Github user Ethanlm commented on the issue: https://github.com/apache/storm/pull/2270 Another Performance testing on ConstSpoutNullBoltTopo. 1. Config: `topology.message.timeout: 300; topology.max.spout.pending: 5000;` 2. 2 VM's as always 3. Launched `4 workers`, `2 spout tasks` and `2 bolt tasks`. ACKing enabled. 4. All experiments ran 300s. 5. For clarity, only show the outputs at 240s. 6. Numbers fluctuate slightly during the experiments. Spout and bolt location like this: ![image](https://user-images.githubusercontent.com/14900612/29385864-eaa6fa28-829e-11e7-97c9-c75b97926ecd.png) #### Experiment 1: No additional network latency Grouping | spout_transferred | spout_acks | spout_throughput (acks/s) | spout_avg_complete_latency(ms) -- | -- | -- | -- | -- LocalityASG | 30810940 | 30810260 | 513515 | 4 LocalOrShuffle(loadaware disabled) | 28185400 | 28182380 | 469756 | 4.7 Shuffle (loadaware disabled) | 30276400 | 30277660 | 504606 | 5 LocalOrShuffle | 27928660 | 27926100 | 465477 | 4.5 Shuffle | 29197460 | 29204280 | 486624 | 4.2 #### Experiment 2: 20ms network latency Grouping | spout_transferred | spout_acks | spout_throughput (acks/s) | spout_avg_complete_latency(ms) -- | -- | -- | -- | -- LocalityASG | 26532600 | 26535020 | 442210 | 3.3 LocalOrShuffle(loadaware disabled) | 13883300 | 13880840 | 231388 | 23.8 Shuffle (loadaware disabled) | 14067160 | 14065380 | 234452 | 22.9 LocalOrShuffle | 13868000 | 13865340 | 231133 | 23.3 Shuffle | 13892060 | 13892460 | 231534 | 23.1 #### Experiment 3: 100ms network latency Grouping | spout_transferred | spout_acks | spout_throughput (acks/s) | spout_avg_complete_latency(ms) -- | -- | -- | -- | -- LocalityASG | 5561480 | 5558060 | 92691 | 26.4 LocalOrShuffle(loadaware disabled) | 2937220 | 2934380 | 48953 | 125 Shuffle (loadaware disabled) | 2935220 | 2934860 | 48920 | 125.1 LocalOrShuffle | 2934980 | 2933620 | 48916 | 125.1 Shuffle | 2934060 | 2935580 | 48901 | 125.4 It's clear to see that `LocalityAwareShuffleGrouping` performs best in this situation. The results are much more clear when there are some network latency.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---