Re: Thoughts on moving app from Oracle to OPS; Caveats, experiences

The main problem is one of contention.

Are you sure you can partition your data so that separate instance do not want to update 'close' data items ?

Since you mention Tuxedo, should we assume you are using XA - there are some problems mixing XA and OPS; especially with multi-threaded server which you would otherwise want to consider with 2,700 user.

Avoid parallel query and large partition views - the combination can be a killer on inter-instance (DLM) lock calls of the 'non-PCM' variety.

You are also quite close to the limit of expansion if you do this, I _think_ 7.3 runs only up to HP-UX 10.20, and the maximum number of OPS nodes for the combination is 4. There is some other combination that goes up to 8 node, but that may only be Oracle 8.

Jonathan Lewis

Clay Isaacs <NOSPAM*asupcsi_at_cyberramp.net> wrote in article <01be1edc$034a1e20$671415ac_at_fusat36_124>...
> Hello,
>
> We are in the process of looking at moving our current OLTP program
from
> Oracle 7.3.2.3 on a single HP K series box to a 3 node OPS cluster
in
> anticipation of a jump in number of users from 150 to 2700.
Assuming the
> application is written well enough to handle the load (we think it
is),
> what caveats, problems, experiences have you had with this kind of
move? I
> have no experience with OPS and have heard it is not all that it
should be.
> Is there anything I should look for in our current DB that could
point to
> possible problems moving over to OPS? Our app uses OCI calls and
Tuxedo in
> the middle.
Received on Thu Dec 03 1998 - 13:32:13 CST