The gravy train that the Windows platform has been enjoying while attacking the vast installed base of Unix servers in the enterprise resource planning (ERP) application software market may be jumping tracks, and a new train appears to be rolling in: the Linux platform.
Companies are increasingly turning to Linux to run SAP, Oracle, and PeopleSoft ERP application suites, according to a survey by Peerstone Research.
These three ERP platforms - now two, as Oracle has just taken control of PeopleSoft - are the crown jewels of the IT sector. The biggest ERP shops tend to use the most advanced technologies, and tend to pick one of these three ERP suites. Peerstone says that the installed base of these ERP suites is comprised of about 40 000 to 50 000 companies and government bodies, which have around 100 000 unique ERP installations, running on anywhere from 700 000 to 800 000 servers.
The company estimates further that some 200 000 of those servers are due for an upgrade in 2005. Getting a small slice of that base is very big business. And that is why Peerstone's independent survey of big ERP shops focused on these three platforms.
Peerstone did in-depth surveys with 252 shops running an ERP application created by SAP, Oracle, or PeopleSoft. While there is some mixing and matching of platforms at some ERP shops, for the most part companies choose one platform and use it for both database serving and application serving, and they tend to replicate those platforms in the various business units or agencies.
According to data gathered between April and September 2004, 162 of the 252 ERP customers surveyed using big SAP, Oracle, or PeopleSoft set ups said that they were running their ERP software on Unix. Although the survey did not gather this information, Solaris, HP-UX, and AIX are the dominant Unixes for running ERP software. Some 71 customers, or 28% of those surveyed, said that they were running their ERP software on Windows. The survey did not peg what version of Windows they were using, but Windows 2000 Server and Windows Server 2003 are probably the main versions in use among such big shops, since Windows NT 4.0 Server is probably far too primitive.
Only five shops, or about 2% of those surveyed, were running their ERP code on Linux, and another 24 companies, or about 5% of organisations, were hosting their SAP, Oracle, and PeopleSoft ERP applications on other operating systems.
These percentages are more or less consistent with quarterly server market share statistics - it takes many years of platform sales to alter the installed base. Unix and Windows server platforms have about equal market share these days, in terms of revenue, but Unix has been selling into the data centre for almost a decade longer.
Legacy platforms - Unix has become one, and Windows is fast becoming one as well - persist because of the human elements of comfort, expertise, and inertia - which is a measure of the difficulty of and the resistance to change. It makes perfect sense for Linux to have a tiny market share at big ERP shops, especially considering that Linux 2.6 is the first enterprise-grade Linux and ERP software. Vendors have only got on board with supporting Linux as an equal to Unix and Windows within the past 18 months or so.
The interesting Peerstone numbers were not where companies are running their big ERP suites today but where they expect to be in three years.
Of the 162 Unix shops, only 126 will be around in 2007, based on current expectations of the ERP customers surveyed by Peerstone.
That drops the Unix share of the survey installed base from 65% to 50%.
Ten of the 71 Windows shops, or 14% of the installed base, expect to move to another platform. However, the Windows platform will gain an influx of customers (mostly from Unix shops that want to get off expensive RISC/Unix iron), and will drive the Windows count back up to 68 customers, or about 27% of the base of survey participants.
Linux is the Cinderella story in 2007, with 38 customers, or about 15% of those polled, expecting to be running their big ERP suites on Linux by then. About 10 customers polled (about 4% of those surveyed) are not sure where they are going to end up by 2007, platform-wise, for their ERP suites, and another 10 customers (again, 4% of the base polled) say they will stick with other platforms.
There is no question that Linux is gradually taking over as the Unix alternative, particularly for shops that want x86 iron but are not so keen to move to Windows, because it is so different from Unix and is perceived as having security issues, which Unix and Linux do not. Over the very long haul, by 2010 or so, according to Peerstone, the high-end ERP space will be roughly split into thirds, with Unix, Windows, and Linux all sharing more or less equal slices of pie of those 50 000 big ERP companies.
By Peerstone's estimates, about 1000 companies worldwide have deployed big ERP applications on Linux, and by the end of 2007, it should be about 8000 companies. That is tremendous growth in the course of only four years.
However, the future ERP platform pie is hard to predict, since Sun Microsystems is just now beginning to push its Solaris Unix on 64-bit x86 iron from Intel and Advanced Micro Devices. Hewlett-Packard is also just starting to get traction on HP-UX on Itanium and could eventually support Xeon and Opteron chips (despite its protestations).
IBM will probably not reverse course and push AIX on X86 iron, although the company could easily do so, since AIX was already ported to Itanium under the 'Project Monterey' effort with Santa Cruz Operation, the market leader in X86-based Unix. AIX was ready to go on Itanium in late 2000, and IBM pulled the plug on it when it launched AIX 5L 5.1 in early 2001.
At about the same time that IBM and SCO were putting the finishing touches on the Monterey 64-bit Unixes for Power and X86 architectures, Linux vendor Caldera International bought SCO, stopped selling Linux, and soon changed its name to The SCO Group. IBM quashed Monterey, only launching the Power version, and SCO soon thereafter sued IBM for allegedly putting Unix intellectual property in its AIX and Dynix/ptx Unixes into the open source Linux operating system.
The reason why I bring all of this up is that, in mid-2000, when the economy had not yet turned solidly south, you would not have been able to guess all of this. The worldwide recession, exacerbated by the 11 September terrorist attacks, put pressure on the Unix market and has helped to foster the move to x86 platforms.
With Windows being the only datacentre-ready alternative for running ERP packages for the past three years, it is no wonder that Windows has done well in big ERP shops. During a recession companies need to cut costs, and until the recent round of Unix announcements from IBM, HP, and Sun, Unix boxes were wickedly more expensive than Wintel boxes. It is safe to assume that RISC/Unix platforms will continue to be under pressure for the next five years, but the future of Unix on 64-bit x86 platforms is getting better every day.
So Unix may not drop as far as those in the Peerstone survey now expect.
Moreover, delays in future Windows releases and pricing issues on the Microsoft software stack could hamper the adoption of Windows if that platform - not just the operating system - is perceived as having a higher price tag than Linux. This is what appears to be happening in the market, by the way, and all of Microsoft Corp's noise about getting the truth about Unix has not changed the fact that Windows is more expensive than Linux because of security issues and Client Access Licenses, which Unix and Linux do not have.
One of the limiting factors to the future adoption of Linux is the pool of experienced programmers, analysts, and system administrators with deep Linux experience in these ERP environments, according to those organisations polled by Peerstone. While Unix skills transfer more easily to Linux than to Windows, Linux and Unix are still very different platforms in the details, where the devils, after all, lurk. I am optimistic about Linux, hopeful about Unix, and skeptical of Windows, and think Windows could fare the worst among the three in the 2008 to 2010 timeframe unless something radical happens with Microsoft's software pricing and the security features of its software.
I think that splitting the big ERP pie by 40% for Unix, 35% for Linux, and 25% for Windows is possible for high-end customers. I do not think Unix is going to get much traction in small and midsized businesses, unless Unix is priced like Linux: free right to use and reasonably inexpensive support options. This is what Sun is doing with Solaris 10, and for good reason: Solaris 10 is better than Linux 2.6, and for the same or less money on an X86 server, it is a better option - particularly for a Unix shop.
But not all of the Unix shops surveyed by Peerstone for its ERP survey may have the same position on the future of Linux. Of the 162 Unix shops surveyed, 78% said that they would stay with Unix in the next three years. Of the 36 Unix shops that expected to jump to another platform within three years, 29 said they would pick Linux, six said Windows, and one said it was not sure.
To say that Linux has knocked the wind out of the Unix-to-Windows migration, which has fueled the high-end x86 server market, is an understatement. Small wonder, then, that Unisys, which has partnered very tightly with Microsoft for its ES7000 Wintel mainframes, is now supporting and pushing Linux on those boxes. And while IBM and HP are pushing Windows on their big x86 boxes, IBM is also pushing Linux on its xSeries 'Summit' servers, and HP is pushing both HP-UX and Linux just as hard on its Integrity line of Itanium boxes. Three years ago, Windows was the only game in the X86 town when it came to data centre-class computing.
What seems clear is that companies wanted cheap iron and had no choice but to adopt Windows. But the resurgence of Unix on x86 (driven by Sun on 64-bit Opterons and Xeons and HP on Itaniums) and the momentum of Linux on just about all platforms will put Windows in a pincher.
Although Peerstone's analysts did not come off as pro-Unix and pro-Linux, as my commentary here makes clear that I am, the following extract from Peerstone's report sums up the situation Microsoft faces.
"Are Microsoft executives in denial about the progression of Linux in the core enterprise stack? While they are acutely aware of the menace Linux represents, it seems they are not prepared to acknowledge that any fundamental change in strategy may be necessary.
"They are still working on the assumption that basic blocking and tackling will carry the day. Microsoft has a unique marketing culture, propelled by an unequalled concentration of very smart, hard working people who are used to grinding out victory after victory. This is how they unseated Novell Netware in the 1990s from a position of incredible dominance in the LAN server market and installed Windows NT in its place.
"For the past decade they have been vigorously applying the same method to Unix in the enterprise, and until now they have known nothing but sustained success, although Unix - unlike Netware - is still a formidable opponent. So it is hardly surprising that Microsoft executives now have trouble switching perspectives and admitting that they themselves may have become the legacy system that is being challenged by a relentless newcomer with an alien culture and technology."
It will be interesting to see what Microsoft does, and how the Unix and Linux communities respond in kind - or unkind, even. And Microsoft will absolutely rule among small and midsized businesses, unless and until its vast partner channel supports Linux, as SAP, Oracle, and PeopleSoft have only recently done.
Microsoft is itself one of the biggest suppliers of ERP software to small and midsized businesses, and it does not need partners in that market in order to promote Windows as much as it needs SAP, Oracle, and PeopleSoft in the enterprise market. Microsoft may just concede the high-end Unix ERP market to Linux, lose that battle, and win the war for the small and midsized businesses.