Re: ORACLE RAC and ORACLE Standby



Gabriel,

In general terms, standby is better for disaster recovery and RAC is
more useful for high-availability and load balancing. Both
configuration provide an ability to write a cold database server backup
with taking any downtime on your primary database instance.

The standby database provides a local or off-site recovery instance.
All database activity (DDL, DML) takes place on the primary instance.
Archived redo logs are transmitted from the primary instance to the
standby instance and applied to the standby instance to replicate the
primary database updates. Although the connection between the two
instances should be fast enough to support timely transmission of the
log files, the functionality of the primary instance is not affected by
a slow connection to the standby. A time delay on the redo log
application to the standby database can provide a good "opps" cover
for the primary (similar capabilities are also provided by flashback
and logminer, depending on your oracle version). If the standby
instance is not open for querying, no additional licensing is required
- the primary instance licensing covers both since only one database
will be open and in use at any time.

RAC is two instances running on two (or more) different machines with
shared or independent storage. Both instances must be licensed because
queries and updates are running in both instances. A high-speed
communication connection is required between the two instances because
the instances are talking to each other constantly. When one database
instance goes down, the "database" stills runs from the other
instance. In 10g a read query can be picked up by the other instance
without interruption. For updates, the connection can fail over to the
second instance, but the transaction must be restarted (your custom
code can detect this error and resubmit the update transaction).

Load balancing capabilities of RAC (beyond normal even loading) include
dedication of one instance for batch processing or one or two instances
to serve one particular application. If one database (datastore or
data warehouse) has a heavy overnight load and a smaller business hour
load while another database (transaction processing) has a heavy
business hour load and is mostly idle overnight, then you may be able
to better utilize your processing power (and licensing costs) by
creating an instance of each database on each server in a RAC.
Overnight your data warehouse database has the cpu capacity of both
servers. Suppose you have purchase enough cpu to run the data
warehouse load jobs overnight (i.e. 4 procs) and then watch that sit
mostly idle during business hours, while you also purchase capacity for
peak business hours loads on your transaction processing database (i.e.
4 procs) that still mostly idle overnight. Instead of 8 procs (4 on
each independent server) a RAC setup with 6 proc (3 proc on each
server), may provide better cycle times at a reduced cost.

Hope this helps,
Margaret

.



Relevant Pages

  • Re: ORACLE RAC and ORACLE Standby
    ... standby is better for disaster recovery and RAC is ... > more useful for high-availability and load balancing. ... > with taking any downtime on your primary database instance. ... > The standby database provides a local or off-site recovery instance. ...
    (comp.databases.oracle.server)
  • DATAGUARD: Primary 2 NODE RAC: Standby Single node Single instance
    ... Primary RAC becomes Standby and standby became Primary. ... Alter database recover managed standby Database disconnect; ...
    (comp.databases.oracle.server)
  • Active Dataguard connections
    ... RAC at our primary data center, and a standby 11g ... Dataguard RAC at our remote data center. ... our primary if it can handle the load. ...
    (comp.databases.oracle.server)
  • log shipping load failure
    ... Select Terminate user an and make sure your database ... secondary set at standby ... >I implemented log shipping but if database is open during ... >load the restore is failing. ...
    (microsoft.public.sqlserver.clustering)
  • Re: Generating one table with a terabyte of data
    ... I didn't know we could split a database file group over multiple HDs. ... > Run perfmon counters to a log and analyze them after a bulk load. ... Another thing is to split the updates across file sets. ... >> loading of daily data in the fastest time. ...
    (microsoft.public.sqlserver.dts)