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 Marcel Reutegger <mr...@adobe.com> on 2012/09/10 12:33:18 UTC
process boundaries
hi all,
while looking through the oak code and also in light of the
recent MongoDB based MicroKernel, I was wondering where
the process boundaries are. right now with the maven build
everything runs in the same process, which is one possible
deployment option. I guess as soon as we have some
alternative MicroKernel implementation that scales well, we
will probably also think about scaling the application on top
of oak. the most obvious solution is to just run the application
multiple times in separate processes and they all connect
to the oak repository.
there's existing information on that topic, like the stack
with the different layers [0], however it doesn't talk about
where the process boundaries are and where we'd need
a protocol (except for the top level protocols to access
the content through JCR or the Oak API directly).
so, the question I have is basically about oak-core. is it
intended and designed to run in multiple processes and
access a single MicroKernel instance? that way an application
running in multiple processes would embed an oak-core,
which talks to the MicroKernen. Or is it rather intended
that oak-core runs in a single process and all clients
connect to that single instance?
regards
marcel
[0] http://wiki.apache.org/jackrabbit/OakComponentStructure
RE: process boundaries
Posted by Marcel Reutegger <mr...@adobe.com>.
hi,
> On Mon, Sep 10, 2012 at 12:33 PM, Marcel Reutegger
> <mr...@adobe.com> wrote:
> > so, the question I have is basically about oak-core. is it
> > intended and designed to run in multiple processes and
> > access a single MicroKernel instance? that way an application
> > running in multiple processes would embed an oak-core,
> > which talks to the MicroKernen. Or is it rather intended
> > that oak-core runs in a single process and all clients
> > connect to that single instance?
>
> IIRC the latter should be the standard client/server deployment.
OK, that was also my impression when I looked through the
code. though, that also means a big piece is missing, right?
currently there's no remoting of the oak-api available.
> applications use the JCR API to access the repository.
>
> the JCR API is exposed by oak-jcr. oak-jcr implements
> the JCR transient space and talks to oak-core for read/write
> and workspace operations (ns reg, node type reg, access control,
> versioning, query etc).
>
> the obvious process boundaries are
>
> 1) between oak-jcr and oak-core (remoting of the oak-core API)
> 2) beween oak-core and oak-mk (remoting of the MicroKernel API)
>
> there's a related jira issue, see [1].
thanks for the link. this actually talks about a concern I share.
depending on where an API is remoted, we have to make sure
communication through that API is minimized to keep the number
of messages low.
regards
marcel
> [1] https://issues.apache.org/jira/browse/OAK-162
>
> >
> > regards
> > marcel
> >
> > [0] http://wiki.apache.org/jackrabbit/OakComponentStructure
Re: process boundaries
Posted by Stefan Guggisberg <st...@gmail.com>.
hi marcel
On Mon, Sep 10, 2012 at 12:33 PM, Marcel Reutegger <mr...@adobe.com> wrote:
> hi all,
>
> while looking through the oak code and also in light of the
> recent MongoDB based MicroKernel, I was wondering where
> the process boundaries are. right now with the maven build
> everything runs in the same process, which is one possible
> deployment option. I guess as soon as we have some
> alternative MicroKernel implementation that scales well, we
> will probably also think about scaling the application on top
> of oak. the most obvious solution is to just run the application
> multiple times in separate processes and they all connect
> to the oak repository.
>
> there's existing information on that topic, like the stack
> with the different layers [0], however it doesn't talk about
> where the process boundaries are and where we'd need
> a protocol (except for the top level protocols to access
> the content through JCR or the Oak API directly).
>
> so, the question I have is basically about oak-core. is it
> intended and designed to run in multiple processes and
> access a single MicroKernel instance? that way an application
> running in multiple processes would embed an oak-core,
> which talks to the MicroKernen. Or is it rather intended
> that oak-core runs in a single process and all clients
> connect to that single instance?
IIRC the latter should be the standard client/server deployment.
applications use the JCR API to access the repository.
the JCR API is exposed by oak-jcr. oak-jcr implements
the JCR transient space and talks to oak-core for read/write
and workspace operations (ns reg, node type reg, access control,
versioning, query etc).
the obvious process boundaries are
1) between oak-jcr and oak-core (remoting of the oak-core API)
2) beween oak-core and oak-mk (remoting of the MicroKernel API)
there's a related jira issue, see [1].
cheers
stefan
[1] https://issues.apache.org/jira/browse/OAK-162
>
> regards
> marcel
>
> [0] http://wiki.apache.org/jackrabbit/OakComponentStructure