Amazon Building a Tech Business for the Long Haul
bits.blogs.nytimes.com
bits.blogs.nytimes.com
Certainly, a good part of A.W.S. software is open source,
which means it could be used elsewhere, and companies
like Microsoft and Google also have big clouds. But
neither of them, nor IBM, which also wants the cloud
business, created more than 500 new features and services
last year, as A.W.S. did.
My guess is Google/MSFT is doing a lot of putting out fires since their offerings are newer than AWS, and they're not getting as much usage (which makes bugs appear earlier). That's been my understanding as to why Dropbox hasn't changed all that much -- they've just been making sure scaling works. But can teams of 10 at MSFT/GOOG really not put out new desirable product with any decent timeframe? They have enough staff members.
Now that Alphabet is the parent company, maybe this part of google can be taken more seriously. They are very capable of succeeding, but they don't really understand B2B/Enteprise and really never have.
Of course, I also think they could work wonders on their horrible attrition rate by ending blind allocation for middle to senior talent, so I'm clearly out of my mind.
After all, it works for their platform, what more could someone want?
One way Amazon adds so many features, though it won’t talk
much about this, is to contract key elements from other
companies. QuickSight, the data analysis tool announced on
Wednesday as a competitor to IBM’s Cognos, uses technology A.W.S.
hired from a start-up called Zoomdata, according to people
familiar with the technology. They asked not to be named in order
to maintain professional relationships.Why maintain and improve your existing services when you can get credit for launching a new shiny thing? Steve Yegge called out this tendency in his (in)famous platforms rant [1]:
They prioritize launching early over everything else, including retention
and engineering discipline and a bunch of other stuff that turns out to
matter in the long run. So even though it's given them some competitive
advantages in the marketplace, it's created enough other problems to make it
something less than a slam-dunk.
For example: http://remcobron.com/cloud-server-review-and-comparison-amaz...
Yegge sez: "a platform-less product will always be replaced by an equivalent platform-ized product."
I guess that's true. If I had to defend that statement, I'd try to argue that a platform, especially an open platform, will be tested, poked, prodded, evaluated, evolved and maybe even documented by many more developers than some monolithic piece of software in some closed code base.
But what other arguments are there?
I'm open to be corrected on this but I don't think it's good to confuse "our different offerings have lots of interoperable synergy" with "it's fucking impossible to migrate off this"
I can attest to this:
ssh user@rsync.net s3cmd get s3://rsync/mscdex.exe
Instagram's post about moving off AWS is enlightening here: http://instagram-engineering.tumblr.com/post/89992572022/mig...
The problem isn't as bad as, say, writing your app in Objective-C or CUDA only to later find you don't want to ship to iOS or NVIDIA cards.
The big question is how cheap will computing get, and how will the demand of that computing increase over time? If prices continue to decrease rapidly, it becomes very difficult to maintain large revenue growth, even with unit-volume increases.
In 20 years, how cheap will core services like S3 and ec2 be, and how will AWS' revenue on those services compare to today?
They are also developing a lot of AWS software that probably won't decrease in price over time. For example, over the past 30 years database hardware has decreased dramatically in price, but Oracle software licenses have not. AWS has a lot of services that are some combination of complex software and hardware components.
If you just run a simple web-application sure it's much much cheaper today than it was in 2000, but as technology evolved that isn't enough for big business today, if you look at companies that still build their own servers then unless they have unlimited space the servers these days tend to be much more expensive and exotic hardware is more and more common (it's not that uncommon to see SQL/BI servers with quite expensive compute/GPU's these days, 5 years ago it was all those "IOPS accelerators" or "specialized" SSD's, in 5 years it will be something else).
As far as looking at the price of S3 or EC2 in 20 years that's completely irrelevant because those services for the most part will be irrelevant in 20 years, yes their might have their own counterparts but it's not like you will still use the same services, just look at how drastically the way you build and deploy applications has changed in 10 years.
It is really strange to think about how the average consumer views Amazon as an online retailer, but that is really only half the picture. As the end of the article states:
>Maybe someday A.W.S. will be the company that also happens to own an online retailer.
I think though that Amazon will probably remain more of a conglomerate, rather than divesting any parts of its business. Instead, it seems to add more and more brands, services, etc. In the long this will probably be good for them, since they'll be able to effectively integrate the experience of customers across all brands/services/etc.
So not sure there's a mismatch problem. Also i wonder - how can Amazon tie all those B2B units into a larger whole ?
Jeff Bezos sees Amazon as the GE for the next 100 years. He has a long view that left American corporate culture in the late 1970s. Japanese corporate culture takes the long view, they have business plans that look out across decades, rather than quarters. Bezos is making amazon a very wide company, because it's harder to fail as a wide company than it is a deep company. Now, in some ways Amazon IS a deep company, with Amazon products delivered through their supply chain, marketed by their website running on their platform, and a rich media ecosystem on Amazon consumer devices. However it will survive damage to one of those sectors because of the breadth of the company.
The idea of cor businesses and splitting up companies that do "to much" is rooted in the quarterly corporate culture, which sacrifices the future of the company in exchange fore shareholder (and executive) rewards today. Business stagnant? Split it up into aligned companies for a quick stock boost. Did you actually create anything new or useful, or change the way the businesses operate? No, but it looks good on paper.
The question is size is far from decided, though. Though these Japanese firms may survive long-term, they often aren't good stewards of shareholder capital as measured by return on equity (ROE) and other financial metrics. But it does seem that conglomerates are making a comeback, with Alphabet, Amazon, and the perennial favorite Berkshire, whose own CEO has said their size is starting to present problems. Definitely an interesting question.
1. Its really expensive: about 5 times as expensive as dedicated servers.
2. Not reliable: they will shoot down servers any time they like. You will need to make sure you have redundancies in place and pay for them.
3. Lock-in: their services (beyond EC2) come with a set of proprietary APIs that your app will depend on. This creates a lock in.
4. Still need DevOps: so you pay a lot, get locked-in and still need to put everything together by yourself and make sure it runs reliably? Yep, you still need devops/admins which at the end is probably the biggest cost of your operation.
So, here are my suggestions:
You have a low-volume and don't mind a lock in: use a dedicated mobile/web cloud ala Firebase/Parse/etc or go with a true PaaS solution like AppEngine/Heroku.
If you have a big volume: you will need to bite the bullet, hire some devops and setup+maintain your own dedicated servers. You will at the end pay much less for servers compared to AWS, have full control over hem and have no lock-in.