Yes, regarding #1, exactly as I outlined. Yet there is something special about
the number 2 for the number of sets of processes.
Especially for the log shipping variety of transmission, having the archived
redo logs be smaller has value in competition with the extra set of processes.
Also, for extreme throughput of updates, there is a value to owning the log
writer.
For example, if the throughput of change through a single PDB is limited by the
write rate to a single chain of data sync on a machine complex, using two
separate destinations with independently operating pathways from memory to
persistence is not possible within a single container. That may dictate even
more than 2 containers.
So I maintain my position that this is indeed not “Ni!” but rather is a
slippery slope where the most productive position of the stake on the slippery
slope should be the result of measurements.
I am glad you are familiar with Monty Python.
mwf
From: Mladen Gogala [mailto:gogala.mladen@xxxxxxxxx] ;
Sent: Thursday, February 25, 2021 7:48 PM
To: Mark W. Farnham; smishra_97@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: Re: 19c Physical Standby
Well, that sort of defeats the purpose. The container databases were invented
for two reasons:
1. To avoid the overhead of running several separate instances on the same
machine.
2. Because every other manufacturer had that feature. Oracle was in the
bind because they were technologically behind.
I don't have a need for the Holly Hand-Grenade of St. Antioch, I am the DBA who
always says "Ni!".
On 2/25/21 8:22 AM, Mark W. Farnham wrote:
A sledge hammer approach requires asking yourself the question of whether
having two distinct container databases on the complex of machinery in question
is reasonable.
--
Mladen Gogala
Database Consultant
Tel: (347) 321-1217
https://dbwhisperer.wordpress.com
