[GitHub] storm issue #2533: STORM-2912 Revert optimization of sharing tick tuple
Github user HeartSaVioR commented on the issue: https://github.com/apache/storm/pull/2533 I ran FastWordCount with 1.1.1, and it is showing similar latency even from 1.1.1. Complete latency is calculated from acker receiving ACKER_INIT to acker completing tuple tree, hence it contains all the latency: tuple waiting a queue, transferring, etc, and so if the topology is heavily overloaded it can be a bit off. ---
[GitHub] storm issue #2533: STORM-2912 Revert optimization of sharing tick tuple
Github user arunmahadevan commented on the issue: https://github.com/apache/storm/pull/2533 +1 Apparently the topology needs to use tick tuple to see the issue. Ran RollingTopWords which uses tick-tuples and can clearly observe the difference. **Before patch** https://user-images.githubusercontent.com/6792890/35416447-f6a09fb4-01dd-11e8-9d73-d4f736865a8b.png;> **After patch** https://user-images.githubusercontent.com/6792890/35416453-feab19fa-01dd-11e8-95c7-901d4b9bc9bb.png;> ---
[GitHub] storm issue #2533: STORM-2912 Revert optimization of sharing tick tuple
Github user arunmahadevan commented on the issue: https://github.com/apache/storm/pull/2533 Ran the FastWordCount with and without this patch, the number dont look much different. FastWordcount is expected not to see any difference ? Maybe we can wait for Alexandre's results. Storm 1.2.0 RC1 ``` uptime: 30 acked: 8180 avgLatency: 3218.882640586797 acked/sec: 272.7 failed: 0 uptime: 60 acked: 8240 avgLatency: 3296.849514563107 acked/sec: 137.34 failed: 0 uptime: 90 acked: 2380 avgLatency: 478.61344537815125 acked/sec: 26.443 failed: 0 uptime: 121 acked: 3420 avgLatency: 1493.2397660818713 acked/sec: 28.264462809917354 failed: 0 uptime: 151 acked: 3420 avgLatency: 1493.2397660818713 acked/sec: 22.649006622516556 failed: 0 uptime: 181 acked: 9680 avgLatency: 4966.5268595041325 acked/sec: 53.48066298342541 failed: 0 uptime: 211 acked: 9680 avgLatency: 4966.5268595041325 acked/sec: 45.87677725118483 failed: 0 uptime: 241 acked: 2140 avgLatency: 569.7102803738318 acked/sec: 8.879668049792532 failed: 0 uptime: 271 acked: 5620 avgLatency: 2827.6298932384343 acked/sec: 20.7380073800738 failed: 0 uptime: 301 acked: 5620 avgLatency: 2827.6298932384343 acked/sec: 18.67109634551495 failed: 0 ``` With the patch ``` uptime: 30 acked: 3280 avgLatency: 717.0731707317074 acked/sec: 109.33 failed: 0 uptime: 60 acked: 3280 avgLatency: 717.0731707317074 acked/sec: 54.664 failed: 0 uptime: 90 acked: 3020 avgLatency: 1077.8543046357615 acked/sec: 33.56 failed: 0 uptime: 120 acked: 11380 avgLatency: 5864.984182776801 acked/sec: 94.83 failed: 0 uptime: 150 acked: 11380 avgLatency: 5864.984182776801 acked/sec: 75.86 failed: 0 uptime: 180 acked: 5640 avgLatency: 2464.45390070922 acked/sec: 31.332 failed: 0 uptime: 210 acked: 5640 avgLatency: 2464.45390070922 acked/sec: 26.857142857142858 failed: 0 uptime: 240 acked: 10120 avgLatency: 3743.3379446640315 acked/sec: 42.164 failed: 0 uptime: 270 acked: 11600 avgLatency: 5629.23448275862 acked/sec: 42.96296296296296 failed: 0 uptime: 300 acked: 11600 avgLatency: 5629.23448275862 acked/sec: 38.664 failed: 0 ``` ---
[GitHub] storm issue #2533: STORM-2912 Revert optimization of sharing tick tuple
Github user revans2 commented on the issue: https://github.com/apache/storm/pull/2533 The test failure looks unrelated. I have seen it happen before and I think it is related to/fixed by #2534 ---