-
Hi
Today my boss asked me what high availability does Oracle offer, I mentioned Parallel Server & Advanced Replication. Our enviroment is already using Sun Cluster technology in a data center so I told him that OPS would offer availability + load balancing now he asks me about Replication I went a bit stoned because I think Replication is really for disasters like the whole data center gets blown up and we would have another replication in another city. And I dont see much point of having two server replicating each other in the same building.
Anyone have a suggestion of using OPS or Replication? I would guess OPS is cheaper to setup?
Having a standby database in an environment with Raid 5 would there be any benefit at all since we have redundant disks in our situation so in case of disk corruption we can still recover. Would standby database provide any improvement in case we lose a disk for whatever reason?
By the way since I am not an expert with Solaris does anyone know if Sun Cluster offers load balancing for database servers? (As far as I know it only offers availability)
Cheers
-
Pando
-
OPS will help in terms of load balancing, but it will only help in case of instance failure and as an application failover mechanism. I would either go with Replication or Standby db.
-
Hi,
The UNIX clusters offer availability (I have worked with AIX-HACMP and HPUX-Service Guard) where the second server takes over when the primary goes down. During normal operations you can use the failover server to run some applications thereby distributing some load. But when the primary goes down, the application on the secondary server has to be shutdown and it will takeover all the functions of primary.
Having Raid5 takes care of your disk failures. What do you have? a storage array or just the raid5 with hotspare?
Replication OR standby server is generally used for site disasters. Standby server is a cheaper solution as you move logs and apply them on the remote database. For replication you have many vedors providing disk to disk replication services. For this you need additional software and good network bandwidth as they replicate the disk writes.
Hope this helps.
Baliga
-
OPS should be chosen if you have two major reqts, scalibility and high availibility together.
Replication should not be used if the amount of transaction and number of objects to be replicated is high.Also you should be aware whether current infrastructure cannot support that volume or not.Also Replication may affect application code .
There are certain third vendor good products.Like Shareplex from Quest.
Standby database can also be used as high availability.
Thanks
GP
-
Hmm as I thought OPS would be a good way to imporve our environment. We need both availability and scalability.
Mbaliga I am not sure about storage array although I am pretty sure we are using A5000 from sun (which is a storage array I think)
Replication I think it's expensive to implement and gpsingh you said that this might affect application code? In which way...? By the way we are working as an Application Service Provider so web based and the application are 100% Java (EJB, JSP etc)
Wont consider third party tools if Oracle offers enough options ;)
May be a standby database should be considered for an disaster recovery.
Thanks for the inputs!
-
Take a look at SharePlex....
it's log based replication for the Oracle database...
works on 7.3.4 to 8i... and replicates in real time...
http://www.quest.com/shareplex/
A number of major companies are using it such as HP, Johnson Controls, Sega, AMD, and even Honeywell...
Or call them at 1-800-306-9329
-
what is the advantage of shareplex over advanced replication?
-
You may also get almost the same availability of OPS by having standby database and having Failover agents (say Veritas Cluster Server) running on different machine.
When one machine fails, the other system will take over the responsibility of the primary server. (There will be a down time of few seconds to few minutes, for the second system to take over)
But again, the Data is residing on the same fileserver (say NetApp) which is accessible by both the servers.
In this configuration, only one machine is accessing the DB at anytime.
I guess the advantage here is low "locence" cost.
-
hi pando
for ha solutions there are three things basically
1)ops
2)replication
3)standby server
of these from dba persprective ops and standby servers are the easisese to maintain and replication is the most difficult one relatively speaking.
hth
