Look – up on that pedestal. It’s Linux, basking in the glow of popular attention. However, one analyst is cautioning that the open source OS isn’t going to be everyone’s saviour.
Jonathan Eunice, analyst at Illuminata Inc. in Nashua, N.H., thinks the recent burst of excitement surrounding Linux is great but he doesn’t want large enterprises getting so caught up in the excitement that they buy into something that is not in their best interests.
“As soon as there was that groundswell of opinion, people were immediately talking about how it could displace Windows NT, how proprietary Unix was dead meat, how Linux would take over everything — and we felt that was rampant over-optimism,” he said.
Linux’ strengths — low cost and overhead, its abilities as a network server and the fun involved in being part of the open source community – are outlined in a report written by Eunice titled Why Not Linux?.
But for large commercial applications, Eunice said, one drawbacks is that, although all Linux variants are based on the same core code, different versions each have their own install procedures, update strategy and tools, and set of fixed and outstanding bugs. Eunice said this can end up creating a “quality control nightmare.”
Linux’s “tiny design centre” is also an issue, according to Eunice — most Linux developers work on single-processor servers with maybe a few attached disks, and Linux is currently a single-CPU operating system.
“Linux does not have the kind of system size or scalability that lets it run a big database or big application,” Eunice said. “So it will do a heck of a job with an Apache Web server or an FTP server…or traditional Unix network server tasks, but a big database requires multiprocessing and it requires a really sustained commitment to high I/O rates. Scalable I/O is really the huge sticking point.”
Bob Young, chairman and CEO of Linux vendor Red Hat Software in Durham, N.C., admitted this is currently a limitation of Linux but said this is going to get better, especially in the area of symmetric multiprocessing.
“With the new kernel — the 2.2 kernel — you’ll see this improve,” Young said.
But, Eunice countered, “I still don’t think that the motivation or the capability is there for Linux to become the same as an HP-UX or a Solaris or an AIX in running big apps, because the people that are doing the development of it are doing it for their own enjoyment,” Eunice said.
“In some cases, groups that are VARs are making some money off of it, but they just do not have the market opportunity for building 16-way or 32-way systems and terabyte disk farms, whereas HP and IBM and the others…are selling these large commercial systems every day.”
Evan Leibovitch, director of the Canadian Linux Users Exchange (CLUE) and partner in Brampton, Ont.-based Starnix Ltd., acknowledged Linux is not an adequate competitor among high-end data-warehousing systems.
“But right now, the operating system is not that old,” he said. “It doesn’t have 64-processor capability yet, it doesn’t have gigabytes of RAM or terabytes of disk — but those things are coming.”
There are also support issues, according to Eunice. You can purchase good support from a number of VARs, but scale and depth are still problematic. Major Unix vendors can bring kernel developers, performance engineers and test-to-scale laboratories in when mission-critical systems and applications fail. But as Eunice wrote in his report, “Linus Torvalds (creator of Linux) and his network-distributed crew cannot be contracted on the same basis.”
Thomas Goguen, product line manager of Solaris servers for Sun Microsystems Inc. in Menlo Park, Calif., said users should consider maturity.
“If you’re an enterprise or a large customer, you’re probably going to look at one of the more mature companies — Sun, HP, IBM — to deliver that kind of support,” Goguen said.
Miles Barel, software product marketing manager for IBM Corp.’s AIX division in Austin, Tex., agreed. “Although you can look at some of the companies that are getting behind particular Linux deliverables — and they certainly are providing very significant and serious support — I believe you’ll find the levels of support are greater when you’ve got tighter control over the product.”
Red Hat’s Young said he has never claimed Linux is everything to everyone. “There are always limitations to all technologies, and all technologies are better suited to some applications than others,” Young said.
Although improvements such as a simpler graphical user interface and SMP support are in the works, Young said, gearing Linux toward the commercial enterprise space isn’t the number one priority for the near future.
He said the vast majority of machines being installed in general are not huge enterprise ones but broader and lower-cost solutions, so that’s not such a bad place to focus on.
“That’s where Linux excels, the departmental server level and the network server level,” Young said.
