ranger-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Abhay Kulkarni (JIRA)" <j...@apache.org>
Subject [jira] [Assigned] (RANGER-2420) Ranger spends 36% of CPU in ObjectMapper
Date Thu, 09 May 2019 15:23:00 GMT

     [ https://issues.apache.org/jira/browse/RANGER-2420?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Abhay Kulkarni reassigned RANGER-2420:
--------------------------------------

    Assignee: Abhay Kulkarni

> Ranger spends 36% of CPU in ObjectMapper
> ----------------------------------------
>
>                 Key: RANGER-2420
>                 URL: https://issues.apache.org/jira/browse/RANGER-2420
>             Project: Ranger
>          Issue Type: Improvement
>          Components: Ranger
>            Reporter: Todd Lipcon
>            Assignee: Abhay Kulkarni
>            Priority: Major
>
> Ranger uses ObjectMapper to convert to/from JSON. In a profile of a workload (an Impala
authorization test) I see that 36% of the ranger CPU is spent in these functions. 26% of total
CPU is in the '_findRootDeserializer' method, which gets cached if a type is deserialized
multiple times. However, the caching is only effective if the ObjectMapper is reused. JSONUtil
appears to create a new ObjectMapper on every call, which defeats the caching.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message