jackrabbit-oak-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Thomas Mueller <muel...@adobe.com>
Subject Re: AW: writing a new DocumentStore implementation
Date Mon, 20 Jan 2014 12:59:54 GMT
Hi,

I think it should be relatively easy to extend the Segment storage to
store to a read from a database. However, I'm not sur if that's currently
planned. 

The philosophy of the Segment storage is quite different from the Mongo
storage (which should probably be renamed to Document storage): the
Segment storage combinies many nodes into a segment and stores larger
blocks, while the Mongo storage keeps each node separately. If you have
multiple cluster nodes, the behavior of the two is quite different: the
Mongo storage detects conflicts very early, but the Segment storage
doesn't. So, in my view, the Segment storage is ideal for standalone
repositories (where you have one process accessing the repository), and
for clusters where the network latency between cluster nodes is high. But
currently not so much for clusters where the latency between cluster nodes
is low, because conflicts are not immediately detected and changes are not
guaranteed to be visible within fixed time bounds.

Regards,
Thomas


On 20/01/14 11:06, "KÖLL Claus" <C.KOELL@TIROL.GV.AT> wrote:

>Hi Thomas,
>
>Sounds good to me !!
>
>One question based on that topic.
>As jukka described with the big-picture the BlobStore we are talking is
>only used with
>mongo-mk as I can see, but what's about the segment-mk.. is it planned to
>have a SQL based 
>Segment Store ?
>
>greets
>claus


Mime
View raw message