Retention Policy

From GridPP Wiki
Jump to: navigation, search

Retention Policy is defined in the [SRM 2.2 Standard]. It is an enum reflecting the "quality" of storage, taking one of three values:

REPLICA

REPLICA - this is the "cheapest", it means storage space of this quality is qualified to hold replicas only: if the space "dies" for some reason, the user (not the site!) is meant to be able to recover the data because it was not the primary copy that died.

OUTPUT

OUTPUT - this is higher quality than REPLICA, referring to storage space that is of sufficient quality to hold job output. Since jobs may run for a long time, it can be relatively expensive to recreate the data, but if the storage area were to die, it can at least be done.

CUSTODIAL

CUSTODIAL - this is the most expensive storage space, meaning that the site will make best efforts to not lose the data. This category is meant for primary copies.

Comments

  • CUSTODIAL is often meant to mean "data is on tape". Data on tape can be lost but the likelihood of this happening is very low. One could construct CUSTODIAL space on disk, e.g., with expensive disks with RAID or other redundancies.