Posted by Sacha in English, IT.
Last week, Amazon made an announcement which didn’t get due credit: they’ve released their own version of Linux. And guess what: it could very well become one of the top two Linux “distributions”.
Much like ORCL did 3 years ago, Amazon took the most widely deployed enterprise Linux clone, CentOS, and customized it for its AWS usage. What does it really brings to them?
From an API/package standpoint, leveraging the CentOS distribution brings them binary compatibility with RHEL, the leading Linux distribution which supports the largest number of ISVs out there. While AMZN’s Linux distribution sounds like a purely technical exercise today, the day they decide to build a proper ISV program, the work that existing RHEL ISVs would have to perform to support that “new” environment would be marginal (unlike, say, if AWS had decided to use Debian or Ubuntu).
From a “back-end” standpoint, AWS has a very specific hardware base to support, this is the dream of any OS distributor and which only the like of Mac OX, Sun, HP UX, etc. could partially enjoy: you don’t need to support thousands of different devices of all kind and their respective incompatibilities: you target a very small set of well defined hardware configurations. This allows you to focus on key drivers and hyper-optimize your OS. No other vendor could ever reach that level of optimization since Amazon is probably the only one to know what hardware runs in their datacenters.
The former point also has an interesting consequence: the cost structure required to maintain such “distribution” is much reduced: they leverage all work and QA that RHT does on the core packages, and only focus on maintaining a well defined set of drivers/hypervisors. As you can imagine, for RHT the situation is radically different: supporting all those random hardware devices, comes at a significant cost.
The net result of this is that AMZN is in a position to deliver a more power-, performance- and cost-efficient Linux distribution than anybody ever could and ISV would have no technical trouble supporting it.
My only criticism? They do need to find a proper brand name for their OS. “Amazon Linux AMI” won’t fly if you want people to remember and name your product…
What does it mean?
Unless you are in denial about what impact the cloud will have on IT, you probably realized that significant computing load will move to the cloud in the next 5 years. Given the current market forces, chances are high that AWS will represent a sizable share of that market. Consequently, AWS Linux could become a key Linux “distribution” in the next 5 years (they don’t “distribute” it per se – they “run” it, the GPL could face interesting use-case loopholes here).
You could even imagine have AWS license their Linux power-pack to other regional IaaS (telco operators with a hosting business to retro-fit would be a prime target I’d imagine).
If such scenario indeed happens, RHEL could end-up becoming the new HP-UX, only surviving in aging on-premise environments (and on “private cloud” comets). In the mid-term, RHT could decide to license a lower-cost cloud-specific RHEL version – but that would only slow their revenues and margin fall and looks more like a IaaS-specific consulting business than anything truly subscription-based. In the longer term, the real question for RHT is whether they want to become a cloud provider themselves, but RHT has a long history of not being ready to trade a strategic bet against the risk of upsetting one of its “partners” (such as good buddies ORCL and AMZN).
As for an acquisition of RHT, this won’t happen at the current price level IMO except if a bidding war takes place (see next paragraph, “Enter Star Wars”, on how this could be triggered at current price levels).
For ORCL and AMZN, as long as RHT can do all of the work and produce RHEL (hence CentOS) for free, the deal is probably good enough for them.
Enter Star Wars: The VMW+SUSE scenario
(Note: it is Sunday the 19th of September 2010 as I write this, cruising altitude of 42’000 feet).
I’ve seen the tweets of a potential acquisition of the NOVL business by VMW. NOVL’s current market cap is very low. Furthermore, if as implied by the rumor, VMW has been able to negotiate to split the deal with a private equity fund and only pay for the SUSE business, this becomes a bargain.
For argument’s sake, let’s say this SUSE asset acquisition indeed happens; as we well see, this would further accelerate the plan described above.
VMW would probably see multiple advantages to this acquisition.
First, they would get to solve a long standing issue: in order to become a true “full-stack” enterprise player, they do need to onboard lots of ISVs, fast. Problem: to date, they haven’t been able to do so successfully. The only way they would have to accelerate this process at warp speed would be to “buy” those ISVs them by acquiring SUSE. While SUSE supports less ISVs than RHEL, their catalogue is significantly broader than VMW’s and I have no doubt an acquisition by VMW would rejuvenate the interest of this distribution by ISVs.
The second reason is that VMW currently present the same weaknesses as RHT do – their predators, once again, are the mega-IaaS, simply because they do not need them. The difference here is that I suspect VMW would have the cojones to become a cloud provider of its own. In doing so they’d probably upset tons of existing partners but that might be the price they have to pay to remain relevant long-term. Don’t forget VMW has strong MSFT DNA in-house and just look at what MSFT is doing. Bottom line, by acquiring SUSE, VMW would get its own cloud-OS, much like Amazon just did. Furthermore, unlike AWS, VMW wouldn’t not need to worry about whether they’ll need to acquire their Linux provider one day. If NOVL was to disappear, this could trigger a bidding war for RHT – no company currently largely dependent on RHEL or CentOS (IBM, ORCL, AMZN, etc.) would want to stay with a big strategic hole in their stack…
Last but not least, I am pretty certain VMW would love to get a few of those delicious NOVL patents in case MSFT get aggressive: firing a scud or two in their Office business would calm them down.
(To keep things simple, I’ll not cover here the challenges that VMW would inevitably face in running an Open Something project, despite their acquisition of SpringSource, FOSS is clearly not in their DNA).
5 to 10 years from now, I don’t see an on-premise Linux distribution/subscription to be a viable business: too limited of a market, too expensive to maintain. While this might be a long way to go, I do think that the (courageous) decisions required to escape from that black-hole must take place now.