[
https://issues.apache.org/jira/browse/HBASE-12529?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14217707#comment-14217707
]
Matteo Bertozzi commented on HBASE-12529:
-----------------------------------------
attached a patch for 98 the ThreadLocal<Random> is probably good enough
compared with what was there before (see the RandomTest)
{noformat}
#threadId [ #distribution ] #time
Random with 10 threads
4[ 999704 1000114 999765 998028 1000434 1000596 1000814 1001225
1000652 998668] time 20485
2[ 999288 999881 999381 1000100 999296 1000340 1001612 998864
1001001 1000237] time 20792
1[ 999125 1000371 1000566 1002153 1001618 999801 999477 999954
998644 998291] time 20838
3[1001213 998487 1001787 999244 1001494 998837 998483 1001043
1000310 999102] time 20870
0[ 998651 1000391 1000057 998980 1001854 998017 1000166 1002317
998841 1000726] time 20896
9[1000119 1000046 1000611 998970 999313 1001388 1000343 1000947
998077 1000186] time 20891
8[ 998982 1000488 1001040 1001141 999592 1000447 999284 998895
999706 1000425] time 20901
6[ 999270 999654 999056 1001046 1000398 1000642 1000405 1000525
999977 999027] time 20916
5[ 999905 999293 998908 999612 999778 999762 1000451 999766
1000366 1002159] time 20919
7[ 999355 998668 1002233 1000987 1000195 1000012 998565 1001572
999989 998424] time 20914
ThreadLocalRandom with 10 threads
2[ 999405 999806 1000033 1000446 999368 999636 1000624 999784
1000161 1000737] time 204
1[1001876 999712 1000384 999411 999139 999210 999709 999235
1000026 1001298] time 208
7[ 998383 1000379 1001579 1000611 998066 999539 999861 1000250
1001622 999710] time 209
4[ 999751 999360 999987 1000412 998773 998454 1002118 1000389
1001025 999731] time 210
0[ 999830 1001159 1000151 1001600 1000539 999356 998167 999599
1000652 998947] time 211
3[ 000553 998186 999984 999025 1001416 999396 1001011 1000951
998675 1000803] time 212
5[ 999025 1001039 998066 1000585 999958 1001476 998990 999766
1000478 1000617] time 213
8[ 999604 999893 1000388 1002748 997180 1000562 997591 1000215
1001002 1000817] time 199
6[1000654 999660 999590 998480 999547 1000924 1000921 1001984
998402 999838] time 214
9[1000614 998997 1001502 1000113 999398 1000529 1000441 1000216
998574 999616] time 205
ThreadLocal<Random> with 10 threads
5[1000763 1000190 1001990 1001534 998429 999888 1000276 999415
998488 999027] time 337
7[ 999155 1000851 999237 1001077 999979 999698 999991 998073
1001693 1000246] time 337
3[1000241 1000288 999260 999745 999260 1000377 1000135 1000982
1000998 998714] time 340
9[ 999109 999947 1000781 999946 1001335 999319 997880 1000301
1000904 1000478] time 331
6[1000217 1001076 1000677 998441 999550 999291 1000330 1000138
999141 1001139] time 342
2[ 999253 1000411 1000383 1000612 1000754 999804 1000456 999144
999163 1000020] time 344
1[ 999640 998994 1001377 1001255 999724 1000206 1000259 999782
998997 999766] time 344
0[ 998073 999640 999735 1001348 1000081 998673 1000735 1000365
1001249 1000101] time 344
8[1000510 1000444 999805 998874 1000165 1000569 1000459 999686
999178 1000310] time 338
4[1001040 999468 1001355 1001486 999909 1000115 998970 997450
998746 1001461] time 359
{noformat}
> Use ThreadLocalRandom for RandomQueueBalancer
> ---------------------------------------------
>
> Key: HBASE-12529
> URL: https://issues.apache.org/jira/browse/HBASE-12529
> Project: HBase
> Issue Type: Improvement
> Components: IPC/RPC
> Affects Versions: 2.0.0, 0.99.1, 0.98.6.1
> Reporter: Matteo Bertozzi
> Assignee: Matteo Bertozzi
> Priority: Minor
> Fix For: 2.0.0, 0.98.9, 0.99.2
>
> Attachments: HBASE-12529-0.98-v1.patch, HBASE-12529-v0.patch,
> HBASE-12529-v1.patch, RandTest.java
>
>
> looks like the Random in RandomQueueBalancer is now the main cause of
> contention in the rpc queue insertion. we should replace that Random with a
> ThreadLocalRandom.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
