flyovercountry
No user record in our sample, but flyovercountry has activity below (stories or comments). Likely we have partial data — the full bulk-load will fill profiles in.
No user record in our sample, but flyovercountry has activity below (stories or comments). Likely we have partial data — the full bulk-load will fill profiles in.
The file format for hadoop tools is a sequence file. For most queries this is the second slowest format, after text. Rcfile or parquet would be a more interesting benchmark.
HDFS now supports placement groups with one use case being columnar storage.