[
https://issues.apache.org/jira/browse/HBASE-13307?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HBASE-13307:
--------------------------
Attachment: gc.png
cpu.png
next.png
Here are graphs of before and after patch applied to branch-1. ~20% more
throughput, less CPU, but more GC.
Here is my scan command (we are NOT returning results):
./hbase/bin/hbase --config ~/conf_hbase
org.apache.hadoop.hbase.PerformanceEvaluation --filterAll --rows=100000
--nomapred scanRange1000 10 &> /tmp/scan/$$.1.log &
Here is how I ran it on 5 different servers:
for i in c2020 c2022 c2023 c2024 c2025; do ssh $i " bash ./bin/scans.sh" ; done
Here is the data I am scanning...:
./hbase/bin/hbase --config ~/conf_hbase
org.apache.hadoop.hbase.PerformanceEvaluation --valueSize=1024 --valueZipf=true
--nomapred=true --rows=10000000 sequentialWrite 10 &> /tmp/scan/$$.1.log &
100M zipfian-sized cells between 0 and 1024 bytes.
Can I get a review? Thanks.
> Making methods under ScannerV2#next smaller allows them to be inlined gaining
> us 7% more throughput
> ---------------------------------------------------------------------------------------------------
>
> Key: HBASE-13307
> URL: https://issues.apache.org/jira/browse/HBASE-13307
> Project: HBase
> Issue Type: Sub-task
> Components: Scanners
> Reporter: stack
> Assignee: stack
> Fix For: 2.0.0, 1.1.0
>
> Attachments: 13307.txt, 13307v2.branch-1.txt, 13307v3.branch-1.txt,
> 13307v4.branch-1.txt, 13307v5.branch-1.txt, 13307v6.branch-1.txt, cpu.png,
> gc.png, next.png, traces.8.svg, traces.9.svg
>
>
> See parent issue for patch and evidence.
> I was looking at graphs of our scan and found that methods were 'too big' to
> be inlined (looking at jvm compilation and inlining output flags -- see
> parent for list). Changing method size helped some. Let me commit the
> resultant patch.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
