Own SGA: Each instance has it owns SGA (shared pool, buffer cache, log buffer, large pool, etc)
Background processes: each instance has it owns main background process (dbwr, lgwr, ckpt, smon, etc). RAC has additional BG process like lms, lmon, diag, etc.
Data and control files are shared: Control files are shared between all instances of the same database. Datafiles the same, with the exception of UNDO and TEMP datafiles.
Own redo logs: each instance has it owns online redo logs to save it owns changes, but they are shared for all instances.
Own set of rollback/undo tablespaces: each instance has it owns UNDO tablespace, but they are shared for all instances.
TEMP tablespace is shared for all instance, I would recommend to use tablespace TEMP groups.
Each instance has it owns Password file
I would recommend that each RAC node has it owns ORACLE_HOME for Clusterware and Oracle Database (including ASM)
Parameter file could be shared and stored in ASM
what is the meaning of followings
Cache fusion: RAC needs to be aware of the oracle blocks used by each instance
Guarantees cache coherency: RAC needs to guarantee cache coherency between instances, if not we can have data loss or corruption because of having multiple instances accessing same database.
No disk I/O costs: RAC will try to access all oracle blocks from memory. For example if Instance 2 wants to read some Oracle blocks and RAC detects that they are on instance 1...Instance 2 will ask for that blocks to instance 1 (memory and interconnect I/O, instead of disk I/O)
Background processes: each instance has it owns main background process (dbwr, lgwr, ckpt, smon, etc). RAC has additional BG process like lms, lmon, diag, etc.
Data and control files are shared: Control files are shared between all instances of the same database. Datafiles the same, with the exception of UNDO and TEMP datafiles.
Own redo logs: each instance has it owns online redo logs to save it owns changes, but they are shared for all instances.
Own set of rollback/undo tablespaces: each instance has it owns UNDO tablespace, but they are shared for all instances.
TEMP tablespace is shared for all instance, I would recommend to use tablespace TEMP groups.
Each instance has it owns Password file
I would recommend that each RAC node has it owns ORACLE_HOME for Clusterware and Oracle Database (including ASM)
Parameter file could be shared and stored in ASM
what is the meaning of followings
Cache fusion: RAC needs to be aware of the oracle blocks used by each instance
Guarantees cache coherency: RAC needs to guarantee cache coherency between instances, if not we can have data loss or corruption because of having multiple instances accessing same database.
No disk I/O costs: RAC will try to access all oracle blocks from memory. For example if Instance 2 wants to read some Oracle blocks and RAC detects that they are on instance 1...Instance 2 will ask for that blocks to instance 1 (memory and interconnect I/O, instead of disk I/O)
No comments:
Post a Comment