Re: RAC, CFS and online backups

I've no doubt at all that RAC is very much more scalable than poor old OPS. But it certainly isn't linear as certain marketing material (mentioning no names) might suggest.

For an excellent discussion on this very issue, have a read of James Morle's paper on www.oaktable.net. I think it's called 'Unbreakable' (if I weren't so lazy, I'd go and check!).

Regards
HJR "PdV" <piet.de.visser_at_cmg.com> wrote in message news:6de2ae57.0212080907.6c64633e_at_posting.google.com...
> Pete, Howard, Connor, Paul, Group,
>
> Thanks for some Good Information.
>
> My current recommendation for backup RAC would be:
> - cold backup if you can.
> - hot backup by a ksh script, that should generate its own commands,
> and execute them at the same time (e.g. generate just before execution,
> to make sure it is up-2-date).
>
> All this under assumptions:
> - a clusterded file system (do not want to run RAC without it).
> - an indetermined nr of nodes (e.g. expandable).
> - you already have a hot-backup ksh, that generates single-node
> commands (like alter tablespace, cp files, and re-cp files to help
during
> restore).
> - as a conservative Dino, I do not want additional overhead of RMAN.
>
> nb: Looking fw to a thread to discuss RAC in general:
> IMHO the SCALABILITY is its STRONGEST point,
> with failover and HA as a good, but far, second argument.
> HA is best assured by Good, Redundant, hardware.
> Whereas scalability can be done more effectively by adding nodes
> as-needed under a running database.
> Think of a news-site reacting to increased hits by _quickly_ adding the
> development- test- and acceptance nodes to support production load!
>
> Ideas anyone ?
>
> Regards,
>
> PdV
Received on Mon Dec 09 2002 - 01:10:42 CST