Last week's Linux World Conference & Expo provided another opportunity for Linux to emerge as a mainstream corporate computing option. However, the elite of mainstream technology providers still overshadowed the open-source evangelists who hold fast to the mantra that software "wants to be free." Keynote speakers included representatives from IBM and Dell. The Enterprise Solutions Center was staffed with representatives from Hewlett Packard (HP), NEC, Intel, Oracle, Sybase, and personnel from high-profile open-source projects (e.g., the communication protocol Samba, the Web server Apache, the email program SendMail).
Linux initially gained attention as an OS server for noncrucial applications (primarily Web serving); however, its suitability as an OS for all computing applications—from handheld devices to data center platforms—is widely debated. Several large-scale projects are exploring Linux's suitability as an enterprise-class platform. A case in point is the Beowulf Project, which is geared toward generating supercomputer performance from commodity computer components running Linux.
The question is no longer whether Linux will find a place in enterprise computing but rather how big its role will be and how quickly Linux will fill it. The buzz at Linux World was that Microsoft has dropped its campaign to discredit Linux and open-source software in general. Microsoft now views Linux as an important fixture in the competitive landscape. That view is correct: Market research company IDC projects that Linux will enjoy a compound annual growth rate of 28 percent through the year 2006. IDC researchers believe that Linux will make gains as both a server and a client operating environment.
Although most media hype about Linux usually focuses on the threat it poses to Windows, Linux's gains have actually come at the expense of UNIX, the OS on which Linux is based. Linux will probably affect the storage market in two ways. First, Linux will replace UNIX in Storage Area Networks (SANs). Second, Linux's growth will stimulate the creation of new storage technologies designed to maximize large computer cluster performance.
According to IDC, currently, 28 percent of UNIX storage occurs on SANs, as compared to 8.5 percent of Linux storage. IDC predicts that over the next 4 years, the value of Linux-deployed SANs will grow sixfold, from a baseline of $58 million in 2001. Moreover, major storage vendors, including Brocade, EMC, HP, IBM, and VERITAS Software, have released Linux-based products.
Despite Linux's gains, many analysts agree that Linux kernel 2.4, the current version, has serious storage shortcomings. Matthew O'Keefe, founder and chief technology officer (CTO) of Sistina Software, believes Linux is limited to 2TB of storage per device and thus per file system. According to O'Keefe, to compete at the high end of the storage spectrum, Linux deployments need to resolve this storage limitation.
A number of storage vendors, including Sistina Software, offer a solution that creates a storage cluster that consists of a SAN, shared storage devices, a cluster file system, and volume manager running on the servers that are integrated into the cluster. A cluster file system lets any server read or write any file on the shared file store. You can map the single, shared cluster file system onto the shared storage hardware to create a scalable, manageable storage pool for all the servers. You can add new servers and new storage devices to the storage cluster without going offline.
Vendors created the cluster file system approach to take advantage of the added computing muscle that clustered computing solutions offer. In the future, the cluster file system approach could be key in enabling Linux to claim a role in the data center, running large-scale crucial applications.
For a decade, Bill Gates has had a vision of the Windows OS running everything—from your handheld device to the data center. Linux proponents hold a similar vision. How successfully the open-source community and its allies can realize that vision depends in part on how effectively they mitigate open-source software's storage limitations.