Posted by Sacha in CloudBees, English, IT.
The cloud is redefining the IT industry as we’ve known it. Amazon’s recent announcement to provide free Windows micro-instances on EC2 is another event that underscores this revolution occurring from the move to the cloud and how even a giant like Microsoft has probably been forced in such a move.
However, I see several roadblocks to truly leveraging “free” on EC2 with the Windows platform. Let’s explore this further… For as long as IaaS providers have existed, using a Windows machine has always been more expensive than using a Linux machine. As an illustration, you can see below the current prices for Amazon Web Services and Rackspace, two leading IaaS providers.
|Amazon EC2 pricing
Using Microsoft Windows is about 33% to 40% more expensive than running Linux in the cloud (from just a per-hour pricing). Note that at that price point, in both cases you are not going to get support from any vendor (you would have to either get support from Microsoft for Windows or pay for a Linux subscription from Red Hat or another provider). This leads to several observations.
First, this demonstrates that current software licensed-based business models are ill-suited for the cloud. In order to be competitive, cloud providers are looking for ways to scale their business at the lowest possible cost, which, most of the time, means they’ll always prefer relying on Open Source software if it is feature-competitive with its proprietary counterparts. At the infrastructure layer, this is almost always true: virtualization, routing, firewall, load-balancing, storage, DNS, SSL, etc. are king in Open Source land. This makes it very hard for vendors like Microsoft and even VMware to remain competitive when up against cloud-providers’ offerings aimed at commoditizing the market. Consequently, in the new cloud era, especially at the infrastructure layer, either you are able to provide a full-fledge SERVICE offering (that is the “S” in “XaaS”) that can be packaged in a way that makes sense to your market, or times will be hard, very hard.
While this post focuses on Microsoft, it doesn’t mean that things will necessarily be rosier for Open Source vendors down the road. Let’s open a quick parenthesis on this aspect. Once a cloud provider has decided to move with Open Source software, the next question is whether they need to rely on a third-party company to get support for that open source components they are using? While they might need help initially, providers will typically have grown the required expertise in-house as they expand, with no need for a high-friction process involving a third-party vendor. Furthermore, since their infrastructure is very homogeneous (hundreds of thousands of identical servers!), they can support their open source software at a much lower price point than any generic subscription provider ever could: they do not have to run gazillions of tests on gazillions of different motherboards, chipset, LAN cards, controllers, disks, CPU, etc. Their domain problem isn’t a problem – as their infrastructure is hyper-focused. This point is illustrated by the recent launch of Amazon’s own Linux distribution, based on RHEL/CentOS: Amazon Linux. Not only is support for that version of Linux provided as part of the default IaaS support cost (and as we know, FREE is a tough price point to beat), but it probably saves costs to Amazon since those customers are using a stack that Amazon understands extremely well, hence reducing the probability of hard-to-diagnose support tickets.
The bottom line is that any infrastructure subscription/license business, open source or proprietary, is a hard business in the cloud era.
But let’s go back to our point with regard to Microsoft Windows. So, Windows is not only up to 40% more expensive to run in the cloud, but the tools available to build a Windows AMI (vs. a Linux AMI), customize it, provision it, build a JEOS (Just Enough OS) are much weaker on Windows than on Linux. Anybody who has tried to properly automate the usage of Windows and Linux in the cloud will tell you that none of this is trivial, and that Windows is particularly hard. Second roadblock to Windows’ adoption in the cloud: Windows’ cloud DNA is weak, very weak.
Let’s pursue on the on-ramping of new cloud users. About a year ago, Amazon launched their “free usage tier” offering. The idea was to provide free compute capacity to new subscribers so they could experience and learn the cloud at no cost. Yet, because this was a free offering, it was only made available for Linux-based AMIs. This meant that for about a year, tens of thousands of developers, DevOps, sys admins, etc. have learned how to use the cloud on Linux, not on Windows. Third roadblock: Windows is not a good candidate for on-ramping new users to the cloud, both for technical and pricing reasons.
Microsoft has probably tried to ignore this for while but this wasn’t sustainable: Amazon announced today that Windows is now also eligible on these free tier instances. For this to happen, MSFT probably had to provide free pass-through licenses to Amazon for that very purpose and maybe even more as this is really a way for Microsoft to tap into AWS’s on-ramping funnel. As we all well know, whenever we read “Microsoft Windows” and “free” in the same sentence, we know that something big had to force that situation.One last roadblock? Well, now you can try to do something meaningful on Windows with 613MB of RAM, which is what AWS’ free micro instances provide…
The cloud is redefining the IT industry as we’ve known it and re-shuffling power among vendors. This is an opportunity for new entrants such as CloudBees to make a difference as well as for dominant positions to shift base. My humble opinion? Amazon will become the new Microsoft and today’s news might be one of the first symbolic events that signal that shift. It will be very hard for Microsoft to turn their ship in a new direction.
Oh, and remember, you can test whatever the cloud revolution means to you, as a Java Developer, on the CloudBees PaaS. Just try us for free! ;)
Posted by Sacha in CloudBees, English, IT.
I always start the New Year with plenty of well-intended resolutions. Never mind, most of them don’t last: as soon as I start hitting the post-break reality, resolutions seems harder to satisfy, exceptions to whatever I have committed to seem to stack up and, oh well…maybe next year.
For now, the same logic doesn’t seem to apply at CloudBees. Here’s why.
In 2010, we committed to ourselves that in 2011 we would offer the first end-to-end, development-to-production Java PaaS in GA. Last year, in January 2011, we announced the first end-to-end, development-to-production, Java PaaS in GA. That was a great way to start the year — without breaking an important New Years resolution.
In 2011, we committed to aggressively communicating to the industry who CloudBees is and what we offer. We did just that. As if it was a mirror image of January 2011, this January–in just a few days—also saw that resolution come to be! In just a few days, CloudBees received three awards!
The next day, Dr. Dobbs announced the JOLT Awards. CloudBees received an Honorable Mention and, according to Executive Editor Andrew Binstock, “…CloudBees was included for special distinction because we strongly feel it indicates an important new direction in coding.”
With such a great start, I have a very good feeling about CloudBees in 2012…and, I can’t wait to see what NEXT January may also bring! After all, good things happen in threes…