Often the biggest consideration is the performance of a virtualized environment once hosted. In most cases, the virtualized environment involves some level of emulation of one or more of the hardware interfaces (CPU, network or disk) of the host environment. The effect is to reduce the effective performance of the virtualized environment compared to running an application natively on the host.
Some core resourcing issues to be aware of include:
Using virtualization does not reduce the amount of CPU required to support a particular application or environment. If your application stack requires 2GB of RAM on an individual machine, the same RAM requirement applies within your virtualized environment. The additional overhead of the virtualization layer and host operating system or environment often mean that you need 2.5GB or 3GB of RAM to run the same application within the virtualized environment.
Configure your virtualization environment with the correct RAM allocation according to your applications needs, and not to maximize the number of virtualized environments that you can execute within the virtualization host.
Virtualization of the CPU resources is more complex. If your MySQL database and application stack do not have a high CPU load, then consolidating multiple environments onto a single host is often more efficient. Keep in mind that at peak times your application and database CPU requirement may need to grow beyond your default allocation.
With some virtualization environments (Xen, Solaris Containers, Solaris LDOMs) you can dedicate CPU or core to a virtual instance. Use this functionality to improve performance for database or application loads that have a high constant CPU requirement, as the performance benefit outweighs the flexibility of dynamic allocation of the CPU resources.
Contention of resources within the host should be taken into account. In a system with high CPU loads, even when dedicating RAM and CPU resources, the I/O channels and interfaces to storage and networking resources may exceed the capacity of the host. Solutions such as Xen and Solaris LDOMs dedicate specific resources to individual virtual instances, but this does not eliminate the effects of the overall load on the host.
If your database application is time sensitive, including logging and real-time database applications, or you are using MySQL Cluster, then the effects of virtualization may severely reduce the performance of your application. Because of the way the virtualized instances are executed and shared between CPUs and the effects of load on other resources, the response times for your database or application may be much higher than normal. This is especially true if you are running a large number of virtualized instances on a single host.
Be aware of the limitation of using a single host to run multiple virtualized instances. In the event of a machine or component failure, the problem affects more than just one database instance. For example, a failure in a storage device could bring down all your virtualized instances. Using a RAID solution that supports fault tolerance (RAID levels 1,3,4,5 or 6) helps protect you from the effects of this.