Maxperformance vs maxavailability

Question

Hi Gurus,
We are in the process of setting up data guard,
what is the way to measure
1) the percent performance impact caused to primary database due to data guard (maxperformance or maxavailability)
2) Business want to go for maxavailability for better protection for our data without compromising the performance much.
we want to measure the performance impact to primary. how to measure? Any pointers would be helpful.
Thank you.

Answer 1

Zero Data Loss protection is Zero Data Loss protection. There is no compromise, either you want it or you do not. If you want it then it does not matter if there is an impact because you want it. If you care more about performance than zero data loss then you don't want it :^)
The only way to measure the impact is to run a workload without a standby and see how long it takes. Then turn on the standby, let it get synchronized with the Primary and then up the Protection to Maximum Availability (SYNC AFFIRM with Standby redo log files at the standby) and run the workload again. Compare the results. The way SYNC works is as follows.
In 11.1 and before:
User commits
LGWR gets the necessary redo from the Log Buffer (all clients who are ready to commit) and writes it to the Online Redo Log (ORL) file.
LGWR tells the LNS process for the SYNC standby to get and send the same redo.
LNS gets the redo from the Log Buffer, sends it to the standby where it is written to the Standby Redo Log (SRL) file and an ACK is sent back.
LGWR gets the ack and tells the clients that they are committed.
Total time calculation is "ORL I/O + Network Round Trip (RTT) + SRL I/O"
In 11.2 and after:
User commits
LGWR gets the necessary redo from the Log Buffer (all clients who are ready to commit) and writes it to the Online Redo Log file.
In parallel to the above, the NSS process gets the redo from the Log Buffer, sends it to the standby where it is written to the Standby Redo Logfile and an ACK is sent back.
LGWR gets the ack and tells the clients that they are committed.
Total time calculation is "Network Round Trip (RTT) + SRL I/O"
So with a low latency network in 11.2 there should be very little impact.
But read and implement the MAA best practices paper for Network Tuning at http://www.oracle.com/technetwork/database/features/availability/maa-wp-10gr2-dataguardnetworkbestpr-134557.pdf paying attention to set the TCP Window sizes to 10MB or the BDP*3 (from the paper) whichever is larger if you arein 11.1 onwards.
Larry