In the 12 years since Linus Torvalds posted to Usenet about his free operating system that was "(just a hobby, won't be big and professional like gnu) for 386(486) AT clones ...," Linux has moved from the basement to the board room and in the process most likely made itself a permanent fixture in the server room.
Linux has also matured from a robust adolescence, a time when those in the server room had those in the board room convinced that it had the potential to be everything for everybody, to being known for being highly effective in certain conditions, such as those that benefit from scaling out, yet less desirable in others, such as those that require scaling up.
So what is the state of Linux in the enterprise, circa 2003?
To answer this question, ServerWatch attended Jupitermedia's Enterprise Linux Forum in Washington, D.C. two weeks ago. (Jupitermedia is the parent company of this publication.)
For one thing, there are fewer key players developing and peddling the operating system than there were even two years ago. Although the number of actual Linux distributions still extends well into the hundreds, Red Hat and Suse account for the lion share of worldwide organizational Linux deployments, and Red Hat alone owns more than 50 percent of the market. As a result, it is these two vendors with whom OEMs and middleware vendors typically partner.
Jim Enright, director of the Linux Program Office at Oracle, offered some differentiation between Red Hat Linux and Suse Linux, describing the Red Hat flavor as more business oriented while Suse's distribution is more technically oriented.
Jonathan Eunice, president and principle analyst at the research firm, Illuminata, said that in an enterprise environment, the choice of which distribution is "based on geography and industry." He added that all of the distributions share a common goal -- beating Red Hat.
As Linux matures, not only have the players in the arena become fewer, but the arena itself has become smaller. For with maturation has come specialization, and the realization Linux cannot be all things for all enterprises.
For example, one area in which Linux does not stack up well compared to Unix is scalability. Although it's much-improved, and the yet-to-be-released but oft referred to 2.6 kernel will offer up to 16-way SMP with support for 32-way (and Red Hat's just released Enterprise Linux 3 scales to 16-way), Linux is no match for a Unix configuration that can scale up to 256 processors.
And although Linux has found a home on the mainframe, most notably on the IBM zSeries, it typically follows a "scaling out" model even there, Eunice said.
While scaling up is not among its strong suits, Linux's modularity makes it a good candidate for infrastructures that scale out, Mike Balma, Hewlett-Packard's Linux strategist told ServerWatch.
This modularity works particularly well in a security environment, where the option to not allow certain privileges (e.g., FTP) is key, Balma said. This advantage has not been overlooked by the financial services community, where Linux is now well-entrenched.
This pattern is not a unique one. Unix flavors also conquered scaling out before scaling up and cut their teeth in the financial services community, Eunice said.
Enterprises that opt for Linux often end up with more, smaller systems rather than fewer large systems, trading in, for example, a 16-way for several 4-way or 8-way systems. In some cases, Balma said, an 8-way could replace a 16-way if the new system is more robust.
Balma believes that Linux will eventually dominate in the Web and high-performance computing space, but Unix will remain king in the database and application server space. Eunice believes that Linux, and by extension open source, is perfect for application developers and for deployment in network services and infrastructure, general utility computing, LAN servers, and scientific and technical computation.
Eunice added that he would not recommend Linux for "a full-board desktop solution or a very high-scale database with high transaction applications," as the databases and middleware that support Linux don't have the speed, functionality, and ease-of-use most users require. In general, however, "in places where there isn't a substantial advantage in a proprietary offering, open source is the way to go."
Balma believes enterprises considering Linux should ask the following two questions: 1) what is the application that it will be running, and 2) what management tools are available? The management software is often the most vital component, and HP's management software (OpenView, Insight Manager, and Rapid Deployment Pack), he claims, scales well horizontally.
With so many hardware vendors investing heavily in Linux, it stands to reason that migrating platforms and operating systems often go hand in hand, an assumption with which Balma concurs.
Oftentimes hardware savings are the driver for an enterprise-level Linux migration, Balma said. Windows-to-Linux migrations on the same hardware usually stem from a security issue, he added.
HP has become a staunch proponent of Linux, to the degree that its road map replicates that of the Open Source Development Lab (OSDL). Blalma said that the among HP's servers, the ProLiant line is typically the hardware of choice for Linux deployments. This supports the general consensus about where Linux's strengths lie.
Balma offered some statistics to illustrate Linux's presence in HP's hardware sales. According to IDC's latest figures, 80 percent of Itanium-based systems that ship with Linux are from HP's Integrity line. To put the number in perspective, however, sales of Linux on Integrity servers comprise slightly more than 15 percent of all Integrity sales.
For any new technology, standardization is typically a calling card of maturation. Linux is no exception to this, and the five-year-old Free Standards Group has stepped up to the plate with Linux Standard Base (LSB). Members of LSB span the IT landscape and include AMD, Connectiva, Debian, Dell, Hewlett-Packard, IBM, Intel, Red Hat, Samba, and Sun. Oracle is the one notable vendor absent. Scot McNeil, executive director of the Free Standards Group, attributes this to Oracle's general avoidance of consortia. (Oracle is also not a member of the OSDL.)
The LSB is not creating a new variant of Linux, rather it is charged with "creating standards for a Linux base to bridge the needs of the free and open source developer community and IT industry," McNeil said.
To that end, it has succeeded. Nineteen products have been certified for LSB v1.3 to date, including all of the major Linux distributions, such as those from Connectiva, Mandrake, Red Hat (with the exception of Enterprise Linux 3, which was released two weeks ago), SCO, Suse, and United Linux.
LSB v2.0 is currently in development.
Making It Work -- Linux in Action
Examining and tracking Linux in and of itself is meaningless. To measure the penetration of the operating system, one must take into account the software that runs on Linux. Stacey Quandt, a principle analyst for the OSDL who has been following Linux since 1998, notes that "ISVs not supporting Linux at this time are at a competitive disadvantage."
Eunice noted that all of the major OEMs support Linux, as do the leading system management software, middleware, and applications. Eunice added that like Unix in its early days, Linux has found a particularly loyal following on Wall Street. It currently has a strong presence in Merill Lynch, Ameritrade, Reuters, and Credit Suisse/First Boston.
Quandt said that E*trade recently migrated from Sun systems to IBM's xSeries 330 running Linux, and the Securities Industry Association division of the New York Stock Exchange relies on Linux.
Server blade vendors Egenera, which targets the financial services community, and RLX elected to go with Linux-based systems. This is not surprising, given that 33 percent of all edge computing is Linux-based. Balma expects that blades, which are typically deployed at the network edge, will mimic that.
Moving away from financial services, Linux continues to be strong. It has become such an accepted and standard component of enterprise computing that Eunice said, "it's getting to the point where you might as well list the people who don't use Linux."
For example, Amazon.com, long the poster child of e-commerce, may turn out to be the poster child for Linux profitability. Red Hat CTO Michael Tiemann, and others, referenced Amazon's long history of successful sales yet failed profitability -- until it switched its operating system to Linux. Prior to the switch, Tiemann explained, "Amazon's cost structure kept it from being profitable. Since switching to Linux it has made money off of every transaction."
HP has put its money where its mouth is with regard to Linux and now boasts about 3,500 Linux-based systems in-house involved in a range of activities, from DNS servers to clustering to electronic design, Balma said.
The Future for Linux
Now that Linux has its footprint in the data center, it is slowly making its way to the desktop. There was nearly unanimous consensus that Linux on the consumer desktop is "not ready for prime time," as Eunice put it. He contends, however, that specific organizations and institutions are ready for Linux on the desktop, and in such situations it would do fine on the desktop.
With such widespread acceptance for Linux, the question being asked by the Linux community clearly is no longer "is Linux ready for the enterprise?" but rather "are enterprises ready for Linux?" In most cases, the answer will be a resounding yes, which then begs the questions of where to deploy to maximize its strong points and what functions can it most effectively accomplish.