The hidden costs of open source

Clusters based on open source software and the Linux® operating system have come to dominate high performance computing (HPC). This is due in part to their superior performance, cost effectiveness and flexibility. The same factors that make open-source software the choice of HPC professionals have also made it less accessible to smaller centers. The complexity and associated cost of deploying and managing open-source clusters threatens to erode the very cost benefits that have made them compelling in the first place. As customers choose between open-source and commercial alternatives, there are many different costs related to administration and productivity that should be considered. These are explored in this paper in order to give a true cost perspective. We also examine how a commercial management product, such as IBM® Platform™ HPC, enables HPC customers to side-step many overhead cost and support issues that often plague open-source environments and enable them to deploy powerful, easy to use clusters.

HPC: Why this debate is different?
The pros and cons of open-source versus commercial software have been debated in corporate IT for years. While sometimes viewed as a decision between Windows® versus Linux®, the issues are more complex than just the choice of an operating system. In HPC circles the Linux versus Windows debate has been largely settled, with Linux enjoying 75 percent market share compared to less than 6 percent share for Microsoft Windows

Linux has an even more commanding position among TOP500® supercomputing sites2. For this reason, in HPC the debate more often centers on whether to deploy a pure open-source management environment or to use a supported commercial product built for open-source Linux environments. Accounting for real costs Despite the many benefits of open-source software, it is not without its pitfalls. This is particularly true for smaller organizations that may lack staff with requisite expertise to operate and maintain open-source software and middleware
environments.