Storage virtualization
Storage virtualization abstracts the internal functions of storage devices or services from hosts and applications, for example by pooling disks or arrays into logical volumes.
Block virtualization maps logical block addresses to backing storage. It can run in a host, a network appliance or a storage system. File-level virtualization operates at a different layer, so block aggregation is an example rather than the whole category. [2]
An exposed capacity is not necessarily a physical reservation. Linux device-mapper thin provisioning allows virtual devices to share a data pool and tracks their mappings separately from the payload. Data capacity and metadata capacity are distinct resources. The kernel's 48 x data-device bytes / data-block bytes formula is metadata sizing guidance, not a guaranteed requirement; the guide rounds estimates below 2 MiB up to 2 MiB, and sharing and snapshot changes affect demand. [3]
Resizing has multiple contracts: enlarging a volume alone does not enlarge its filesystem. With resize2fs, ext2/3/4 shrinking requires an unmounted filesystem, while mounted expansion depends on support. Reducing an LV before its filesystem has been successfully shrunk to fit can destroy data; deleting files alone does not shrink the filesystem. [4] [5]
Relations
- related
- Logical volume, SAN, Block storage, LVM, Volume manager, Thin provisioning
Advantages
- A common storage interface can hide changes in physical placement from applications. [2]
- Thin snapshots can share unchanged blocks instead of immediately copying an entire volume. [3]
Limitations
- Online changes and supported volume layouts depend on the implementation, not on the virtualization label. [2, 4]
- A snapshot retained in the same storage system does not protect against loss of that system. [2]
Citations
- Bunn, Simpson, Peglar and Nagle. Storage Virtualization. SNIA Technical Tutorial; preface August 2003, booklet copyright 2004.
- Thin provisioning, Linux kernel documentation (device-mapper)
- e2fsprogs resize2fs(8): filesystem growth and shrink support
- LVM2 lvreduce(8): filesystem ordering and data-loss warning
Further reading
Sources & caveats
- The SNIA booklet has an August 2003 preface and a 2004 copyright notice. It supplies the layer taxonomy, not current product capabilities or utilization measurements. [2]
- Linux device-mapper and ext2/3/4 document concrete implementations; other filesystems and storage systems need their own resize contracts. [3] [4] [5]
- The older c1 citation is retained for compatibility. Historical utilization figures are not used as a present-day baseline; this entry reports no performance or resize experiment.