Re: low-latency multi-master replication reliability?

Comments inline.

Pete

rcordingley_at_my-deja.com wrote:

> In article <384B46F1.351987DB_at_us.oracle.com>,
> Pete Sharman <psharman_at_us.oracle.com> wrote:
>
> I'd love to have the luxury of volume testing the system in advance.
> However, I work for a very small company and we don't have the equipment
> to do it very effectively. That's why I'm looking for somebody who's
> done it before me. 8)

Good theory, but in reality your data load, complexity etc. is going to be unique. You might get some ballpark figures, but that may not be accurate enough.

> As far as log-sniffing vs trigger-based replication, in this case the
> comparison is like oranges and tangerines, I think. Shareplex can do
> multi-master replication in its latest incarnation. On the face of it,
> shareplex seems to be the better choice for load distribution, since it
> relplicates transactions continuously, instead of every "N" units of
> time in a batch.

You mean it's synchronous in nature? You can set up replication in exactly the same way if you can really rely on your network. Alternatively, you can drop the frequency of refresh real low to give you the impression of real time replication. However, most sites I've seen implementing replication don't have a need for such immediacy of data. It could also mean that you're looking at the wrong tool to do what you want. There are other approaches that you can use for load balancing (like OPS for example), but that opens up another kettle of fish.

> The question is, does it work worth a damn?

You expect an Oracle employee to give an answer to that? ;) I'll leave that to others in the group.

>
>
> Robert
> > One of the things you should always do when setting up replication is
> volume
> > test it. I haven't got a configuration to do this with now, so I
> can't
> > comment too much on your first question.
> >
> > On your second question, you're comparing apples with oranges.
> Shareplex
> > (from my understanding, anyway - please correct me if I'm wrong) is a
> > product that uses log based replication rather than trigger based
> > replication, so it's more akin to standby database than advanced
> > replication.
> >
> > cord_at_ragesoft.com wrote:
> >
> > > Can y'all give me any ideas of the reliability of Oracle8i's
> > > multi-master replication when the latency is cranked as low as
> possible?
> > >
> > > I need to have two (or more) databases in which all changes are
> > > replicated as close to realtime as possible. I set up a similar
> system
> > > a few years back and it was horribly unreliable. Of course, that
> was on
> > > Oracle 7.3, so things have hopefully improved dramatically since
> then.
> > >
> > > Also, does anybody have any experience with Shareplex from Quest?
> Is it
> > > better or worse than Oracle8i EE for fast replication?
> > >
> > > Robert Cordingley
> > >
> > > Sent via Deja.com http://www.deja.com/
> > > Before you buy.
> >
> > --------------4471071215613804206EFA91
> > Content-Type: text/x-vcard; charset=us-ascii;
> > name="psharman.vcf"
> > Content-Transfer-Encoding: 7bit
> > Content-Description: Card for Pete Sharman
> > Content-Disposition: attachment;
> > filename="psharman.vcf"
> >
> > begin:vcard
> > n:Sharman;Pete
> > tel;cell:+1.650.868.9969
> > tel;fax:+1.650.633.1506
> > tel;work:+1.650.607.0109
> > x-mozilla-html:FALSE
> > org:Advanced Technology Solutions;Oracle Corporation
> > adr:;;500 Oracle Parkway M/S OPL-A4019;Redwood Shores;CA;94403;USA
> > version:2.1
> > email;internet:psharman_at_us.oracle.com
> > title:Managing Principal Consultant
> > note;quoted-printable:**** The statements and opinions expressed here
> are my **** <br>=0D=0A **** own and do not necessarily represent those
> of **** <br>=0D=0A **** Oracle Corporation. **** <br>=0D=0A
> <br>=0D=0A "Controlling application developers is like herding cats."
> <br>=0D=0A Kevin Loney, ORACLE DBA Handbook <br>=0D=0A "Oh no it's
> not! It's much harder than that!" <br>=0D=0A Bruce Pihlamae, long term
> ORACLE DBA <br>
> > x-mozilla-cpt:;9088
> > fn:Pete Sharman
> > end:vcard
> >
> > --------------4471071215613804206EFA91--
> >
> >
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.

Received on Wed Dec 08 1999 - 10:19:59 CST