Companies Want Good Enough IT, Not 'Best of Breed'

by Timothy Prickett Morgan

If you are an IT organization accustomed to assembling your systems and software piecemeal using the so-called "best of breed" components on the market, or are one of the major IT players used to supplying such customers, watch out. Best of breed is a dying breed. With IT budgets tight for going on three years now, and decades of history during which so many well-intentioned IT projects failed, the best-of-breed approach to IT is being downshifted to "good enough."

Good enough is becoming fashionable, and it is not just suitable for governments anymore, but also for financial services firms, healthcare organizations, manufacturers, distributors, and various kinds of communications services providers, like telecoms and Internet hosting companies. For companies just like yours, in fact.

Good enough IT is not the same thing as using commodity components, but the two concepts are related. While high-end servers, storage, operating systems, middleware, and applications software have continued with their increasing number of bells and whistles, seemingly low-end alternatives to each of these IT components have matured to the point that they can rival the functionality of best-of-breed components from the late 1990s or early 2000s. To many a company, having IT systems and software that was as good as the top-of-the-line systems in 2000 is a serious upgrade. For these customers, today's good enough systems are actually better than they could have dreamed of affording in the recent past.

It comes as no surprise that we see the widespread adoption of X86 servers (both Xeon and Opteron) as core data processing systems, usually running Windows or Linux, in favor of more expensive RISC/Unix systems, which arguably have more scalability, reliability, and flexibility. They also cost two to three times as much to acquire. Similarly, midrange Unix boxes are displacing many aging and proprietary mainframes and midrange servers, because they offer functionality that is good enough and substantially lower costs to justify the replacement. To be sure, IBM continues to get mainframe enthusiasts to buy great gobs of mainframe capacity on a quarterly basis, which is making zSeries sales soar in recent quarters. But these companies are exceptions to the rule, and can benefit financially by consolidating vast networks of Unix and Windows servers they have sprawled around the world onto giant mainframes, with which they have expertise in administering very effectively. Similarly, the very high-end of IBM's OS/400 server customer base is trying to cope with a mix of platforms and can now consolidate OS/400, Windows, Linux, and AIX workloads onto a single box, with logical partitions or co-processor cards supporting these workloads. The number of such zSeries and iSeries customers measures in the thousands, not the millions.

For the millions, the Windows Server System from Microsoft, with its complete stack of software running on X86 iron, is good enough. Yes, Windows security is a nightmare, but Windows 2003 is a far cry better than trying to glue together Unix, Windows NT, and maybe NetWare systems into a cohesive whole. For customers who don't want Windows, the new commercial Linux variants from Red Hat or Novell present a good enough alternative, especially now that these companies are working on building out their middleware stacks through partnerships and by tapping into various open-source middleware projects.

Microsoft is betting that its various business software will be perceived as good enough for such customers, who might have otherwise contemplated using PeopleSoft, Lawson Software, Oracle, or any number of other midrange application suites, as they grew up into real accounting software from running themselves out of QuickBooks. (No offense to QuickBooks, which is good enough for tens of millions of small businesses that were previously running themselves from Excel spreadsheets.) And it seems likely that it will not be long before Red Hat and maybe even Novell wake up to the fact that the myriad customers that want good-enough IT really want a stack of open source ERP, CRM, and SCM software for specific tailoring, for a dozen of so industries.

Pick any component of any IT system and you can see the choice between best-of-breed and good-enough playing out. People buy Celeron desktops, not Xeon workstations. Companies are forcing IT vendors to debut hot-plug ATA drives and arrays when they want to continue to sell the more expensive, established, and reliable SCSI that have dominated the industry for a decade. IBM, Sun Microsystems, the Fujitsu Siemens partnership, and the Intel-Hewlett-Packard partnership continue to plough huge piles of money into their Power, Sparc, Sparc64, and Itanium systems, when what most people would be happy with is a 64-bit Xeon or Opteron box that is expandable, cheap, and supports the widest possible variety of operating systems, middleware, and applications. MySQL or ProgeSQL is good enough for a vast number of Web applications, and is being chosen over Oracle, DB2, and SQL Server because of this.

Whether opting for a best-of-breed or a good-enough approach, the one thing that IT managers should not do (and will not do, if they like their jobs) is to automate bad or dated processes. There is no benefit to fossilizing a business in time. Good enough does not mean being lazy. It means being smart and practical about what is good enough to get the IT job done. Any savings that come from this good-enough approach can be pumped back into the IT organization for those areas where a best-of-breed approach is still warranted. Like buying an iSeries to run core ERP applications in many OS/400 shops, for instance.