The database is the quintessential data dependency for any application. Databases in production environments tend to be performance sensitive and rely on consistent and predictable performance from their underlying infrastructure. Virtualization, despite its wide adoption and various benefits, has limited penetration in this space due to the hypervisor overhead and its inability to guarantee SLAs. Consequently, these databases are deployed on bare metal servers leading to sprawl, low hardware utilizations, over-provisioning for peak demand, and poor isolation.
All is not wrong with virtualization. Virtualization makes development and test databases agile and portable. Databases in VMs are fast to deploy and clone, have improved consolidation and hardware utilization by sharing infrastructure, are fully isolated, benefit from VM high availability and failover, and are easy to move across data centers and clouds.
So how can we achieve bare-metal-like performance, while retaining benefits of virtualization, all from the same management layer and at a significant cost saving?