[HADOOP-1007] Names used for map, reduce, and shuffle metrics should be unique

• Type: Improvement
• Status: Closed
• Priority: Trivial
• Resolution: Fixed
• Affects Version/s: 0.12.0
• Fix Version/s: 0.12.0
• Component/s: metrics
• Assignee: Nigel Daley
• Reporter: Nigel Daley
• Created: 09/Feb/07 22:21
• Updated: 02/Mar/07 23:02
• Resolved: 09/Feb/07 23:34

Description

The names used for map, reduce, and shuffle metrics currently overlap:

Map metrics:
input_bytes
input_records
output_bytes
output_records

Recduce metrics:
input_records
output_records

Shuffle metrics:
input_bytes

I propose that the metric names be unique:
map_input_bytes
map_input_records
map_output_bytes
map_output_records
shuffle_input_bytes
reduce_input_records
reduce_output_records