Difference Between Redo logfiles,Undo Tablespace,Archive logfiles, Datafile

Question

Can some please highlight difference between Redo and Undo logs
Also why do we need seperate archive logs when the data from redo logs can be directly written to Datafiles..
You help will be highly appreciated...

Answer 1

Also why do we need seperate archive logs when the data from redo logs can be directly written to Datafiles..
Redo logs are the transaction logs of yours database , it is exactly a recorder of yours database activity ,by using redo logs i can take my database to a particular state, for further detail lets take an example.
Suppose i have three disk (disk1,disk2 and disk3),i have one datafile in my database which is f1.dbf on my disk1,i have 2 redo log file i.e r1.log and r2.log which are on disk2.At disk3 i copy my database file f1.dbf from disk1 daily at 12:00 PM. On Monday i have successfully copied the f1.dbf file from disk1 to disk3 at 12:00 PM , at 1 PM on monday i insert some record in a table "test".
Insert into test values (1)
It means on monday at 1:00 PM within my datafile f1.dbf i have modified a block and this block contain the record of table test (1 values) , at 1:00 PM if you compare datafile f1.dbf at disk1 from disk3 , you will find only that f1.dbf will contain a new record of test table with values 1 but this will not be within backed up datafile f1.dbf at disk3 . At disk2 r1.log contain those record which i inserted on monday at 1:00 PM, it means i have redo (transactions log) of this datafile f1.dbf within r1.log, i can redo (again inserting) that record from redo log r1.log which i have done at 1:00 PM.
Suppose at 2 PM on monday my disk1 smoked out , if disk1 smoked out f1.dbf has been also smoked out , then what will i do , i will put a new hard disk disk4 and will copy the database file f1.dbf from backup disk3 , this f1.dbf would not have record of "Insert into test values (1)" which occured after backup at 1:00 PM.After coping f1.dbf to new hard disk disk4 i will start reading and applying my redo log file r1.log after 12:00 PM which contain the (redo entry i.e redo vectors) "insert into test values (1)" to my new restored file f1.dbf at new disk4 , in short it is going to replay from recorder of my transactions log (r1.log) and after applying the redo from r1.log i will have exact data before crashing the hard disk1 at 2:00 PM.The only purpose of redo log file in this scenario is to recover the database in case of media failure , if i do not have r1.log then i can only recover my database till 12:00 PM which was copied from disk1 to disk3 but this recovered backup file does not have entry of "insert into test values (1)", i can only bring back my database to before failure current backup at 12:00 PM not just exact before the failure.
You may think why the wastage of three disk for one information , one is at disk1 (f1.dbf),another one is at disk2 (r1.log) and last one is at disk3 for coping f1.dbf from disk1.You may think that it is ok with two disk (disk1 and disk3) why do we need another storage area for r1.log , basically r1.log is the transactions log until between yours current backup to the next backup.It does not contain the whole redo (transaction logs) of yours database but have information (transaction log) till yours next backup which is going to be start on next day tuesday at 12:00 PM ,when it gets filled it is switched to another redo log r2.log, when r2.log is filled then it switched back to r1.log , it overwrites whatever written at 12:00 PM "insert into test values (1)" with new database transactions , for the safety of this overcome this redo is archived and keep save till yours next backup.Redo log switching depneds on yours database transaction pace and its size, the more database activity the quickest redo filling and in turn redo log switching.
Thats all why we need redo and archived log files, as far as the data from redo logs can be directly written to Datafiles is wrong , whenever you do any activity i.e DML , this DML modify the two block one is in within DB buffer cache another one in redo buffer cache , at every 3 seconds or 1/3 filing of redo buffer or commit occurs whichever occur first LGWR write transactions log from redo buffer to online redo log file.
DBWR writes data (dirty buffer) from database buffer cache when checkpoint occurs (checkpointing a process which updates alls datafile headers with the SCN) before Oracle 8.0 it was LGWR which does the job of CKPT to update alls datafile header with SCN , but after Oracle 8.0 CKPT is now responsible to do this activity cause LGWR already keep involved in writing redo buffer data to redo log file.