[ https://issues.apache.org/jira/browse/HBASE-9969?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13826388#comment-13826388 ]
Matt Corgan commented on HBASE-9969: ------------------------------------ Not sure how i didn't see it before, but here is the optimization that avoids sending the scanner through the heap when the next KV is from the same scanner: KeyValueHeap.pollRealKV() : line 350 {code} // Compare the current scanner to the next scanner. We try to avoid // putting the current one back into the heap if possible. KeyValue nextKV = nextEarliestScanner.peek(); if (nextKV == null || comparator.compare(curKV, nextKV) < 0) { // We already have the scanner with the earliest KV, so return it. return kvScanner; } {code} Does the LoserTreeKeyValueHeap do a check like that? > 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: 9969-0.94.txt, KeyValueHeapBenchmark_v1.ods, > hbase-9969-pq-v1.patch, hbase-9969-v2.patch, hbase-9969-v3.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)