ABSTRACT

However, once corruption is detected, the single-server setting does not necessarily allow data recovery, simply because the server failure has caused irrecoverable loss of data. us, RDIC has to be complemented with storing the data redundantly at multiple servers. In this way, the veri er can use RDIC with each server and, upon detecting data corruption at any of the servers, it can use the remaining healthy servers to restore the desired level of redundancy by storing data on a new server. To ensure long-term data reliability in a distributed storage system, aer data is redundantly stored at multiple servers, we can loosely classify the actions of a veri er into two components: prevention and repair. In the prevention component, the veri er uses RDIC protocols to ensure the integrity of the data at the storage servers. In the repair component, which is invoked when data corruption is detected at any of the servers, the client uses data from the healthy servers to restore the desired redundancy level. Over the lifetime of a storage system, the prevention and repair components will alternate.