Replication and Erasure Coding Explained
Abstract: Unlike what most storage vendors will try to make you believe, data loss in storage systems is unavoidable. It is right up there with death and taxes. However, by selecting the right solutions and applying the appropriate parameters, we can control to a large extent both the probability of data loss happening, as well as the size of the data loss when it actually does happen. This paper provides a qualitative, straightforward view into the specifics of this process as related to the Caringo object storage software, Swarm.