Re: What are the differences between Real Application Clusters, Guard I and Guard II?

> >I have heard rumours of a RAC in Beijing where the nodes are separated by
> >18kms. Quite why anyone would do that, I have no idea. And it could just
be
> >rumour.
>
> I can clearly see the point to have a cluster span over different
builgding,
> even cities.
> If you have RAC on 2 or more servers, they can all go down together
> (remember 2 years ago in NYC?).

I think what Howard meant, was that it would be cheaper to have a replicated or otherwise distributed environment when dealing with such distances, instead of RAC, which works effectively only with low latency and high throughput networks.
OTOH, todays dark fiber networks are quite fast as well.

Also, if you want true reliability, you should keep your redundant systems as independent as possible. That means different databases, independent instances, etc.. RAC isn't that.

> And yes OpenVMS can do it, you can have a cluster (and those are real
> clusters, the 1st and still the best after over 15 years) with up to 96
> nodes spread up to 500 miles apart (I never knew why this limitation in
> distance though but I would assume it is sufficient for most
companies...).
>
> You have one database, and you can split the client load on multiple
servers
> all serving that database. Very scalable, etc, and you don't ever see
> downtime on your database not even for OS upgrades or hardware
maintenance.
> Of course for Oracle upgrades you have downtime, I don't think there's a
way
> around thats, is there?

Yes, there is - have more than one database and have some logic to capture changes when one of the databases is down.

Tanel.

>
>
> --
> Syltrem
>
> OpenVMS 7.3-1 + Oracle 8.1.7.4
> http://pages.infinit.net/syltrem (OpenVMS related web site, en français)
> ---zulu is not in my email address---
>
>
Received on Mon Sep 08 2003 - 12:09:37 CDT