gluster-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gluster-devel] GLUSTER and Databases


From: Geoff Kassel
Subject: Re: [Gluster-devel] GLUSTER and Databases
Date: Fri, 7 Aug 2009 14:42:15 +1000
User-agent: KMail/1.9.9

PostgreSQL does exist in a more traditional cluster format - search for 
PGCluster and its fork, CyberCluster.

This is synchronous multi-master replication, just like MySQL Cluster, with 
roughly about the same amount of effort involved in setting it up. (No 
setting of triggers/altering database structure required.)

I'm using PGCluster on my servers, and have dabbled with CyberCluster. You 
have to manually recover if you have an extended outage of one master, but 
beyond that it's a bit more transparent for clients to use than MySQL 
Cluster. (i.e. no schema changes required, no need to manually set up 
databases/users on each master, etc.)

It also doesn't have the (quite large) fixed memory overhead of MySQL Cluster, 
which is a big plus in a multi-role machine.

It won't solve performance problems though, as there's a bit of overhead that 
makes a single PostgreSQL server faster for most tasks. (But obviously not as 
reliable or providing to-the-second replication of data.)

What I can recommend in this direction is fast, persistent, distributed 
key-value databases like Redis.

If the key-value pattern isn't a good match for your data model (i.e. you 
require full ACID compliance) it works well as a fast distributed caching 
system to cache database calls and internally generated data.

Cheers,

Geoff.


On Fri, 7 Aug 2009, Gordan Bobic wrote:
> Dan Farrell wrote:
> > On Wed, 05 Aug 2009 11:27:25 +0100
> >
> > Gordan Bobic <address@hidden> wrote:
> >> Either way, it is unlikely that any database performance problems
> >> will be solved by a cluster file system.
> >
> > mysql can cluster, although I don't believe postgreSQL has support
>
> PostgreSQL can cluster, you just have to write your own remote
> connectivity functions and triggers to do the replication exactly how
> you want it. It's much more flexible and powerful, although as you say,
> clustering/replication isn't built in.
>
> Note, however, that I said "cluster file system", not a database cluster.
>
> Gordan
>
>
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel




reply via email to

[Prev in Thread] Current Thread [Next in Thread]