While backing up to the Oracle Fast Recovery Area (FRA), you determined the backup is taking too long and suspect a performance bottleneck. Which three are true about diagnosing and tuning these…

Question

Options

• AIf an RMAN BACKUP VALIDATE command takes roughly the same time as an actual backup,
• BSetting DBWR_IO_SLAVES to a non zero value can improve backup performance when using
• CIf an RMAN BACKUP VALIDATE command takes noticeably less than an actual backup, then
• DIf an RMAN BACKUP VALIDATE command takes roughly the same time as an actual backup,
• EData files with a high value in V$BACKUP_SYNC_IO.DISCRETE_BYTES_PER_SECOND are a
• FSetting DBWR_IO_SLAVES to a non zero value can improve backup performance when using
• GData files with a high value in V$BACKUP_ASYNC_IO.SHORT_WAITS are a potential performance

• A
  3% (1)
• B
  69% (22)
• E
  3% (1)
• F
  16% (5)
• G
  9% (3)

Explanation

B, C, and D capture the three core truths about RMAN backup performance diagnosis.

Why B is correct: Setting DBWR_IO_SLAVES to a non-zero value helps when the platform uses synchronous I/O - it spawns I/O slave processes to simulate asynchronous behavior, unblocking the backup channel from waiting on each write to complete.

Why C is correct: BACKUP VALIDATE reads data files but writes nothing to the backup destination. If it finishes noticeably faster than a real backup, the read side is fine - the bottleneck is in writing to the FRA.

Why D is correct: Conversely, if BACKUP VALIDATE takes roughly the same time as a full backup, the write side is not the issue - the bottleneck is on the read side (slow data file I/O).

Why the distractors are wrong: Option A shares D's setup but draws the wrong conclusion about where the bottleneck lies. Option F mirrors B but claims DBWR_IO_SLAVES helps with asynchronous I/O - it does not; async I/O already avoids the blocking that the parameter addresses. Option E wrongly flags high V$BACKUP_SYNC_IO.DISCRETE_BYTES_PER_SECOND as a problem - high throughput there means good performance. Option G incorrectly marks high V$BACKUP_ASYNC_IO.SHORT_WAITS as a bottleneck - many short waits means I/Os completed quickly, which is desirable.

Memory tip: Think of BACKUP VALIDATE as a read-only rehearsal - if the rehearsal is as slow as the real show, the stage (disk reads) is the problem; if the real show is much slower, the audience (backup writes) is the bottleneck.

No community discussion yet for this question.