Linux: Not Ready for the Data Center

Computerworld - The virtues of Linux are well known: It's free, open source and pretty cool, especially if it's running in a bunch of high-volume locations, like a chain of retail stores. Issues of support and reliability have been addressed to some degree as well. For example, over the summer, Oracle ported its database engine and apps to Red Hat's Linux distribution and announced that it would offer enterprise-level operating system support, thus joining Veritas, IBM, Hewlett-Packard and Dell on the Linux bandwagon. Even Linux's performance is starting to look good.
But despite these advances, it's still unclear whether going with Linux (your CFO's interest in anything free notwithstanding) can save you money inside the data center.
Brian Richardson, program director at Meta Group, says that while Linux has moved from "bleeding-edge" to "early-adopter" status, multiprocessor CPU deployments are rare.
For a serious Linux analysis, he recommends doing two things. First, make a sensible comparison: "Wintel vs. Lintel isn't as true a comparison as RISC vs. Lintel," he says. After all, Microsoft's Advanced Server Edition has already been proven to eight processors. And don't compare an overconfigured Unix setup with a smaller Intel platform. Apples-to-apples comparisons are necessary because the costs for storage, servers, middleware, applications and support are about the same for Linux and Windows.
Second, Richardson calls for CIOs to check the total cost of ownership for Linux vs. Windows, because the real comparison should be at the database pricing level. For example, running Oracle vs. SQL Server.
At the application level, there's nary a difference between Unix and Linux, and the same holds true for the database level. No, it's at the RISC and Intel hardware level that real cost comparisons make sense.
So look at your IT infrastructure and your cost structure closely before going with Linux.
Consolidating a lot of little boxes into one large server sounds enticing, but Richardson says that generally, he has yet to see a large ROI from such efforts. Moving from 100 boxes to 10 won't change end-user support charges, nor will it modify database administrator and application costs.
And finally, Richardson argues, the shadow support costs (unlogged time doing work-arounds and enhancements, for example) might be higher for Linux than for Windows.
Talking up the merits of Linux is an attention-getter, especially in a tight budget environment. But at the current stage of Linux development, it's incumbent upon IT to investigate how many months it will take to build the distribution tools, utilities and enhancements that will make Linux ready for the data center.
Pimm Fox is a freelance writer in San Francisco. Contact him at pimmfox@pacbell.net.

Read more about Linux and Unix in Computerworld's Linux and Unix Topic Center.