[
https://issues.apache.org/jira/browse/STORM-1731?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15258485#comment-15258485
]
ASF GitHub Bot commented on STORM-1731:
---------------------------------------
Github user HeartSaVioR commented on the pull request:
https://github.com/apache/storm/pull/1362#issuecomment-214815353
I found at throughput 22,000 current 1.x-branch doesn't keep up the SLA at
some moment.
```
uptime: 181 acked: 678,400 acked/sec: 22,613.33 failed: 0 99%:
88,866,815 99.9%: 119,668,735 min: 9,134,080 max: 154,402,815
mean: 18,610,826.60 stddev: 11,680,541.13 user: 110,933 sys: 37,832
gc: 375 mem: 499.06
uptime: 211 acked: 413,040 acked/sec: 13,768.00 failed: 0 99%:
93,126,655 99.9%: 129,564,671 min: 9,093,120 max: 168,820,735
mean: 18,817,505.02 stddev: 12,642,217.74 user: 111,099 sys: 37,859
gc: 377 mem: 432.22
uptime: 241 acked: 678,720 acked/sec: 22,624.00 failed: 0 99%:
102,105,087 99.9%: 189,267,967 min: 9,117,696 max: 237,109,247
mean: 19,483,050.58 stddev: 15,679,529.59 user: 111,690 sys: 38,045
gc: 526 mem: 479.78
```
I guess it cannot keep up throughput higher than 22,000. Keeping up
throughput 20,000 seems fine.
throughput | version | min | max | mean | stddev | 99% | 99.9%
---- | ---- | ---- | ---- | ---- | ---- | ---- | ----
20000 | before patch | 9,150,464 | 140,378,111 | 17,525,115.09 |
6,993,362.18 | 57,769,983 | 101,187,583
So I guess current 1.x-branch can keep up with throughput around 20000 ~
22000. That's it.
I excluded current 1.x-branch and continue testing.
throughput | version | min | max | mean | stddev | 99% | 99.9%
---- | ---- | ---- | ---- | ---- | ---- | ---- | ----
30000 | STORM-1731 | 7,729,152 | 58,359,807 | 15,619,704.64 | 1,665,745.05
| 19,824,639 | 27,738,111
35000 | STORM-1731 | 9,035,776 | 114,032,639 | 17,340,166.53 | 5,343,955.29
| 40,009,727 | 71,368,703
At throughput 35,000 STORM-1731 doesn't keep up the SLA at some moment.
```
uptime: 211 acked: 1,132,000 acked/sec: 37,733.33 failed: 0 99%:
36,470,783 99.9%: 63,438,847 min: 8,089,600 max: 105,316,351
mean: 16,146,505.98 stddev: 4,753,921.20 user: 153,736 sys: 58,653
gc: 653 mem: 707.02
uptime: 241 acked: 737,120 acked/sec: 24,570.67 failed: 0 99%:
34,897,919 99.9%: 67,108,863 min: 8,781,824 max: 114,753,535
mean: 16,564,099.80 stddev: 4,599,850.66 user: 155,660 sys: 58,905
gc: 645 mem: 853.44
uptime: 272 acked: 1,105,560 acked/sec: 35,663.23 failed: 0 99%:
40,009,727 99.9%: 71,368,703 min: 9,035,776 max: 114,032,639
mean: 17,340,166.53 stddev: 5,343,955.29 user: 156,610 sys: 57,884
gc: 664 mem: 678.84
```
So I guess STORM-1731 can keep up with throughput around 30000 ~ 35000.
Btw, let's compare the numbers when they doesn't keep up throughput SLA.
throughput | version | min | max | mean | stddev | 99% | 99.9%
---- | ---- | ---- | ---- | ---- | ---- | ---- | ----
22000 | before patch | 9,134,080 | 157,155,327 | 18,968,494.02 |
13,125,204.89 | 95,551,487 | 127,729,663
35000 | STORM-1731 | 9,035,776 | 114,032,639 | 17,340,166.53 | 5,343,955.29
| 40,009,727 | 71,368,703
STORM-1731 with throughput 35000 still has low latency compared to
1.x-branch with throughput 22000. I'd like to conclude that STORM-1731 ensures
similar or over than 13,000 higher throughput SLA compared to 1.x-branch in my
test environment.
> Avoid looking up debug / backpressure enable flags within critical path
> -----------------------------------------------------------------------
>
> Key: STORM-1731
> URL: https://issues.apache.org/jira/browse/STORM-1731
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-core
> Affects Versions: 1.0.0
> Reporter: Jungtaek Lim
> Assignee: Jungtaek Lim
> Priority: Critical
>
> While profiling the result of STORM-1729, I also found that there're many
> places in critical path which look up the value of flags which are not
> updated dynamically.
> ("get from map" is on top 5 from each spout / bolt thread.)
> This should be fixed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
