On 12/7/06, Otis Gospodnetic <[hidden email]> wrote:
> Now that Terracotta JVM clustering has been open-sourced (and works with Lucene's RAMDirectory), who is going to be the first to write something to support HA Solr? :)
I'm very skeptical of the performance of such a solution.
> If I understand the significance of this, this means more or less real-time Solr replication would be possible, eliminating the need for the Master/Slave setup, shapshots, pull scripts, etc. Ja?
It might be OK for very small incremental changes, but I think you
would still need to designate one server a master to coordinate
updates. Larger updates would become network bound, including when it
wasn't necessary (writing out every change to a segment that is going
to be merged anyway).
Maybe someone can do something interesting with it, but I don't think
it can replace our current replication strategy (at first glance).
The problematic part might be the attribution requirements.
Otherwise, the ASF board has determined that MPL binaries are OK for
redistribution as long as they have not been modified (as MPL requires
changes to be made available).
I don't believe into any "real-time" for enterprise. Everything is measured
by response time, which is very good in case of 0.5-3 seconds, and
acceptable for up to 10 seconds...
BEA offers 'real-time' WebLogic, JRockit uses 'determenistic garbage
Indeed, we have 'asynchronous' index for our data via SOLR.
Why real-time??? Why not 'transactional' ;)
From: Otis Gospodnetic
Sent: Thursday, December 07, 2006 1:58 PM
To: [hidden email] Subject: Solr e Terracotta
Now that Terracotta JVM clustering has been open-sourced (and works with
Lucene's RAMDirectory), who is going to be the first to write something to
support HA Solr? :)