Mailing-List: contact common-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Wed, 20 Jul 2016 05:15:21 +0000 (UTC)
From: "He Tianyi (JIRA)" <jira@apache.org>
To: common-issues@hadoop.apache.org
Message-ID: <JIRA.12991002.1468991661000.80211.1468991721531@Atlassian.JIRA>
In-Reply-To: <JIRA.12991002.1468991661000@Atlassian.JIRA>
References: <JIRA.12991002.1468991661000@Atlassian.JIRA> <JIRA.12991002.1468991661743@arcas>
Subject: [jira] [Created] (HADOOP-13390) GC pressure of
 NetworkTopology.isAncestor call can be eliminated
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Wed, 20 Jul 2016 05:15:23 -0000

He Tianyi created HADOOP-13390:
----------------------------------

             Summary: GC pressure of NetworkTopology.isAncestor call can be eliminated
                 Key: HADOOP-13390
                 URL: https://issues.apache.org/jira/browse/HADOOP-13390
             Project: Hadoop Common
          Issue Type: Improvement
          Components: net
    Affects Versions: 2.7.0, 2.6.0, 2.8.0
            Reporter: He Tianyi
            Priority: Minor


{{NetworkTopology.isAncestor}} is called in {{NetworkTopology.getLeaf}}, which is excessively used in block placement policies ({{chooseRandom}} for example).

Currently, the implementation calls {{getPath}} twice. And {{getPath}} performs string concatenation on the fly. On a busy NameNode, this introduces more GC pressure and CPU overhead for block allocation. 
Given that network location and node name does not generally change frequently, we can cache path as a properly of {{Node}} and update accordingly while network location or node name is changing.
Also, one of these {{getPath}} calls in {{getLeaf}} can be eliminated because we are expecting identical result in both calls.


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

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org