[
https://issues.apache.org/jira/browse/HBASE-9969?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13823272#comment-13823272
]
Vladimir Rodionov commented on HBASE-9969:
------------------------------------------
{quote}
I just ran StoreScanner with 8 store files and the same test after compaction.
All data is in block cache in both runs. The results I can not explain. Scanner
after compaction is slower: 3.7 sec vs 3.5 sec. The effect of KeyValueHeap
sub-par implementation is probably negligible.
{quote}
Sorry, that was a wrong test. Actual results: 1.9sec before compaction and 1.5
sec after ~ 20% improvement. 2M rows (400M cache size). Full scan time. All
data cached in block cache. The potential win for Loser tree is ~ 20%
currently, but if HBASE-9778 will be resolved and ScanQueryMatcher will be
optimized ...
> Improve KeyValueHeap using loser tree
> -------------------------------------
>
> Key: HBASE-9969
> URL: https://issues.apache.org/jira/browse/HBASE-9969
> Project: HBase
> Issue Type: Improvement
> Components: Performance, regionserver
> Reporter: Chao Shi
> Assignee: Chao Shi
> Fix For: 0.98.0, 0.96.1
>
> Attachments: hbase-9969-v2.patch, hbase-9969.patch, hbase-9969.patch,
> kvheap-benchmark.png, kvheap-benchmark.txt
>
>
> LoserTree is the better data structure than binary heap. It saves half of the
> comparisons on each next(), though the time complexity is on O(logN).
> Currently A scan or get will go through two KeyValueHeaps, one is merging KVs
> read from multiple HFiles in a single store, the other is merging results
> from multiple stores. This patch should improve the both cases whenever CPU
> is the bottleneck (e.g. scan with filter over cached blocks, HBASE-9811).
> All of the optimization work is done in KeyValueHeap and does not change its
> public interfaces. The new code looks more cleaner and simpler to understand.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
