Video narration
Have you ever moved a huge video off a nearly full laptop and wondered: does one computer have to hold the whole file? With HDFS, Hadoop's distributed file system, a cluster-a group of computers-works together to store it. But HDFS splits the file into blocks, smaller pieces, spreads them across machines, and keeps replicas. The NameNode tracks where each block lives; it manages the map, not the file data. DataNodes are the computers that store those blocks. When you need the file, that map tells HDFS where to find its pieces. Replicas are extra copies of blocks on other machines. If one machine fails, copies elsewhere can help keep the file available, though replicas don't protect against every failure. So does one computer have to hold the whole huge file? No. HDFS spreads the file's blocks across machines and keeps copies, so one machine failing need not take the whole file with it.