[
https://issues.apache.org/jira/browse/IGNITE-288?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14347044#comment-14347044
]
Sergey Evdokimov edited comment on IGNITE-288 at 3/4/15 3:51 PM:
-----------------------------------------------------------------
This approach increases memory usage, because there are overhead for array,
overhead for integer that describe array content, overhead for instance of
java.lang.Long class for ttl and expireTime (now ttl and expireTime is a 'long'
field).
Also search in bit groups and the array take more time then simple field getter.
was (Author: sevdokimov-gg):
This approach increases memory usage, because there are overhead for array,
overhead for integer that describe array content, overhead for instance of
java.lang.Long class for ttl and expireTime (now ttl and expireTime is a 'long'
field).
> Refactor grid cache entry extras
> --------------------------------
>
> Key: IGNITE-288
> URL: https://issues.apache.org/jira/browse/IGNITE-288
> Project: Ignite
> Issue Type: Task
> Reporter: Sergey Evdokimov
> Assignee: Sergey Evdokimov
> Fix For: sprint-2
>
>
> I think all our extras classes can be replaced with a single class having
> int and array. int will contain 8 of 4bits blocks describing what is stored
> in each array element.
> example (each digit is 4bits block turned to decimal):
> 0000 0001 - ttl and mvcc at index 0
> 0000 0031 - readers at index 1, ttl and mvcc at index 0
> This way we can move:
> rmts from distributed cache entry
> readers from dht cache entry
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
