[Linux-cluster] DB Clustering Question

[Chris Darroch]

Hi --

David Teigland wrote:

> I believe Ingres is the only cluster database that's open source.  When I
> looked a few months ago there was some work needed to hook it into our
> cluster/lock managers, but that didn't look too bad as they were already
> able to switch between different clustering/locking infrastrutures.

   That's very interesting -- I should have looked more closely
at Ingres R3!  :-)  Naturally, their site is mostly down now that
I want to look at it.  Seems like they adopted OpenDLM last year.

   I can't quite tell, but if the Ingres Grid Option is their "single
DB clustering" option, it seems to not support things like row-level
locks and "update mode locks".  (The Distributed Option appears to be
a DTP solution for heterogeneous DBs, and the Replicator Option one
for replicating between Ingres DBs, both based on two-phase commits.
It looks like you turn of two-phase commits when using the Replicator
and Grid Options together.)  Errors are mine due to overly quick
scanning of documents.


I wrote:

>    It's perhaps worth noting that PostgreSQL and Oracle face special
> complexities regarding data consistency and locking because they
> provide MVCC (Multi-Version Concurrency Control) ...

   One small pointless correction to my own tangent is that Oracle
actually calls their version MVRC and it works a little differently
than PostgreSQL's, but the rough idea is the same.

Chris.

-- 
GPG Key ID: 366A375B
GPG Key Fingerprint: 485E 5041 17E1 E2BB C263  E4DE C8E3 FA36 366A 375B