This is a pure OLTP Database. The decision makers are under the impression
that having multiple read replicas and spreading the load would solve the
problems they are facing ( Can you please comment on this).
They are considering that AWS provided tools would make the migration
hassle free.
They got a Camaro with a slow driver and thinking that switching cars would
make them go faster.
On Thu, Apr 5, 2018 at 5:55 PM, Mladen Gogala <gogala.mladen@xxxxxxxxx>
wrote:
Comments in-line
On 04/05/2018 08:11 PM, Ravi Teja Bellamkonda wrote:
Hi Mladen,
First of all thanks for the response.
The primary reason they are considering this migration is to attain a
better performing database considering the scaling capabilities of Aurora
would solve all the issues which makes a technological decision not a
business decision.
Better performing doing what? OLTP? Reporting? Data Warehouse? Data Mart?
Oracle, as opposed to various MySQL variants, is completely instrumented
with a very rich set of tools to help you diagnose the performance
problems. So, where is the problem with Oracle? The answer to that
question is crucial for answering your question. What kind of workload is
bringing the Oracle RDBMS with 40 processors and 160 GB RAM to its knees?
And going from Oracle to something else is ALWAYS a business decision, not
a technology decision. If the performance was a problem, a good consultant
could probably solve it for a small fraction of the migration cost. You
will have a very demanding migration project on your hands that is likely
to cost some serious money. Just to copy all the data from Oracle RDBMS -->
Aurora will take some hard manual labour and scripting. Remember, you need
to replicate the table structure, along with all foreign keys, study all
the triggers and see how to replace them and copy all the data in a
consistent manner, all while keeping the original database running. You are
very unlikely to just disconnect the original DB and plough on with Aurora
until the development is done. You are also very likely to need a
heterogeneous replication software, like Golden Gate. We are talking about
a major project and shelling out some real money, before you can even think
of switching to Aurora.
I have done a migration project of migrating Oracle 11.2 --> DB2 9.7 and
it was hard, despite the fact that DB2 is much more feature rich than any
variant of MySQL. It is easier now with DB2 10 and 11 because those
versions can execute PL/SQL natively, but it would still be a major
undertaking. Aurora is an order of magnitude more complex.
My disagreement is about the idea of scaling big would fix all the
problems.
This is what I think about our scenario:
"If Camaro is not fast enough for you, definitely moving to a 18 Wheeler
will not help". I might be completely wrong here.
Well, what it all boils down to is what are you doing with Camaro? What
kind of tasks do you expect? Do you need the vehicle to pick up your kids
and go shopping or do you need a vehicle to do some serious transport?
--
Mladen Gogala
Database Consultant
Tel: (347) 321-1217
