Dataguard three modes

Dataguard three modes
Recently we want to create an oracle distributed uard database on the production system. If you have a problem, let's discuss it. In 9ir2, the master database can be placed in the following three modes:
-Maximum protection
-Maximum availability
-In maximum protection, maximum performance ensures that the slave database and master database have the same data, so that zero data loss transactions are committed at the same time on both the master and slave databases. If the slave database goes down or the network is faulty, the master database cannot communicate, and the master database goes down immediately. This method has the highest protection level. However, this mode has a great impact on the performance of the master database and requires high-speed network connections. In maximum availability mode, if the connection to the slave database is normal, the running mode is the same
In maximum protection mode, transactions are committed simultaneously by the master and slave databases. If the slave database loses contact with the master database, the master database is automatically switched to maximum performance mode to ensure maximum availability of the master database. In maximum performance, the master database transfers the archived log to the slave database through the arch process. In this way, the master database has the highest running performance, but cannot guarantee that data is not lost, the lost data is affected by the size of the redo log. When the redo log is too large, one log may not be archived in one day. You can manually switch the log to reduce data loss.
My question is: what kind of mode do you usually choose when you are doing the dataguard database?

At present, China is basically the largest performance model, and other models will crash you

Maximum availability is also quite good. If there is no problem with the network, like maximum protection, if the network is not good and maximum performance is the same

First, maximum protection is generally not used when there is only one standby database. On the one hand, it has a great impact on the performance of the master database, and on the other hand, it must ensure a fast and secure network speed, if the network is disconnected or the standby database fails, it will cause the master database to go down. Although the data can be protected to the maximum extent, it is still insecure. If there are multiple standby databases, consider

Maximum performance; although it has little impact on the performance of the master database, but it does not provide good data protection. 9i generally uses the default log size of 100 mb. If the disk of the master database is not broken, at least 100 MB of log data is lost, which cannot protect data.

Now maximum availability is better. It runs normally in maximum protection. If the network or standby dababase is faulty, it will automatically switch to maximum performance; X, however, during the test, I found that if I read standby only open, the master database will not send logs, and the stanndby database will become invalid.

I use maximum performance mode.

I think that, depending on the production needs, different application needs are different, so our database cannot be down, so we cannot use the protection mode.

I think maximum protection switching is troublesome. It took me a long time to switch over successfully.

Use the third mode! Mature!

The key to this should be the need. The customer must consider the downtime and data loss conditions of existing machines and equipment. No best, only the most suitable

A better way is to use Max performance, but log transfer uses lgwr async noaffrem.
In this way, a redo log is retained at standby to ensure performance. Although data may still be lost during crash, it can be controlled within a certain range ......
However, when lgwr is used, although Max performance is used, when standby is down abnormally, primary hung may still stay for a certain period of time. In my tests, if the standby network is disconnected, primary hung stays for 3 minutes. Although the Hung time can be shortened by modifying the DB and OS parameters, there is still a certain risk.
I finally adopted another compromise, Max performance, and arch optional transmission mode. In addition, a redo log is saved on each storage and local disk to prevent storage crash ......