MapReduce for Dummies Say you need to grep for "Hi There" across a petabyte of text spread over thousands of files. A single machine reading that much data sequentially could take hours — I/O alone becomes the bottleneck, before you even count the CPU time to scan every line. The fix: split the work. Hand different files to different machines (workers) so they scan in parallel, and use a central coordinator to hand out tasks, track progress, and reassign anything that fails. That's Map...