Backup of data across network of devices

摘要

A distribution plan that indicates how to encode and distribute backup data across multiple data centers may be generated. The distribution plan may be generated such that one or more characteristics of the distribution plan, such as costs, are optimized while constraints on the plan, such as protection level, recovery point objective (RPO), and recovery time objective (RTO) are satisfied. The protection level may indicate the number of the data centers that are to remain available such that the backup data is recoverable from encoded fragments of the backup data stored in the data centers that remain available.

主权项

1. A system to distribute data across a network, the system comprising:
a plurality of storage units, wherein each of the storage units is included in a corresponding one of a plurality of data centres interconnected via the network; a memory comprising: a parameter component configured to receive a protection level parameter indicative of a number of the data centres that can fail and backup data remains recoverable from encoded fragments stored in the storage units of the data centres that did not fail, wherein the backup data is encoded into the encoded fragments, and a set of the encoded fragments is for each respective one of a plurality of customers; a constraint satisfaction component configured to generate a distribution plan that satisfies a set of constraints that includes the protection level parameter and a plurality of recovery point objectives for the customers, wherein each of the recovery point objectives is for a corresponding one of the customers, the distribution plan comprises a number of the encoded fragments to distribute to each respective one of the data centres, and each of the recovery point objectives indicates an amount of time within which an incremental backup is completed for the corresponding one of the customers; and a backup component configured to distribute the encoded fragments across the data centres according to the distribution plan, wherein the constraint satisfaction component is further configured to: select a time window that is at least as long as the longest of the recovery point objectives of the plurality of customers; divide the time window into time quanta, each one of the time quanta representing a time slot in the time window; and associate each of the recovery point objectives with a corresponding one of the time quanta in which a respective one of the recovery point objectives falls.