[
https://issues.apache.org/jira/browse/HBASE-9778?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13913119#comment-13913119
]
Lars Hofhansl commented on HBASE-9778:
--------------------------------------
Some further observations.
When we reseek for a column we pass a KV that would be located just before the
first KV for that column, in the various scanners, we then seek forward in the
file until we're *past* the KV passed in, then we go back one KV discarding the
current KV. So when we seek forward through the a file we'll scan every KV
twice.
I'm planning to test passing a special KV so that in the scanners can tell when
we're *on* the KV we're looking for. For example when looking for column we can
scan forward until we see the first KV for that row, fam, col, and then we can
stop. No need to need to scan one more, remember the previous, and then go
back. For cases with few versions/columns that should shave off a large portion
of the time. Will report back.
> Avoid seeking to next column in ExplicitColumnTracker when possible
> -------------------------------------------------------------------
>
> Key: HBASE-9778
> URL: https://issues.apache.org/jira/browse/HBASE-9778
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Assignee: Lars Hofhansl
> Attachments: 9778-0.94-v2.txt, 9778-0.94-v3.txt, 9778-0.94-v4.txt,
> 9778-0.94.txt, 9778-trunk-v2.txt, 9778-trunk-v3.txt, 9778-trunk.txt
>
>
> The issue of slow seeking in ExplicitColumnTracker was brought up by
> [~vrodionov] on the dev list.
> My idea here is to avoid the seeking if we know that there aren't many
> versions to skip.
> How do we know? We'll use the column family's VERSIONS setting as a hint. If
> VERSIONS is set to 1 (or maybe some value < 10) we'll avoid the seek and call
> SKIP repeatedly.
> HBASE-9769 has some initial number for this approach:
> Interestingly it depends on which column(s) is (are) selected.
> Some numbers: 4m rows, 5 cols each, 1 cf, 10 bytes values, VERSIONS=1,
> everything filtered at the server with a ValueFilter. Everything measured in
> seconds.
> Without patch:
> ||Wildcard||Col 1||Col 2||Col 4||Col 5||Col 2+4||
> |6.4|8.5|14.3|14.6|11.1|20.3|
> With patch:
> ||Wildcard||Col 1||Col 2||Col 4||Col 5||Col 2+4||
> |6.4|8.4|8.9|9.9|6.4|10.0|
> Variation here was +- 0.2s.
> So with this patch scanning is 2x faster than without in some cases, and
> never slower. No special hint needed, beyond declaring VERSIONS correctly.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
