| Title: | DEC Rdb against the World |
| Moderator: | HERON::GODFRIND |
| Created: | Fri Jun 12 1987 |
| Last Modified: | Thu Feb 23 1995 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 1348 |
| Total number of notes: | 5438 |
I have heard & read that Oracle's implementation of After Image
Journalling does not work correctly in a VAXCluster.
Could someone comment on what its actual behaviour is when one node
dies? Does it depend on WHICH NODE dies, or are they all equal?
And finally, how does all this apply to a LAVC or a new mixed-mode
Cluster?
We want to be sure that we are right to recommended staying away
from a LAVC when Oracle is being used. This is not a new sales
situation, so the decision to go Oracle was made some time ago.
Thanks
Janet
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 138.1 | clustered block... | HSK01::MANNISTO | Olli Mannisto, SWAS/SW Technology, Digital Finland | Tue May 24 1988 15:24 | 20 |
My current understanding is that Oracle does not support AI
in (any) clusters. Oracle says that a future version will.
In a cluster they draw the system saying that each node contains
a kernel making it symmetrical. Oracle uses it own lock mechanisms.
Don't know how it recovers. All nodes have BI journalling.
A cluster is a cluster, independent on how it's composed. Because
of performance reasons I'd personally would think twice before
putting the db on a served disk. A local or a HSC disk is better.
The shorter the trip between the kernel and the disk the better.
You might use also SQL*NET in a cluster meaning that you pick
up a (largest) node to work as a db server. The cluster itself
would serve for system management purposes. Could provide some
licence savings.
-- Olli
| |||||
| 138.2 | AIJ on Kernel Node only | BIOMIC::BLOOMFIELD | Pen is Mightier than the Pencil! | Wed May 25 1988 19:24 | 9 |
I would go along with previous reply in that one approach is to use SQL*NET on satelite nodes in a LAVC and restrict Oracle database activity to the boot node which can then use AIJ (since there is no contention for the database from any other nodes). The satelite nodes can process the SQL*FORMS activity (from local disks if necessary) which, depending on the application, may be significant. This method worked fine on a recent benchmark (see NOTE 137) but number of users was relatively low. Peter | |||||