OPS/RAC vs. Standby Database

RenatoD
RenatoD used Ask the Experts™
on
Hi all,

I'm trying to weigh the cost and benefits of going with Oracle Parallel Server / Real Application Cluster vs. implementing a Standby DB.  This will also mean an upgrade from Oracle 8.0.5 to 8i or 9i.

I'm leaning toward the Standby DB because it strikes me as much more affordable solution (more bang for the $).  As I see it a Standby DB means saving a bundle of $, and forgoing the seamless failover for resignation to the half-hour or so of downtime that the swith-over to the Standby DB may take.  

If anyone out there has been thru this kind of decision I'd really appreciate your input.

If there are holes in my calculations below, please help me fill them.

HARDWARE:

(OPS\RAC)
Two Servers
Storage Array (RAID)
Hardware required for interface between the two servers and eachother, and the servers and the storage array.

(StandbyDB)
Two Servers

SOFTWARE (& licenses):

(OPS/RAC)
OPS/RAC license from Oracle
Cluster Package from Server Mfgr
Standard Oralce License
Standard Server Support/License

(StandbyDB)
Standard Server Support/License

Thanks.
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Commented:
OPS/RAC makes your filesystem the critical part. If your budget is limited, you might not be able to get a RAID that gives you the required reliabilty and being suitable for RAC. In terms of availabilty OPS/RAC seems to be the best solution otherwise. You have virtually no downtime in case a single machine crashes.

A Standby-DB needs a (short) period of time to get online. If you can live with that it will be cheaper if the machine costs are left out of question - e.g. you have a bunch of boxes that are not used at all, since you need app. twice the CPU-Power compared to a RAC solutuion.

A key benefit of a Standby-DB is that you can use it also as a reporting DB. My recent CTO was very keen on that. Another point is that you can put the standby DB in a different location. Think of 9/11 and you know why...

Commented:
In principle you have to ask one question to yourself: Do you want a High Availability solution or dou you want disaster recovery. This is quite a difference. After you have made that decision, it's quite clear what are the solutions.

DR: Standby (now called Data guard)
RAC: High availability.

With Rac you have two options, actualy using both servers in an active/active configuration or using a failover solution kind of active/passive. Often it's quite difficult to justify a second server only for failover.

By the way storage is also very important for single instance. If the storage can't cope with the workload you also have a problem in single instance. Don't look at the size of the disks, look at the number of i/o's it can provide.
WE tend to use stand by or for more mission critical 24/7 databases the EMC cabinets and fail over. Backups are always taken from the copied database which is also used to populate report server and 'development' or test environments for mission critical premigration testing so that the real run has a high confidence factor switch is cabinet internal or Perl script depending on client availibility. A backup control file to trace and transfer is need for some elements or all of them if different naming conventions are used. The primary tape copies can be offsited for disaster recovery or get a weekly copy and apply transactions via permanent recovery mode this can be enhanced by logfile switching to provide a fairly up to-date copy. Racperf asked the pertinent question what is it for exactly. He also mentioned IO performance which is difficult to manage on raided systems that haven't got tools for 'hotspot' analysis. The other item that desrves real attention is what are you trying to protect against disk failure or program/operation errors. The former is easier than the latter although controllers have been known to smack both the original and mirrored copies. Both items are worrying in parallel environments and I'm not confident enough with the technology yet to put my butt in the firing line. Take a good look at the errors occuring and negative reports before commiting to new technology it's usually better to be conservative and let others do the pathfinding.

Commented:
Hello,

Since this question is still active, I'm wondering if you are still looking for an accepted/acceptable answer?

thanks.
No comment has been added lately, so it's time to clean up this TA.
I will leave a recommendation in the Cleanup topic area that this question is:
[PAQ with csar's comment as answer.]
Please leave any comments here within the next seven days.

PLEASE DO NOT ACCEPT THIS COMMENT AS AN ANSWER!

MathiasMagnusson
EE Cleanup Volunteer

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial