You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-dev@jackrabbit.apache.org by Jukka Zitting <ju...@gmail.com> on 2013/11/01 21:28:41 UTC

Optimizations in and around SecureNodeBuilder

Hi,

I spent this afternoon looking at SecureNodeBuilder with the goal to
fix potential bugs and reduce the amount of duplicate work that's
going on.

I explicitly avoided touching the actual permission evaluation as
Angela and Toby are already working on it, but I was still able to
identify and fix a number of issues. See issues from OAK-1139 to
OAK-1142 for details.

The result of these changes is a nice performance boost. We're still
slower than Jackrabbit 2.x for the GetNodeWithAnonymous benchmark, but
the performance difference has dropped from 3x to just above 2x.

# GetNodeWithAnonymous             C     min     10%     50%     90%
  max       N
Oak-Tar (before)                   1      58      59      60      61
  100     991
Oak-Tar (after)                    1      42      42      43      45
  130    1372
Jackrabbit                         1      18      19      19      20
   39    3087

BR,

Jukka Zitting

Re: Optimizations in and around SecureNodeBuilder

Posted by Angela Schreiber <an...@adobe.com>.
hi jukka

that's great... that's for the effort.

one related thing that i planned to look at soon: can't we drop the
SecurityContext altogether in favour of the TreePermission? as far
as i can see there is no real functionality left on the SecurityContext
that would justify that extra wrapper. wdyt?

kind regards
angela 

On 11/1/13 9:28 PM, "Jukka Zitting" <ju...@gmail.com> wrote:

>Hi,
>
>I spent this afternoon looking at SecureNodeBuilder with the goal to
>fix potential bugs and reduce the amount of duplicate work that's
>going on.
>
>I explicitly avoided touching the actual permission evaluation as
>Angela and Toby are already working on it, but I was still able to
>identify and fix a number of issues. See issues from OAK-1139 to
>OAK-1142 for details.
>
>The result of these changes is a nice performance boost. We're still
>slower than Jackrabbit 2.x for the GetNodeWithAnonymous benchmark, but
>the performance difference has dropped from 3x to just above 2x.
>
># GetNodeWithAnonymous             C     min     10%     50%     90%
>  max       N
>Oak-Tar (before)                   1      58      59      60      61
>  100     991
>Oak-Tar (after)                    1      42      42      43      45
>  130    1372
>Jackrabbit                         1      18      19      19      20
>   39    3087
>
>BR,
>
>Jukka Zitting