Previous | Next --- Slide 37 of 43
Back to Lecture Thumbnails

If a node crashes, the computations will be completed by other nodes as if they got assigned those partitions.


It won't take too much time to recompute for failure nodes if the number of such nodes is limited and those nodes can be re-assigned to other nodes and calculate parallel.


The time it will take to recalculate the lost data depends on the number of steps in the pipeline