Oracle instance & media recovery structure

1、large pool：备份恢复时使用(rman) 若不设置则使用sharedpool
2、表空间备份策略：
        system：备份频度轻，用户数据不要放在system
        temporary：不需要备份
        undo：当作用户数据表空间备份恢复，非常重要
        readonly：每次修改后备份
        highly volatile data:用户表空间，非常重要
        index data:不需备份，可以重建
3、日志文件优先数据文件写入，commit是把redologbuffer写入redolog，数据文件不一定写入
4、checkpoints:把redo相关的数据从dbbuffer写入磁盘 决定从何处开始recovery
      checkpoint position: redolog 的分界线 where recovery starts
      checkpoint queue: 内存中的已被修改的块,增量检查点
5、checkpoint types:
        full： shutdown normal|immediate|transactional, alter system checkpoint
        incremental:periodic writes,only write the oldest blocks
        partial:tablaspace level,alter tablespace backup|offline|normal
6、ckpt process：写 control file 和datafile的文件头
7、instance recovery process:  datafiles out-of-synch->roll forward(redo)->committed and non-committed data in files->roll back(undo)->committed data in files
8、crash recovery: all instance crash | instance recovery: one failed instance in rac
9、tuning the duration cr/ir:
        set init para influence number of redo log and data block
        size the redo log file influence checkpoint frequency
        issue sql statement to init checkpoint
        parallelize ir, recovery_parallelism,fast-start on-demand rollback
10、init para: fast_start_mttr_target 指定ir恢复的秒数后，系统自动进行其他相关参数设置


数据字典
v$sga v$sgastat
v$instance
v$process
v$bgprocess
v$database
v$datafile
v$instance_recovery
v$fast_start_servers
v$fast_start_transactions