[
https://issues.apache.org/jira/browse/HBASE-17877?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15961794#comment-15961794
]
Vikas Vishwakarma commented on HBASE-17877:
-------------------------------------------
I did some tests with JMH. So overall it looks like the hadoop version gives
better throughput with random byte arrays, but it hits a worst case when two
the two byte arrays are almost similar and have just 1 byte diff like say last
byte in which case it gives 30-40% lower throughput. I will do some more
experiments with the guava version also. The below result table is with Random
byte arrays. I have used 20 warmup and 20 measurement cycles of 1 sec each. I
will leave a longer running test for the weekend.
|Byte Array Size|byte array diff index|Mean HBase (ops/ms)|Hadoop (ops/ms)
|%diff|
|----|----|----|----|----|
|Random|4|440977.678|439558.547|0|
|Random|8|437781.625|439089.09|0|
|Random|16|441573.466|439052.733|-1 |
|Random|20|433731.957|449067.897|4 |
|Random|32|441486.237|448821.457|2 |
|Random|50|432720.686|447027.454|3 |
|Random|64|436713.574|440450.684|1 |
|Random|100|430155.538|439399.306|2 |
|Random|128|441413.074|440321.381|0 |
|Random|200|450140.537|448164.678|0 |
|Random|256|376015.788|457719.989|22 |
|Random|512|437394.161|438975.893|0 |
|Random|1024|442672.023|440269.773|-1 |
|Random|2048|441500.328|439325.611|0 |
|Random|4096|428583.365|439881.571|3 |
|Random|8192|449889.732|439456.918|-2 |
|Random|16384|439970.375|439271.189|0 |
> Replace/improve HBase's byte[] comparator
> -----------------------------------------
>
> Key: HBASE-17877
> URL: https://issues.apache.org/jira/browse/HBASE-17877
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: Vikas Vishwakarma
> Attachments: 17877-1.2.patch, 17877-v2-1.3.patch,
> ByteComparatorJiraHBASE-17877.pdf
>
>
> [~vik.karma] did some extensive tests and found that Hadoop's version is
> faster - dramatically faster in some cases.
> Patch forthcoming.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
