Data Erasure codes - Resilience
Erasure codes enables the capability to recover data with less space that with replication. The implementation uses Jerasure library (same as CEPH) and uses the NUM_REPL
environment variable to choose how many servers are used to store ECs.
Activation of the code is done using CMAKE GKFS_ENABLE_EC
.
ECs are only stored/generated with a complete chunk line (num_servers - num_repl).
Additionaly the user can enable the -DGKFS_ENABLE_READ_ERRORS
to inject a
50% of read errors into the workflow.
Edited by Ramon Nou