##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
Instead of storing full copies (x storage overhead), erasure coding splits data into data chunks plus parity chunks. Any chunks can reconstruct the original.
Example: , gives x overhead (vs x for replication) while tolerating failures. Trade-off: reconstruction requires reading chunks and CPU for decoding. Use for cold data where reads are rare.
