[jira] Commented: (TS-43) INK_BLOCK_SIZE is 512 but drives are moving to 4096 as a native size

["John Plevyak (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TS-43?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12797704#action_12797704 ] 

John Plevyak commented on TS-43:
--------------------------------

Simply increasing INK_BLOCK_SIZE to 4K would potentially mean
a great deal of waste for small objects.   While 4K might be fine for
HTTP it might be wasteful for other applications.   An better implementation
would (for example) be to separate out the aggregation write header from
the document header and try to pack objects into 4K chunks and then search
the chunk targeted by dir_offset for the object.  Consideration should also
be taken to ensure memory alignment so that disk transfers do not require
unnecessary copies.

> INK_BLOCK_SIZE is 512 but drives are moving to 4096 as a native size
> --------------------------------------------------------------------
>
>                 Key: TS-43
>                 URL: https://issues.apache.org/jira/browse/TS-43
>             Project: Traffic Server
>          Issue Type: Bug
>          Components: Cache
>         Environment: all
>            Reporter: John Plevyak
>            Assignee: John Plevyak
>   Original Estimate: 72h
>  Remaining Estimate: 72h
>
> INK_BLOCK_SIZE is currently a constant 512, but drives are moving to 4096 as a block size.
> We need to make this a variable or move to 4096 and look into the consequences of such a change.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.