Mailing-List: contact issues-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Date: Tue, 5 May 2015 19:43:01 +0000 (UTC)
From: "Andrew Purtell (JIRA)" <jira@apache.org>
To: issues@hbase.apache.org
Message-ID: <JIRA.12825498.1430220610000.6808.1430854981089@Atlassian.JIRA>
In-Reply-To: <JIRA.12825498.1430220610000@Atlassian.JIRA>
References: <JIRA.12825498.1430220610000@Atlassian.JIRA>
 <JIRA.12825498.1430220610252@arcas>
Subject: [jira] [Commented] (HBASE-13579) Avoid isCellTTLExpired() for
 NO-TAG cases
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HBASE-13579?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14529130#comment-14529130 ] 

Andrew Purtell commented on HBASE-13579:
----------------------------------------

bq. Minor, unquantified perf improvement. Didn't think 0.98 would be interested.
Ok, thanks. 
I recall reading on one of these JIRAs that vInt parsing of tag length was prominent in some workloads. Let me profile and see if this patch is useful.

> Avoid isCellTTLExpired() for NO-TAG cases
> -----------------------------------------
>
>                 Key: HBASE-13579
>                 URL: https://issues.apache.org/jira/browse/HBASE-13579
>             Project: HBase
>          Issue Type: Sub-task
>          Components: Scanners
>    Affects Versions: 2.0.0
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>         Attachments: HBASE-13579_1.patch, HBASE-13579_2.patch, HBASE-13579_KVExtension.patch, HBASE-13579_branch-1.patch, HBASE-13579_storelevel.patch
>
>
> As observed in this JIRA's performance test, we are always calling the isCellTTLExpired() for every cell and internally it is parsing the keyLength, valueLength() to get the tagsLength after which we decide whether Cell level TTL is present are not.
> This JIRA aims to avoid this check if all the readers of the storescanner knows that there are no tags to read.  Note that, for the memstore scanner we will do that in another JIRA, which I suppose Stack had already raised to avoid tag length while flushing (for the NO-TAG) case.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)