Scalability, Clustering
I've searched Mosix, Dipc, Postgres mailing-lists and used google about the
possibility to cluster ,load balance, Postgresql databases.
Mosix isn't good to clucter Postgresql because of shared memory.
Can be Dipc (http://wallybox.cei.net/dipc/) suitable for this task without
changing postgres' sources (probably only cpu balancing, no data)?
If PostgreSQL Inc. will do a replication server, will be possible?
And Mariposa (http://mariposa.CS.Berkeley.EDU/download.html) ?
My question isn't an academic one, i (probably WE) really need this feature.
thank you in advance for you reply.
valter
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Share information about yourself, create your own public profile at
http://profiles.msn.com.
On Sat, 9 Sep 2000, Valter Mazzola wrote:
If PostgreSQL Inc. will do a replication server, will be possible?
And Mariposa (http://mariposa.CS.Berkeley.EDU/download.html) ?
My question isn't an academic one, i (probably WE) really need this feature.
thank you in advance for you reply.
Depends on what kind of scalability you need. Replication does not
usually equal scalability.
I know that someone was working on a commercial extension to PostgreSQL
to add clustering based on a shared disk system. Basically he was added a
raw storage manager to PostgreSQL plus a lock manager to co-oridinate
access to the shared disk. That way the two nodes could co-ordinate
access to the shared disk. This is very similar to Oracle Parallel
Server.
Replication is a different beast.
Tom
I know that someone was working on a commercial extension to PostgreSQL
to add clustering based on a shared disk system. Basically he was added a
raw storage manager to PostgreSQL plus a lock manager to co-oridinate
access to the shared disk. That way the two nodes could co-ordinate
access to the shared disk. This is very similar to Oracle Parallel
Server.
This is sad. Good Cluster DB design is based on shared nothing architecture
and "function shipping". OPS is known to have a bad and antiquated
architecture
that only works well with extremely well thought out application design.
Andreas
Import Notes
Resolved by subject fallback
Mariposa (http://mariposa.CS.Berkeley.EDU/download.html) has a BSD licence
but it refers to Postgres95.
Mariposa is a patch aganist postgres sources and alpha release, there are a
lot of Papers describing this.
I've compiled under linux but no success.
It's possible that there isn't a solution or a trick to load-balance
postgres (at least cpu load balancing with a central high speed location for
data-bases) ?
thank you for your reply.
valter
From: Zeugswetter Andreas SB <ZeugswetterA@wien.spardat.at>
To: "'Tom Samplonius'" <tom@sdf.com>, Valter Mazzola <txian@hotmail.com>
CC: pgsql-hackers@postgresql.org
Subject: AW: [HACKERS] Scalability, Clustering
Date: Mon, 11 Sep 2000 10:41:50 +0200I know that someone was working on a commercial extension to
PostgreSQL
to add clustering based on a shared disk system. Basically he was added
a
raw storage manager to PostgreSQL plus a lock manager to co-oridinate
access to the shared disk. That way the two nodes could co-ordinate
access to the shared disk. This is very similar to Oracle ParallelServer.
This is sad. Good Cluster DB design is based on shared nothing architecture
and "function shipping". OPS is known to have a bad and antiquated
architecture
that only works well with extremely well thought out application design.Andreas
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Share information about yourself, create your own public profile at
http://profiles.msn.com.
Import Notes
Resolved by subject fallback