hadoop general

- schema on read vs RDBMS schema on write - data flow - splits, split size tends to be HDFS block size to avoid split spanning two nodes which are difficult to data locality data locality. same node -

>>阅读原文<<