This document discusses best practices for virtual machines (VMs), storage area networks (SANs), and SQL Server. It provides three "nevers" for VMs: never overallocate virtual CPUs, never use automatic settings, and never assume VMs are alone. It also gives three "always" for SANs: always know your neighbors on the SAN, always test storage performance first with SQLIO before SQL Server, and always be checking performance metrics. Key metrics discussed include processor queue length, SQL Server memory page life expectancy, and physical disk read/write average time. The document emphasizes testing storage, understanding competition from other workloads, and monitoring for subtle performance changes.