Re: Oracle 10 Failover

On Thu, 11 Sep 2008 11:39:04 +0200, "Chris Seidel" <cseidel_at_arcor.de> wrote:

>joel garry wrote:
>
>> Could you be more specific as to the replication you referred to in
>> the original post? I suspect you may have meant a non-Oracle thing?
>
>Location 1: Primary database
>Location 2: Standby database
>
>Both location connected via WAN (1 Mbit/s), bandwith not exclusively for the
>database.
>
>100 % oracle 10.
>

Bandwith might be insufficient for a standby database, depending on redolog volume.
There is a series of whitepapers on OTN discussing MAA architecture. One of these papers deals with network issues, and has a formula to calculate required bandwith.
For 200 Mb per 15 minutes 2.5 Mb is required. The problem is, I was having only two, and I had this setup

     L1                                                             L2
     P1 --------------------------->                    S1
    S2  <-------------------------                       P2

So I needed 2 times 2.5 Mb

I conducted a test once, and boy I had fun. What I did was

- copied all_objects to a standard table
- ran enough updates to force a log switch
- multiplied the updates by a factor 3
- created a stored procedure with this code
- ran this procedure using dbms_job at both sites simultaneously.

My performance went completely to the dogs, as both sites were waiting on each other. It didn't crash or stop though, only the end-user (me) experienced the database as being in coma.

-- 


Sybrand Bakker
Senior Oracle DBA