Table 3
Cumulative performance metrics for the 20 Map/Reduce jobs required to run five LS iterations for the initialisation of the photometric system.
| Metric | Map | Reduce | Total |
|---|---|---|---|
| Wall clock time | 26.19 h | 61.22 h | 68.21 h |
| CPU time | 3194.28 h | 4650.61 h | 7844.89 h |
| Number of tasks | 296,047 | 46,005 | 342,052 |
| Data local tasks† | 96.2% | – | – |
| Failed tasks | 141 | 1 | 142 |
| Records I/O* | 7.74 × 1012 | 3.78 × 1012 | – |
| HDFS I/O* | 274.77 TB | 132.42 TB | – |
Notes. A single iteration is composed of job B to E, as described in Sect. 6.1. See the text for the discussion. 
The concept of a data-local task is only meaningful for the Map stage. 
When reporting I/O figures, the Map stage reports the input figure (since the data are read off the distributed file system), and the Reduce stage reports the output figure (since the job results are written to the distributed file system).
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.