Is Sun subverting Linux from the inside?

Industry watchers claim Sun Microsystems is playing a dangerous game with its decision to position Solaris as open source – a move which will see it go head to head with Linux
Written by Cath Everett, Contributor

Sun plans to open source the latest release of its Unix operating system, Solaris 10, during the second quarter of this year under the moniker OpenSolaris. (Some say it has already taken initial steps with a performance management component called Dtrace).

Thomas Goguen, vice-president of product marketing for Sun's operating platforms group, explains the rationale: "We want to expand the ecosystem for Solaris. The vibrancy of an operating system depends on the number of organisations associated with it. We've spent half a billion dollars engineering Solaris and that investment was made for free, but we've now made it open source and this will all contribute to the vibrant environment around it."

But Andy Butler, a vice-president at analyst Gartner, sees the move as somewhat "opportunistic" as Sun bids to ride the "fashionable" open source wave in a bid to counteract the threat posed by Linux. He believes it reflects a desire to reverse the widespread view that the vendor is "very proprietary", and to point up the fact that it has been a major contributor to the open source community for years.

The issue Sun is trying to address is that, over the last few years, Linux has been eating away at the lower end of its Solaris-on-Sparc-based business, which is still its cash cow. The freeware OS has increasingly been deployed in the workstation arena and in infrastructure areas such as file and print and Web hosting where Sun was traditionally strong.

But, as Butler explains, the issue is not a software one per se. "For customers, it's more of a hardware than a software-based decision due to cost, but because Solaris has not been seen as a viable x86 operating system, the result has been, without any real animosity, that people have been moving towards something native to x86. It could be Windows, but it's generally Linux."

Sun has see-sawed between whether to run Solaris on x86 over the years or not, but went down this route again in the first quarter of last year. The decision to open source Solaris is simply the next step, aimed primarily at expanding adoption of the OS on x86 -- although Solaris' code base is the same whether it runs on x86 or Sun's own Sparc processor, which means that the move covers both environments.

Since 2002, the vendor has also hedged its bets by supporting Linux on both Sparc and x86.

But Sun's situation has not been helped by the fact that over the last few years its core customer base in the financial and telecoms sectors has been reining in expenditure and going for cheaper options.

Chris Ingle, group consultant for IDC's systems unit, explains: "Sun traditionally had strong growth, but in a limited customer base -- it's very reliant on telcos and finance. It's not got as diversified a base or product portfolio as rivals such as IBM and the problem is that attempts to diversify into software and services have not resulted in the revenues it needed."

So to try and mitigate the effect of these trends, the vendor is attempting to promote Solaris on x86 as a Linux alternative. "2005 will see Sun going to war with Linux. Although it will continue to sell Linux on its servers too, it recognises that it's a major competitive threat to Solaris," says Butler.

Goguen prefers to pitch it differently, however, saying that the move is not so much about counteracting the Linux threat as supporting a volume business model to increase market share. "It's noise on the fringe that we're taking on Linux and trying to destroy it," he attests.

Instead, the idea is that proliferating Solaris on x86 will generate pull-through sales of other products from an enlarged user base.

"It's not just about volume servers, it's about volume around Solaris. The most successful model for expanding sales is offering a free distribution and services and support around that. Having our technology in the hands of third party developers, both commercial and academic, is better for the health of the OS and the market size of the company," Goguen says.

Butler, meanwhile, believes that, although Sun is keen not to be viewed as waging war with the wider open source community, it is, nonetheless, trying to drive a wedge between it and the Linux world.

"It doesn't want to be saddled with the baggage of being seen as a rabid anti-open source vendor, although personally I think it will be seen that way. Taking on Linux is an incredibly dangerous strategy, but while it will be difficult to stem the bleeding and stop defection, Sun will give it a damn good try," Butler says.

The vendor's aim is to show that "we do Linux better than Linux" and it will, therefore, emphasise the fact that, as a result of its Project Janus initiative, Solaris 10 can run native Linux binary applications without any changes being required and without customers and developers having the "aggravation of learning a new OS".

"Sun still has an enormous portfolio of applications and it wants to change most of them into native x86 ones," Butler explains. The challenge here is that, up until a couple of years ago, Solaris was both a key target platform for ISVs and "the development OS of choice", but has now moved to being "a significant port".

But, while Sun is keen to "reawaken some developers' loss of commitment", in practice, the world has moved on, with the focus no longer being on which OS to write to, but on which over-arching development framework fits requirements best.

Today, this choice boils down to Microsoft's .Net or Sun's Java, but the Java community tends to use Linux rather than Solaris as its underlying OS.

"It's pretty hard. The value of Java is in the notion that it is OS-independent, so there's a danger that Sun will be seen as a bit hypocritical if it pushes Solaris too hard here," Butler points out.

But Sun is also taking another tack to woo developers -- and customers -- over. It is pitching the notion that tools and utilities such as workload management are more functionally-rich and mature in the Solaris world than in the Linux one. "Sun's view is that Linux is building its credentials here, but it's weak, whereas it can give the best of both worlds," Butler explains.

Also inherent in the binary compatibility message, however, is the idea of scalability. "Sun is saying that you can run your Linux or Solaris application on a Solaris-based Opteron machine, but as it grows you can also stick it on a giant system. The problem is that people have been scaling down, not up," says Ingle.

Another challenge that Sun faces here is winning ISVs such as Oracle over to endorsing their Linux-based applications on Solaris machines. "ISVs are notoriously risk-averse and cautious about making glib guarantees of support. Sun is going to have its work cut out convincing them that Linux binaries will run without the slightest change, but if it can't convince them, the strategy will ultimately fail," Butler says.

While he believes that Solaris over time will become a "viable" x86 OS, he is not convinced that success here will lead to growth in Sun's overall Solaris-based revenues.

"Though Sun will gain a degree of sales on x86, it will come at the expense of its traditional Sparc business. It says that sales on x86 will be totally incremental, but we don't buy it. We believe there will be some attrition of the Sparc base and in 2009, Solaris sales will be little, if any, higher than today," Butler says.

Some 95 percent of Sparc processors currently go into machines with less than four CPUs, he attests, but it is just this area of the market that is being dominated by x86 boxes. This means that Sun's moves to promote Solaris on x86 will simply contribute to this trend. Moreover, because Sparc-based machines command higher margins than x86 ones, Sun will have to increase volumes proportionately to avoid overall revenues being eroded.

But because Sun is positioning itself as a Solaris champion and its "preferred gambit is Solaris everywhere regardless of the chip", the end result "will be to marginalise somewhat Sun's recognition as an independent x86 vendor. If you've never bought from Sun before and want a Linux box, why would you go to Sun?" Butler asks.

As a result, the majority of x86 sales are likely to come from "existing customers that know it and have a good reason to buy from it rather than from Dell or IBM that have more open credentials".

This means that, while the open sourcing of Solaris will "help to keep it alive and viable, it won't dramatically improve Sun's fortunes".

But the move also raises question marks over whether the remaining 5 percent of Sparc shipments at the medium to high end will be enough to sustain the life of the processor into the long term.

Sun has already deferred some research and development expense here as a result of cancelling its five-year long UltraSparc V project and allying with Fujitsu, the other major proponent of Sparc and Solaris, to jointly develop high-end servers based on Fujitsu's Sparc64 processors.

But as Butler points out: "Staying in the chip business is not cheap and Sun still has a lot of complex investment ahead, so it needs to keep the cash cow viable. It talks a lot about free Solaris, but there's no such thing as a free OS because someone has to pay to develop it. This means it has to rely on the health of continued demand for Sparc servers as the engine room to fund philanthropic investment."

Goguen refutes the suggestion that the future for Sparc is anything but robust, however, and also denies that an apparent lack of clarity about the positioning of its various offerings may confuse customers.

"It's about choice for customers and developers and now they do have a choice. If they want Solaris, we can do that, if they want access to OpenSolaris code, they can get under the covers, or if they want to use a different OS for different tasks, we can offer them that too," he says.

But Neil Ward-Dutton, director of Ovum's technology practice, believes that the company is sending out mixed messages.

"On the one hand, Sun is talking about how advanced Solaris 10 is, but then it gives it away," he says. "It's a double-edged sword and you can see it in two ways -- that it's giving away fantastic stuff or that it's desperate because it's giving away fantastic stuff."

The situation has not been helped by the fact that "Sun's overall marketing has been so confused over the last two to three years" and its financial results have been "mediocre".

Butler agrees. "People are nervous about doing business with Sun and they debate its corporate longevity, but we spend a lot of time reassuring them that there's no need. Sun is still financially viable and well managed and, while there's a degree of risk with the strategy it has in place, we don't think it's wrong given Sun's situation in the market," he says.

Nonetheless, open source is not for everyone, and as Butler points out, many organisations simply mistrust it as a business and development model, preferring to stay with proprietary operating systems.

This means "it's pretty important that, when Sun manages to profile these companies, it plays down the open source element and says that 'Solaris still belongs to us. We control how it's improved and are remaining firmly in control of where Solaris will go.' Sun will try not to rock the boat here," Butler says.

As a result, those customers with the majority of their IT infrastructure based on Solaris-on-Sparc are likely to remain loyal because "you have to be very frustrated with a vendor to kick it out and Sun hasn't done enough to be thrown out of accounts".

While Butler acknowledges that big Sun users are "looking at the situation with some concern," his advice is to "undertake investment with your eyes open and be cautious of x86 until it's proven that there's no risk". Nonetheless, he expects the "vast majority" of Solaris-on-x86 sales to be generated within the supplier's existing installed base.

On the flip side, however, Butler does not believe that the move to open source Solaris will win Sun many new customers. "For those that have been doing business with IBM and HP for 20 years, but have had a shorter relationship with Sun, they'll look at the relative risks and are likely to say that doing business with IBM looks easier. They're not going to kick Sun out, but they're less likely to make a new business investment," he says.

On the plus side though, the move will make it easier for the vendor to sell to government customers, one of its stated key target markets and one in which many organisations are making open source credentials mandatory. "With the public sector, Sun's ensuring it's not seen on the wrong side of the proprietary/open source divide," Butler explains.

But Ward-Dutton is sceptical of Sun's open source intentions as evidenced by its choice of licensing agreement.

Rather than adopt one of the widely-used and already available licences such as the GNU General Public License (GPL) under which Linux is made available, Sun has fashioned its own. The Common Development and Distribution License (CDDL) is based on the Mozilla Public License and was approved by the Open Source Initiative (OSI) earlier this month.

Unlike the GPL, which stipulates that all code developed and used under the licence must be made publicly available, the CDDL prohibits the release of proprietary code used in any project under its remit.

Goguen explains Sun's rationale. "The problem with the GPL is it's extremely viral and it's not clear how to combine source code using other licences because it all ends up being covered by the GPL," he says.

This situation meant that the vendor was obliged to craft a licence that was "friendly to developers and business and would offer them protections".

"It's about enabling people to build on the developments of others so you can keep the intellectual property yourself if need to. It's a far more liberal licence, which is much more friendly to innovators and business," Goguen adds.

But Ward-Dutton sees it more that Sun is "reacting to pressure from partners and investors to create an alternative foundation environment for open source projects that is more in keeping with making profits".

As such, he views the move as "an attempt to distort the whole open source process and I can't help thinking it will fall flat on its face", not least because, under the CDDL license, "Solaris code won't be able to intermingle with Linux".

As Russell Nelson, an OSI director, points out, however, this inability to swap code is a double-edged sword. "You can turn on the conspiracy machine and it would crank out 'but it keeps OpenSolaris from being subsumed by Linux'. The trouble with that theory is that it also keeps Linux drivers from being incorporated into OpenSolaris," he says.

Butler, meanwhile, sees the creation of CDDL as Sun's attempt to play the open source card, but not give away its crown jewels. "With the CDDL, Sun can portray itself as an open source champion, but still retain complete ownership and control of where Solaris goes," he says.

Not true, argues Goguen, who claims that the OpenSolaris development process will be "less restrictive than other open source communities" as it will have "no specific gatekeeper".

"No one person will decide what goes in, which allows for similar and new innovation. We'll occasionally do a distribution that we'll call Solaris and do testing of new releases and that will most likely include a subset of code from the open source community," he says.

This release will be made available for free, Goguen adds, but customers will pay for support.

As to how successful Sun's decision to open source Solaris will be into the long-term, Butler has mixed feelings. On the one hand, he views the move as broadly positive because it will help the vendor ride the open source bandwagon and gain visibility among public sector organisations. On the other, he thinks it unlikely that broader industry interest in Solaris will be dramatically reawakened.

"Sun is trying to hold on to what it's got and take any new business opportunities coming its way. This is an initiative to maintain the viability of Solaris, but it's unlikely to stop the low-end defection to Linux," he concludes.

Editorial standards