monetize along the way.
IMO, the biggest/best companies are the one's that take run of the mill daily business expenses and turn them into money makers.
Google needed email to run it's business. They built a customer facing service around it.
Computing answers is their thing. With their machine learning cloud services, anyone can compute answers for their complex problems.
Companies that stagnate tend to iterate, with little to show for it, on a few "core" products.
While they're too big to collapse altogether, I'd put Oracle, MS, Yahoo, and numerous others in that group.
Possibly Facebook. They're no where near as diversified as Amazon or Google. If the bottom falls out on buying eyeballs, they're in for a world of hurt.
I'd wager they are much less vulnerable to security breaches because of that. It forces them to treat all their tenants as potentially hostile whereas otherwise they might fall in to the usual trap of thinking that behind the corporate firewall life is good and everybody can be trusted.
I don't think that's the right way to approach this.
AWS is like any other hosting provider that also happens to have one really large customer, Amazon the e-commerce company.
Now, a nuclear powerplant is not going to co-locate with a regular hosting provider because they do not need to move any of that data off-site and co-locating others on their network would mean co-locating those others on the far side of their firewall, or in some kind of DMZ. Nuclear power plants tend to concentrate on what is their main reason to exist: to produce power, not to get into the hosting business so for them it would make absolutely no sense to allow others on their networks, besides any security risks that might cause. You could even argue that the only safe way to connect a nuclear plant to the internet is 'not'. Airgap the thing and call it a day, don't even try to secure it. For Amazon this is - for obvious reasons - not an option.
So for Amazon the company the decision to create a web hosting service which then has their e-commerce arm as one of the tenants (each behind their own (virtual) firewall) is no different at all than Amazon the company buying their hosting from any other provider (Microsoft, Google, Rackspace and so on). They have to host somewhere, no matter what. In which case they have all of the downsides of the current situation and none of the upsides. The impression that you are on Amazon's internal network when you are co-locating with AWS is where things no longer align with how things work from what I know about their infrastructure, AWS allows the creation of VLANs for segregation purposes which is roughly the level of separation that you would expect of any multi-tenant network.
So in that sense Amazon is not less secure than any other corporate setup, the only way in which that sort of situation could lead to trouble is if you managed to find an exploit in the network stack (for instance: the firmware of a switch or router) that Amazon operates that would expose traffic from one tenant to another.
But I'm not aware that any such breach has ever happened, in theory it could happen but that goes for any other hosting provider as well.
What level of compartmentalization would you be comfortable with?
He's basically saying that Amazon is a big logistics company. Except he calls it meat as a service.
It's an interesting take. But it is also how many analysts thought that Walmart would kill grocery stores. I personally don't think that Amazon will fare much better than Walmart... shopping for groceries is fundamentally harder online, and Amazon hasn't demonstrated any magic that makes that easier.
