230 likes | 377 Views
0 KB MB GB TB. 0 KB MB GB TB. 0 KB MB GB TB. 0 KB MB GB TB. 10 -8 10 -4 1 10 4 10 8. 10 -8 10 -4 1 10 4 10 8. Number of jobs. 1.0 PB 0.5 PB 0. Map + Reduce data size. Jobs submitted.
E N D
0 KB MB GB TB 0 KB MB GB TB
0 KB MB GB TB 0 KB MB GB TB 10-8 10-4 1 104 108 10-8 10-4 1 104 108
Number of jobs 1.0 PB 0.5 PB 0 Map + Reduce data size Jobs submitted Jobs submitted Sum data size Sum data size 20 TB 10 TB 0 1.0 PB 0.5 PB 0
Number of jobs 20 TB 10 TB 0 Map + Reduce data size
0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB
0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB
0 KB MB GB TB 0 KB MB GB TB 0 KB MB GB TB
Input data in DFS Shuffle Map 1 Output data in DFS Partition 1 Reduce 1 Partition 1 Map 2 Partition 2 Partition 2 Reduce 2 Partition 3 Map 3 Partition 3 Reduce 3 Partition 4 Map 4 Figure 1. The logical data and processing flow of a MapReduce job.
Figure 2. Hourly job submissions over a week in a Facebook MapReduce workload.