Hierarchical Storage Management (HSM)
HSM define strategies and systems that work with automatic backup of data based on access sample. The idea is to secure data automatic and transparent for users based on rules. Rarely used data are moved to slower media but references stay on the origin system. By user access to moved data data where automatically moved from archive to mainsystem.
HSM is indirect connected to backup. Data are only moved but without high protection. Many backup systems offer extra that deliver this HSM function.
By archiving no copies of data are created but the place of storage changes. Data are moved and removed from origin system. Availability and restore is a main point by archiving. Archiving media are longlife media (e.g. WORM tape in a safe). Is the archiving system defect all data are lost. In case of archiving legal terms like revision security play a role. So data are stored on one time writeable media with high lifetime. By archiving facts like:
- lifetime of media
- readability of media (who can read out magnetic drum storage?)
- data interpretation (who can interprate data from punch card?)
- access to archived data (HSM), fileconversion in current formats
- other points
Archiving is often misunderstood relating to backup systems as outhousing of backup storage media to an external place (bank, another computer center, safe). Archiving is not to confound with backup. But archiving system can be integrated into backup concepts.
Mirroring increased availability of current data. To that, all changes of data are done nearly in time on a mirror system. It is divided into synchronous and asynchronous mirroring:
During synchronous mirroring writing will end if data is on primary and mirror system. This offers highest protection.
Writing process will end if data is written on primary system. Writing process on mirror system is done time delayed. Time delay depends on write load on primary system. Asynchronous mirroring offers a better performance for lower replication band width. But data can be lost in case of failure of primary system during top loads.
Data replication takes place on logical block level (underneath filesystem) or on filesystem level. Redundancy on the block level allows operation in case of failure of complete storage array. But it would not proteect in case of filesystem corruption. Also human mistakes (logic mistakes) can lead to complete failure. That is why it is necessary to provide more versions of one data stock.
Mirroring does not replace backup!