Re: Oracle running on AWS RDS (or SQL Server on AWS RDS)

From: Matthew Zito <matt_at_crackpotideas.com>
Date: Thu, 14 Nov 2013 16:26:18 -0800
Message-ID: <CAJ7936wWO3wEPKHKcVR2VD39OYemWBEE8U4qRfgoBCMs6g9rnw_at_mail.gmail.com>



So - I've never used Amazon RDS in production, but I have a lot of experience with the technology. Effectively, it's just an Amazon EC2 image with the software on it and Amazon managing a lot of the care and feeding - patching, backup and recovery, and so on. The goal is simplicity, not flexibility, so a lot of the tuning and customizations that you can do these days with EC2 go away. You pick your size, set some options around snapshots and replication and click go, and a few minutes later you have an up-and-running oracle/sql server database.

As far as the performance goes, you have to size your hardware appropriately for the workload. One of the nice things about RDS is that it's fairly easy to size your boxes up and down - it does incur a downtime, but you can start with a small or medium and hit a button to convert it to an extra-large high-memory, for example. Behind the scenes, they cleanly shut down your VM, increase the CPU and memory and restart it.

The other performance angle you can use with Oracle, but not with SQL Server, is Provisioned IOPS. It costs a little more, but you can actually define how many IOs per second you want to be able to do to your data volumes. I suspect that your DBA who tried it just did SQL Server on a small instance, which is a fairly wimpy VM - I'd compare it to something you'd use for development, rather than production.

For more flexibility and power, you could always just spin up an EC2 image of your choice and install Oracle directly on there. This gives you the ability to pick from a much broader set of VMs, including some of the newly announced ones with SSDs and extra-fast CPUs, but now you're responsible for all the management of those boxes - no automated backups, snapshots, patching, etc. It also requires that you understand a little bit more about what's going on behind the scenes - expect to get real familiar with EBS volumes, security groups, VPC, and so on.

My instinct is that for the lion's share of tier 2 databases - stuff that's either non-production, or production but non-mission critical - RDS is an excellent option. For really high-end stuff, you're probably better off rolling your own and just using EC2 images, so you can get things exactly the way that you want them.

But it's a very slick offering, and now that they support PostgreSQL, you've got all of the major database platforms aside from DB2 available, which is a nice touch.

Let me know if you have any other questions, Matt

On Wed, Nov 13, 2013 at 12:49 PM, Sandra Becker <sbecker6925_at_gmail.com>wrote:

> We are currently evaluating running a new database on Amazon Web Service
> Relational Database Service. The decision hasn't been made whether to run
> an Oracle or a SQL Server database. Another DBA stood up a SQL Server
> database on AWS, but was unimpressed with the performance.
>
> Does anyone have any experience setting up Oracle on AWS? What are the
> pros/cons? We must run EE 11.2; no other versions are acceptable at this
> time.
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Nov 15 2013 - 01:26:18 CET

Original text of this message