[
https://issues.apache.org/jira/browse/HBASE-13291?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14383366#comment-14383366
]
stack commented on HBASE-13291:
-------------------------------
bq. Yeah. Was thinking we use the index I build there to avoid the repeated
getKeyValueLen calls.
k.
bq. I'm surprised that that StoreScanner.next() is only in there with 6% and
StoreScanner.peek() not all.
peek is way down... < 1%. SS#next is yeah 6%/8%.
bq. Is there anything different about P/E?
Rows here are ten column and on average 180k so if I do my math right, we are
going at about same speed (I'm slower by some). Maybe I should move to smaller
cells because then changes will be more noticeable?
Issue now is reading the mvcc off the end of the Cell. Its where I am spending
time:
18.70% perf-16373.map [.]
Lorg/apache/hadoop/hbase/io/hfile/HFileReaderV2$ScannerV2;.readMvccVersion in
Lorg/apache/hadoop/hbase/io/hfile/HFileReaderV3$ScannerV3;.readKeyValueLen
I've refactored it some. Will look more. And then SQM#match... and
StoreScanner#optimize
> Lift the scan ceiling
> ---------------------
>
> Key: HBASE-13291
> URL: https://issues.apache.org/jira/browse/HBASE-13291
> Project: HBase
> Issue Type: Improvement
> Components: Scanners
> Affects Versions: 1.0.0
> Reporter: stack
> Assignee: stack
> Attachments: 13291.inlining.txt, Screen Shot 2015-03-26 at 12.12.13
> PM.png, Screen Shot 2015-03-26 at 3.39.33 PM.png, hack_to_bypass_bb.txt,
> nonBBposAndInineMvccVint.txt, q (1).png, traces.7.svg, traces.filterall.svg,
> traces.nofilter.svg, traces.small2.svg, traces.smaller.svg
>
>
> Scanning medium sized rows with multiple concurrent scanners exhibits
> interesting 'ceiling' properties. A server runs at about 6.7k ops a second
> using 450% of possible 1600% of CPUs when 4 clients each with 10 threads
> doing scan 1000 rows. If I add '--filterAll' argument (do not return
> results), then we run at 1450% of possible 1600% possible but we do 8k ops a
> second.
> Let me attach flame graphs for two cases. Unfortunately, there is some
> frustrating dark art going on. Let me try figure it... Filing issue in
> meantime to keep score in.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
