The cloud wars of 2016
Amazon, Microsoft, Google, and everybody else
(If you like my writings, you might subscribe to “Fabrica”)
Short Summary, in case you don’t want to read it all (MNMN):
Most companies pretend to have built a “cloud”, but they haven’t. Cloud Computing, an automated, API-driven, on-demand and pay-as-you-go infrastructure, and on-demand software on top of it, are the true innovation.
Amazon Web Services (AWS), the market leader, needs to expand into the private IT through an appliance.
Microsoft is gaining share and pushing Azure through existing enterprise deals.
Google is waking up, gaining momentum in the public cloud space. They should improve their efforts to get customers and to create the right environment for developers to stick around.
Conclusions: Cloud Computing is changing, money can be made in Enterprise IT, large companies will suffer.
AWS will dominate. Microsoft and Google might be relevant if they stay focused.
[Discuss on Hacker News]
And now, the actual article.
Everybody has built a cloud
Everybody claims to have built a cloud these days. Even my dog.
There is a Cloud CRM.
A Cloud Database.
A Creative Cloud.
A Cloud ERP.
What the hell is a “Cloud” ERP (Enterprise Resource Planner), anyway?
It’s just an ERP that you can run on someone else’s server. I wouldn’t call this the “Cloud revolution”; you are just removing the need to run your buggy software on my servers, and you’re charging more for the “cloud” version that I can now run on yours.
Thanks for NOT solving my real problems.
There is so much smoke and so little substance in the “Cloud” space.
Then, there’s the now quasi-obsolete subdivision of Clouds into Software as a service (SaaS), Platform (PaaS) and Infrastructure (IaaS).
What the hell is a Cloud, anyway?
“Cloud” simply refers to something that sits “somewhere” on the Internet. Something that, no matter where it is, it can still offer you some kind of service or functionality.
Is this enough to call it a revolution? No.
Modern Cloud Computing, on the other hand, is much more interesting.
Ten years ago, if you wanted computing capacity to run your software, you had to go through a painful process that usually required several weeks.
That, and many other things were MANUAL.
Exactly ten years ago, Amazon Web Services (AWS) popularized the “on-demand” infrastructure, such as storage (S3) and virtual machines (EC2).
I still remember that day. I started using it right away, and two years later I joined AWS and spent six years there.
You could get capacity when you wanted, through a SOAP or REST API, and you only paid for what you used (“Pay as you go”), with your credit card.
Now, if you pair this with software and licenses, then things get more interesting.
If your “Cloud ERP” allowed me to scale my capacity on demand, and pay only for what I use, and security patches and updates would be automatically applied to my systems without breaking anything, Oracle would be out of business and a lot of customers would be happy.
Oracle doesn’t do that, because it would kill its main business, which still sells licenses year over year.
The same is true for the infrastructure powering that software.
It is a known fact that AWS, as an example, enjoys huge margins on the “on demand” part of the business. The TCO (Total Cost of Ownership) of servers and storage in a properly run Data Center are a small fraction of what you pay to AWS to get them “on demand”.
It’s like the food you get on planes: way more expensive, and way lower quality than what you could get at home.
But also way more convenient.
Unfortunately, there is still a lot of confusion about costs and benefits, and the reality is that THE RIGHT MIX OF THINGS (e.g. some AWS, some private infrastructure, etc.) is usually the best bet.
And in fact, companies like Dropbox do exactly that. And rumors suggest that even Apple might do that (update March 23rd: Apple is doing it).
The real Cloud Computing players
Despite “Cloud” still being so hyped, a few companies are REALLY changing the world.
The big players are Amazon Web Services, Microsoft Azure, and Google Compute Platform. You could call them 90–9–1, as these are approximately their market share at the moment (despite whatever claim —e.g. Microsoft used to include some “free” Azure in their ELA despite customers didn’t want to use it, thus inflating the Azure sales figures).
We can say that there is an actual war going on between them.
Why?
Because the lock-in effects of their solutions are so powerful that, once you have moved to one of their “Cloud”, it’s going to be very hard to move elsewhere.
You also have less bargaining power towards your sole “drug dealer”, which means you can dictate less discounts when trying to negotiate a better deal.
Amazon, and the Appliance
AWS is the 900-pound gorilla, the undisputed king of the Cloud Computing market… For now.
Andy Jassy (AWS’ GM) is a machine: I’ve worked with him and I know that he’s focused, relentless, detail-oriented, and he knows how to scale a business. You don’t expect much less from Jeff Bezos’ first Technical Assistant. (update March 19th, 2016: he’s been nominated Financial Times’ person of the year)
AWS is already dominating the “Public Cloud” market. So, what’s next?
Simple: the “Private” IT, which still represents more than 90% of total IT spending.
How? AWS needs to build an “appliance”, a device that lets you to buy your own “private” AWS cloud, so that your compliance can be met, your security can follow your strictest guidelines, and your occasional need for extra capacity can be met either by “bursting” into the Public cloud, or ordering a new set of servers and SSDs with an API call, delivered at your doorstep with Amazon drones, and ready to be instantly plugged into your AWS appliance and instantly turned on.
An additional neat idea would be to have an API to get rid of them as well, at a pre-fixed price based on standard depreciation.
AWS can’t just TRY, though. It has to nail it at the first round, or it would lose trust and momentum.
AWS needs to find the right “insertion point” into old and clunky Enterprise IT, without pissing off Oracle and SAP too much.
What is it? Is it Monitoring? Is it Resource Management? Or is it a more long term battle, focusing just on containers and riding their future adoption?
Would it be through acquisitions (Sensu, SysDig; SaltStack, Splunk — if you want to get acquired, just name your company S-something)?
Not likely. They need to build something unique, therefore they would target talent, not products, unless the product aligns with the strategy and the team is good enough.
Microsoft, “we do everything and then some”
Microsoft Azure is trying to gain momentum, and the recent open source initiatives (one, two, three, four, and many others) show that Microsoft is actively using its huge installed base and deep penetration in the enterprise world to fight Amazon’s dominance, and gain developers’ mindshare.
It is also using an old tactic: feeding Azure into ELAs (Enterprise License Agreements) with existing customers, with the hope that they will increase adoption and then get locked into the new Azure platform.
It’s a strategy, and it’s doing well. Still to be seen if it will be enough.
However, Microsoft has a shorter path to an “Azure Appliance”, because they don’t need to teach anybody how to use Windows or to write software for it.
Problem is, they don’t know how to build an Appliance, and nobody would trust them right now.
My guess is that Microsoft will simply try to beef up their offering, and sell the “hybrid cloud” story to as many enterprise companies as possible.
Google is finally waking up
Google? Well, Google is becoming interesting.
In December 2015 it acquired Diane Greene / BeBop, and put her at the helm of Google Compute Platform (GCP). I don’t know Diane personally, but I know that she “gets” the enterprise. Google doesn’t. That’s why Google needed her. She might not be up to date with the latest cloud trends, but she has Brian Stevens. She’ll be fine.
Early 2016 signs indicate that GCP (can you drop the acronym and find a cool name? Microsoft Azure; Google… ?) is now focusing on winning “logos” that REALLY use their cloud (sorry AWS). “Fake” logos don’t go too far; I would name another company that did exactly this, but I won’t.
These are the reasons why this is a winning approach:
1) Product teams will interact with REAL customers and see REAL usage of their service. They will better understand what features to implement next, what to fix, what to improve.
2) The executive team will understand which services are useless, or simply don’t contribute much to the bottom line.
3) Future customers will bet on GCP because Spotify and others did.
I expect Google to announce several other huge wins over the next few months… And I’m sure that they are DEEPLY discounting GCP in order to achieve that. (my bet is 80% to 90%, without considering “free” assistance in setting it up and helping them get onboard).
However, Google still doesn’t understand (despite good folks like Greg Wilson) that in order to get widespread adoption, lower-than-AWS costs (and a GREAT infrastructure) isn’t enough.
They should:
1) Offer MUCH better documentation and examples (CoreOS does it superbly);
2) Team up with companies that offer Cloud Computing training (e.g. CloudAcademy — disclaimer: I’m an investor), in order to make it easier for people to learn how to do stuff on… GCP (you haven’t changed your name yet?) and increase adoption;
3) Offer an amazing CLI and amazing APIs — this is simply necessary;
4) Hang out on Hacker News more (when you do, I love it and so do many other developers — just don’t get your PR in there, please).
How do you do #1, #2, #3 and #4? You “throw people at it”, just make sure they are great people who speak “geek” language and who don’t get lost in the interview process. A “friend of mine” (read: you might guess who this is) refused TWO written job offers at Google whose process was a complete screw up.
This one is the “worstest” ever.
Just fix your interview process and hire great people. You will need them.
Other than that, Google is still just competing in the Public Cloud space. It’s a crammed space, and Kubernetes isn’t enough to get people onboard GCP (yes, that’s what it’s primarily sponsored for).
They will need a long term strategy. Diane Greene might be able to offer that. I have some ideas, but it’s a long discussion and I’m not ready to share all these ideas just yet (for what they are worth).
Yes, that strategy will at some point include an appliance, but it’s also very early for Google to think about that.
Conclusions
The Cloud Computing landscape is rapidly changing, there is a lot of money to be made, especially in Enterprise IT, and a few large companies to be destroyed in the process.
AWS will clearly dominate the next few years. There’s space for one or two other protagonists, and both Microsoft and Google have a shot at it.
Focus on what matters is what will count.
Well, what do you think? Feel free to discuss on Hacker News too.
[disclaimer: I have worked for Amazon Web Services from 2008 to 2014, and for VMware from 2014 to 2016; I currently work as CTO for a stealth startup in the Bay Area. The opinions expressed above are mine, and they do not represent my past or current employer’s views, nor do I share sensitive and/or confidential information that I have obtained during my past service at these companies.]
(If you like my writings, you might subscribe to “Fabrica”)
