Most important, all the system and management tools that have made proprietary UNIX boxes the systems of choice for enterprise-class data centers are available on Linux. These include clustering/fail-over, logical-volume management, fiber-channel shared storage, SMP support into the tens of CPUs, and kernel features required to support high performance, high capacity, and high reliability.
A typical Linux migration engagement will begin with a review of the current data-center configuraion, operating process, and applications being supported, to determine system and middleware product requirements and to identify potential process issues. We'll then develop a migration plan based on the value and risks for each application/server. Usually this will begin with servers that primarily run "off-the-shelf" applications such as Web Servers, File Servers, and small data-bases.
A "Linux test-bed" will be implemented, and the performance and functionality of each application being migrated is evaluated on that test-bed before the pre-Linux application is replaced in production by the Linux-based version.
Generally, improvements to data-center configuration and operations will also be made during the migration process. These include server clustering for high availability, implementation of network shared storage, system/network/application monitoring, automated system installs, etc.