December 2009


After my previous post “Cloud to boost proprietary software use?”, Tim Bray questioned whether the pricing comparison of “WebSphere/SUSE vs. JBoss/RHEL on EC2 was a transient anomaly”. JBoss’ Rich Sharples commented that I was comparing apples and oranges.  That was not my intention.  I simply picked the only two application server Amazon Machine Images (AMIs) that I could easily find pricing for.  And in retrospect, my intention was not to compare proprietary versus open source pricing in the cloud.  But rather to compare the price differential of proprietary versus open source products in the cloud versus on-premise.

Let me try again with Windows versus Linux.  Specifically, I looked at the price of Windows Server 2008 R2 versus Red Hat Enterprise Linux (RHEL) on-premise and on Amazon’s Elastic Compute Cloud (EC2).  I wanted to evaluate how, if at all, the Windows price premium differs on-premise versus in the Amazon cloud.  One can argue that “you need 2 Windows servers to do the work of a RHEL server.” Such an argument has no impact on this analysis.  If you do in fact need 2, or a higher number of Windows servers per RHEL server, this ratio would hold equally well on-premise or on Amazon EC2.

Here’s what I found:

On-premise license:
Windows Server 2008 R2 Datacenter Edition: $2,999
Windows Server 2008 R2 Enterprise with 25 Client Access Licenses: $3,999
Red Hat Enterprise Linux Premium Subscription for 1 year: $1,299
Windows price premium: 130% to 208% [See UPDATE below]

Amazon EC2 license on Standard-Small AMI:
Windows Server 2008 R2:  $0.12/hr
Red Hat Enterprise Linux: $0.21/hr plus $19/month per customer
Windows Price premium: -43% [See UPDATE below]

If you’re surprised that the Windows Server AMI is 43 percent less expensive per hour than the RHEL AMI raise you hand [See UPDATE below].

Maybe you think I’ve missed some important or potentially hidden costs for the Windows AMI.  I may have. I’m by no means an operating systems licensing expert.  However, it’s difficult to accept that these costs would add up to Windows being 130% to 208% premium priced versus RHEL on EC2.  Even if I’ve missed a pricing component that doubles the “true” price of a Windows AMI in a production setting, that would roughly put Windows and RHEL at par in terms of EC2 per hour pricing.  That’s a far cry from the 130 percent to 208 percent premium for Windows over RHEL in an on-premise environment.

Hat tip to William Vambenepe for astutely pointing out that the license cost differential between proprietary and open source products narrows in the cloud.

[UPDATE:  2009-12-11 @ 5:45p EST — PLEASE Read]

Based on public & private comments here is some new information for readers:

1] The version of RHEL on EC2 is supported by Red Hat at the Red Hat “Basic Subscription Web support” level.  This includes  2 business day response, and unlimited incidents.  Red Hat charges $349/year for this license.  As previously mentioned the equivalent RHEL AMI (with an equivalent level of support) is $0.21/hr plus $19/month.

2] The version of Windows 2008 offered on EC2 is Microsoft Windows 2008 Datacenter R1 SP2 64-bit. The AMI is not supported as part of the $0.12/hr AMI fee.  However, to receive an equivalent level of support for this AMI as Red Hat offers for the RHEL AMI, customers can purchase the AWS Premium Support at the Silver level.  The AWS Silver Premium level support is $100/month, or the equivalent of $0.14/hr. Alternatively, to receive 24×7 support for this Windows AMI, customers could purchase the AWS Gold Premium level of support for $400/month, or the equivalent of $0.55/hr.

3] The price comparison now becomes:

On-premise license:
Windows Server 2008 R2 Datacenter Edition: $2,999
Red Hat Enterprise Linux Basic Subscription for 1 year: $349
Windows price premium: 759%

Amazon EC2 license on Standard-Small AMI:
Windows Server 2008 R2 ($0.12/hr) with AWS Silver Premium support ($0.14/hr):  $0.26/hr
Windows Server 2008 R2 ($0.12/hr) with AWS Gold Premium support ($0.55/hr):  $0.67/hr
Red Hat Enterprise Linux with Basic Subscription: $0.21/hr plus $19/month per customer
Windows Price premium: 23% to 219%

Key point to take away:
Holding the product version and support level constant across an on-premise license and Amazon EC2 instance, the price premium of Windows vs. RHEL, if X% for on-premise, will be less than X% on the Amazon cloud.  Said differently, the license cost differential between proprietary and open source products narrows in the cloud.

[ /UPDATE]

Follow me on twitter at: SavioRodrigues

PS: I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.”

See update at the bottom of this post.Based on public & private comments here is some new information for readers:

1] The version of RHEL on EC2 is supported by Red Hat.  The support level is: “Basic Subscription Web support, 2 business day response, and unlimited incidents”.  Red Hat charges $349/year for this license.  As previously mentioned the equivalent RHEL AMI is $0.21/hr plus $19/month.

2] The version of Windows 2008 offered on EC2 is Microsoft Windows 2008 Datacenter R1 SP2 64-bit. The AMI is not supported as part of the $0.12/hr AMI fee.  However, to receive an equivalent level of support for this AMI as Red Hat offers for the RHEL AMI, customers can purchase the AWS Premium Support at the Silver level.  The Silver level support is $100/month, or $0.14/hr.

3] The price comparison now becomes:

On-premise license:
Windows Server 2008 R2 Datacenter Edition: $2,999
Red Hat Enterprise Linux Basic Subscription for 1 year: $349
Windows price premium: 759%

Amazon EC2 license on Standard-Small AMI:

Windows Server 2008 R2 ($0.12/hr) with AWS Silver Premium support ($0.14/hr):  $0.26/hr
Red Hat Enterprise Linux with Basic Subscription: $0.21/hr plus $19/month per customer
Windows Price premium: 23%

Last week Eben Moglen, founder and executive director of the Software Freedom Law Center (SFLC), submitted an independent opinion on the Oracle/Sun merger to the European Union (EU). Moglen summarized his submission as follows:

“The GPL was designed specifically to ensure the permanent freedom of software, and the ability of everyone to improve and share their improvements to the program, no matter who acquires the copyrights to the code.  The whole point of GPL as a copyright license is to deal with every contingency that could result in hobbling or destroying the freedom of code shared under it. The drafters of GPL versions 2 and 3 considered scenarios very similar to the ones that the Commission is concerned about now. The design of the license, and the experience we have had using it, show that it can be counted upon to operate as intended in situations like this one.”

Moglen issued the 11 page opinion, pro bono and without the charge, at the request of Oracle’s counsel.  Moglen clarified that Oracle is an ongoing contributor to the SFLC, while Monty Widenius has contributed in the past.  However, neither the contributions from Oracle nor Widenius have exceeded 5 percent of SFLC’s funding since inception.

I found the following paragraph from Moglen’s submission particularly interesting:

“MySQL is now and always has been an atypical GPL software project, because its copyright was highly centralized inside a small commercial firm that considered dual licensing its only commercially attractive strategy for survival. But even MySQL AB’s atypical business model, which was highly unreflective of the mass of GPL’d software development, occurred within the parameters of the GPL’s overall design, which is to ensure the freedom of the software it protects regardless of the commercial motivations or behaviors of the parties distributing the primary upstream version.”

On the other end of the debate, Florian Mueller announced that he has submitted a 31 page rebuttal to Moglen’s position.  Mueller provided a summary of the highlights via email, from which I selected these comments:

“Fundamentally, his paper offers a different prediction as to what would happen post-acquisition. He simply expresses his firm belief that whatever made MySQL successful in the past is not really an indication for the future. In fact he believes MySQL AB had a very suboptimal business model…

If he were right that MySQL AB and all of the companies that succeeded around MySQL didn’t do it right and that a GPL-only approach works best, then actually there would be no point in Sun having acquired MySQL last year nor in Oracle acquiring it now because then the future would at any rate be that someone has to fork it and do a GPL-only project dependent on voluntary contributions. Interestingly, that approach would have been possible during all of those almost 14 years that MySQL has been available and no one, not even Eben Moglen, decided to seize that opportunity.”

Both Moglen and Mueller make strong and weak points.

First, Moglen is too quick to dismiss MySQL as an atypical GPL project.  As Mueller points out, whatever you think about MySQL and their business model, you can’t simply conclude that another business model would be more appropriate.  Just because Linux is licensed under the GPL and Linux vendors, namely Red Hat and Novell are closing in on a combined $1B in revenue, does not mean the GPL is the best license for every open source product with commercial aspirations.  The Linux ecosystem is very different than say, application servers or web content management.  Different markets with different ecosystems require different license considerations.

While Moglen appears to be arguing for a “pure GPL” MySQL, departing from the dual-licensed status quo, Groklaw reports that Mueller and Widenius would like to see the MySQL open source license changed from GPLv2 to the Apache Software License.  According to Groklaw, page 19 of an unreleased submission to the EU from Mueller/Widenius stated:

“We would like to draw attention to the fact that some major concerns about the effects of the proposed transaction could be somewhat alleviated by requiring that all versions of MySQL source code previously released under the GPLv2 license …must be released under a more liberal open source license that is usable also by the OEM users and would also create an opportunity for other service vendors to compete with offerings comparable to MySQL Enterprise. A good candidate is the Apache Software License (ASL).”

Something doesn’t feel right about Widenius proposing a license that MySQL could have chosen “over the past 14 years”.  Clearly MySQL decided against this move as the GPL/dual licensing approach led to a competitive advantage that the ASL v2.0 would not provide MySQL.  But I guess that’s why Widenius suggests Oracle should be forced to re-license MySQL under a permissive license such as the ASL v2.0.

We haven’t heard the last from Widenius & Mueller. Enjoy your holiday season ;-)

Follow me on twitter at: SavioRodrigues

PS: I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.”

I’ve argued that Cloud vendors are in a better position to monetize open source products than open source software vendors themselves.  I’m not alone in this thinking.  In William Vambenepe’s post, “Cloud + proprietary software = love”, there may be yet another reason for open source vendors to drink the Cloud Kool-Aid with caution.

Vambenepe’s writes:

“If Cloud providers get it right and software vendors play ball, the ‘proprietary vs. OSS’ debate may become more favorable to proprietary software in the Cloud than it is on-premise.”

Vambenepe’s argument centers on Cloud computing alleviating buyers from the hassle of license management.

“The removal of license management as a concern may not make a big difference to large corporations that have an Unlimited License Agreement, but for smaller companies it may take a chunk out of the reasons to use open source software. Not only do you not have to track license usage (and renewal), you never have to spend time with a sales rep. You don’t have to ask yourself at what point in your beta program you’ve moved from a legitimate use of the (often free) development license to a situation in which you need a production license.”

With license management being equally unnecessary for open source or proprietary software in the Cloud, buyers can focus on other selection criteria such as costs. Unlike development or unsupported usage scenarios, both open source and proprietary products have an associated license costs in the Cloud.  This is obviously for products delivered and supported by the Cloud provider or the Cloud provider’s partner.  To make Vambenepe’s point, I decided to compare pricing for WebSphere Application Server and the most closely related Red Hat product available on Amazon EC2.  Amazon charges $0.795 per hour to run the IBM WebSphere Application Server on Novell SUSE Linux on a Standard Small Amazon Machine Image (AMI) in the US.  This charge includes Amazon EC2 costs and the per hour license costs of Novell SUSE Linux and WebSphere Application Server.  Alternatively, Amazon charges $1.21 per hour to run JBoss Enterprise Application Platform on Red Hat RHEL on a Standard Small EC2 instance plus a monthly $119 recurring fee.  [UPDATE: 2009-12-11: Rich makes the point I am comparing apples to oranges.  That was not my intention. I simply picked the only two application server Amazon Machine Images (AMIs) that I could easily find pricing for. And in retrospect, my intention was not to compare proprietary versus open source pricing in the cloud, but rather to compare the price differential of proprietary versus open source products in the cloud versus on-premise.] Obviously this is simply one comparison and other comparisons will play out in favor of open source product pricing on the Cloud.  But that’s not the point. As Vambenepe argues, both open source and proprietary products require a license in the Cloud, and the license cost differential doesn’t always favor open source product in Cloud environments.

With license management a wash and license cost differential being much more balanced than one may instinctively think, Vambenepe argues that choosing proprietary software in a Cloud deployment is likely to be an easier decision than on-premise.

Tim Bray wrote “Makes all sorts of sense to me” when tweeting about Vambenepe’s post.

What do you think?

Follow me on twitter at: SavioRodrigues

PS: I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.”

I previously wrote that open source freedom helped me choose Western Digital’s MyBook World Edition (MBWE) over Apple’s Time Capsule as our home backup solution.  I hadn’t considered the value of open source freedom to Western Digital’s product planning process.
I initially selected the MBWE over the Apple Time Capsule, even at a $30 price premium, with the following justification:

“…since MBWE is running a Linux kernel, the ability to add functionality to the device was almost limitless. There’s a strong community of MBWE users that have everything from BitTorrent clients, to PHP, to a PBX running on MBWE devices.”

Just after I wrote the post I decided to check for a firmware upgrade.  Lo and behold, the MBWE firmware now supports Time Machine and an open source BitTorrent client.  These two features were top of my ‘wish list’ with the MBWE.

Then it hit me.  Western Digital is reading my mind, and maybe my email.  Well, maybe not.  But they are absolutely learning from MBWE users and adding features to the supported firmware that advanced users have added to their customized MBWE devices.  Reading the MBWE forums as I have for the past 3 months, Time Machine and BitTorrent clients are very popular customizations to the MBWE.  Before Western Digital added ssh access to the MBWE, it too was a popular advanced user customization to the MBWE.  I differentiate between advanced users and typical users.  Unlike typical users, advanced users tinker with products and tailor the product to their own needs. Western Digital is smartly looking at the customizations that advanced users seek and offering these customizations within the base product. User actions trump user surveys in product and feature roadmap discussions.  A typical user may not have gone through the effort, nor had the technical wherewithal, to install a BitTorrent client on their MBWE.  But the typical user is likely very happy to see it as a feature in the MBWE firmware.

I’m convinced that consumer electronics vendors have much to gain from using and exposing open source within their products. Sure, there’s a risk that another vendor could repurpose the open source software to build a competing consumer product.  Considering how tightly integrated and optimized the software and hardware is in a consumer electronics device, access to the software alone isn’t a compelling competitive issue.

I’ve previously argued that consumer device manufactures should open source their firmware.  As a very happy MBWE customer, I’d reiterate this call.  And remember, I paid a price premium over an Apple product for the MBWE.  I’d happily do so again and highly recommended the MBWE.

Kudos to Western Digital for using open source to offer user freedom, a great user experience and value for (my) money, whilst serving Western Digital’s own profit motivations.  A win-win, enabled by open source software.

P.S. My only issue with the MBWE is the network transfer speed, in the 15-20 megabyte  per second range, pale in comparison to the 125 megabyte per second theoretical speed of the advertised Gigabit network adapter. The speed is however faster than the 10-12 megabyte per second I achieved with the Gigabit-rated Apple Time Capsule on the same Cat5e Ethernet home network.

Follow me on twitter at: SavioRodrigues

P.P.S.: I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.”

Earlier this week OpenLogic announced commercial support for CentOS, a Linux distribution derived from Red Hat Enterprise Linux (RHEL).  The CentOS distribution is ostensibly RHEL source code with Red Hat’s branding and artwork removed.

Some asked if OpenLogic was “stabbing Red Hat in the back”.  OpenLogic positioned the move, not unexpectedly, as healthy competition.  OpenLogic’s Kim Weins explained that OpenLogic customers have been asking OpenLogic to provide Linux support for some time now.

“We do have a value-based approach for OSS support – customers want OSS support, but want it cheaper.

How, if at all, will OpenLogic’s support for CentOS impact Red Hat’s revenue?

Well, Red Hat has already proven itself in the face of Oracle offering a Linux distro based on RHEL.  Novell also offers low-cost RHEL support for up to 3 years during a migration from RHEL to Novell SUSE Linux Enterprise Server.  And yet, Red Hat’s Linux revenue is growing in the 20 percent year-to-year range; very respectable, especially in today’s economic climate.

For some customers, the OpenLogic CentOS offering becomes a negotiation tactic with Red Hat.  But customers already had the option of using Oracle Unbreakable Linux (OUL) during Red Hat negotiations.  And well, Red Hat still sports a 20 percent plus growth rate.

Then there are joint OpenLogic and Red Hat customers.  Based solely on Red Hat’s market share, I’d guesstimate that a large portion, upwards of 70 percent, of joint OpenLogic and Red Hat customers are using RHEL today.  These joint customers will begin to think seriously about their RHEL investments.  Some will begin by categorizing their RHEL deployments into “mission critical” and “less than mission critical”.  RHEL will remain the operating system of choice for the former, while CentOS could begin to displace RHEL in the latter.

The more interesting consideration is how RHEL customers unfamiliar with OpenLogic will react to this news.  Some RHEL customers were surely looking to reduce operating system costs by using CentOS over RHEL.  Until now, there hasn’t been a well established CentOS support option.  For instance, the “Commercial Support” page on centos.org reads:

“This Section or Page is coming soon. This page is a holder for Content that is not yet available for publication.”

While OpenLogic isn’t a market gorilla, they are definitely an established open source support vendor.  Additionally, by now, RHEL customers, and their management chains, are no longer concerned about the “risk of using Linux in the data center”. As such, the price premium for the enterprise Linux market leader, RHEL, is no longer the sacred cow.  Comfort with RHEL and Linux opens up the discussion of moving some workload from RHEL to CentOS in search of lower costs.  Matt Asay pointed out that Orbitz and others are running mission critical applications on CentOS.  While these customers are a thorn in Red Hat’s side, they represent a very small portion of Red Hat’s customer base.

It’ll be interesting to check back in with OpenLogic in 6 months.  Who knows, maybe OpenLogic’s success will make it an attractive acquisition target, maybe even for Red Hat?

Follow me on twitter at: SavioRodrigues

PS: I should state: “The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies or opinions.”