A volume manager is software within an operating system (OS) that controls capacity allocation for storage arrays. Most OSes provide an option for managing volumes.
A volume manager is more flexible than traditional partitioning methods for storage volumes. Volume managers can stripe together or combine partitions and block devices into bigger, resizable virtual partitions without downtime. This can help with storage performance and capacity planning.
Because storage administrators can partition the capacity of disk drives on an array and pool them as needed, volume management can be considered a form of storage virtualization. Examples of popular commercial volume managers currently on the market include Hitachi Universal Volume Manager, Linux Logical Volume Manager and Veritas Volume Manager.
What volume managers do/what they're used for
Most volume manager deployments have the same design. They start with physical volumes, which can be hard disks, hard disk partitions or logical unit numbers (LUNs) of an external storage device. Physical volumes are treated as if they are made up of physical extents (PEs). Volume managers, such as those in HP-UX and Linux OSes, have uniform PE sizes, whereas others, such as Veritas Volume Manager, have physical extents of variable sizes that administrators can merge and break apart as they please.
Physical extents usually map to logical extents (LEs) on a one-to-one basis. Mirroring allows multiple PEs to map to the same LEs by drawing physical extents from a physical volume group (VG), which is a group of PEs that are all the same size and behave like hard disks in a RAID array. Physical VGs normally live on different disks to provide maximum redundancy.
The volume manager pulls LEs into a VG, which can be joined into virtual disk partitions known as logical volumes (LVs). Logical volumes can function as raw block devices the same way disk partitions do. Logical volumes that have been striped together map each logical extent from a different physical volume, which can improve performance for multiple reads.
A hybrid volume is any volume that uses two physical volumes. A workload might be made up of random seeks to a solid-state drive (SSD) that permanently stores frequently accessed or written data, whereas a higher-capacity spinning disk would store long-term or infrequently accessed data.
Some volume managers also offer snapshots through copy-on-write (COW) to each logical extent. The volume manager copies and preserves the old version of the LV, which can be recalled by laying the COW version on the current logical volume. The volume manager must support both thin provisioning and discard; otherwise, once the LE from the original volume is written, it is stored permanently in the snapshot volume. Making the snapshot smaller than its origin -- which is a common practice -- may break the snapshot.
Logical volumes can fragment when storage devices don't contiguously allocate PEs, which can reduce input/output (I/O) performance on magnetic and spinning disks. But volumes that use all the same size physical extents usually make them all fairly large.