Using virtualization can be an effective way of better utilizing the hardware of your machine when using MySQL, or to provide improved security or isolation of different instances of MySQL on the same machine. In some circumstances, virtualization may be a suitable solution for scaling out your database environment by enabling you to easily deploy additional instances of a pre-configured MySQL server and application environment to new virtualization hosts.
With any virtualization solution there is often a tradeoff between the flexibility and ease of deployment and performance, or between the potential performance advantage and complexities of effectively configuring multiple instances of MySQL to reside within a single physical host.
Different issues are experienced according to the virtualization environment you are using. Virtualization generally falls into one of the following categories:
Native virtualization, including products like VMware Workstation, Parallels Desktop/Parallels Workstation, Microsoft Virtual PC and VirtualBox, all work by acting as an application that runs within an existing operating system environment. Recent versions can take advantage of the virtualization extensions in the Intel and AMD CPUs to help improve performance.
The application-based solutions have a number of advantages, including the ability to prioritize CPU usage (including multiple CPUs) and easily run multiple virtualized environments simultaneously.
With these solutions, you also have the ability to easily create a virtualized environment that can be packaged and shared among different virtualization hosts. For example, you can create a MySQL environment and configuration that can be deployed multiple times to help extend an existing scalability or HA environment.
The major disadvantage of this type of virtualization environment is the effect of the host on the performance of the virtualization instances. Disk storage is typically provided by using one or more files on the host OS which are then emulated to provide physical disks within the virtual instance. Other resources on the host are similarly shared, including CPU, network interfaces and additional devices (USB). It is also difficult to directly share lower-level components, such as PCI devices and that the ability to take advantage of RAID storage solutions.
Paravirtualization (Hypervisor), including Xen, Solaris xVM (based on Xen), VMware ESX Server, Windows Server 2008 Hyper-V, and Solaris Logical Domains (LDOM), work by running a specialized version of the host operating system. The host OS then allows slightly modified versions of different operating systems to run within the virtualized environment.
With paravirtualization, the level of performance and the control over the underlying hardware used to support the virtualized environments is higher than native virtualization solutions. For example, using paravirtualization you can dedicate individual CPU cores, RAM, disk drives and even PCI devices to be accessible to individual and specific virtual instances.
For example, within a paravirtualized environment you could dedicate a physical disk drive or subsystem to a particular virtual environment and gain a performance benefit over a typical file-based solution virtual disk.
Operating system-level virtualization, including BSD jails, and Solaris Containers/Zones, offer methods for isolating different instances of an operating system environment while sharing the same hardware environment. Unlike the other virtualization solutions, operating system level virtualization is not normally used to run other operating systems, but instead to provide a level of security isolation and resource control within the core operating environment.
The isolation of these different instances is the key advantage of this type of virtualization. Each virtualized instance sees its environment as if it were completely different system. The solution can be an effective method to provide isolated computing resources for different departments or users, or to provide unique instances for testing and development.
The main reasons for using virtualization, particularly with a database or an application stack that includes a database component, include:
Security — separate instances of different operating systems running within a single host but with effective isolation from each other. When used with MySQL, you can provide an increased level of security between different instances of each server.
Consolidation — merging a number of individual systems with a relatively small load onto a single, larger, server. This can help reduce footprint and energy costs, or make more efficient use of a larger machine. Performance is the main issue with this solution as the load of many MySQL databases running in individual virtual instances on a single machine can be considerable.
Development/QA/Testing — by creating different instances of different environments and operating systems you can test your MySQL-based application in different environments.
Scalability — although using virtualization imposes a performance hit, many virtualization solutions allow you to create a packaged version of an environment, including MySQL and the other application components. By distributing the virtualization environment package to new hosts you can often very quickly scale out by adding new hosts and deploying the virtualized environment.
The remainder of this chapter looks at common issues with using MySQL in a virtualized environment and tips for using MySQL within different virtualization tools.
For advice on common issues and problems, including performance and configuration issues, when using virtualized instances, see Section 14.3.1, “Common Issues with Virtualization”.