RDS Oracle DataPump Import Performance Comparison

 

Test Environment

• Workload: ~1GB schema (200K rows, indexes, packages)
• Test Date: April 7, 2026

Results Summary

Test	Scenario	Duration (sec)	Redo Generated (MB)	Speed Improvement	Redo Reduction
TEST1	Default Import	245	4,349.27	Baseline	Baseline
TEST2	DISABLE_ARCHIVE_LOGGING	125	953.00	49% faster	78% less redo
TEST3	NOARCHIVELOG Mode	111	2,135.92	55% faster	51% less redo

Key Findings

1. DISABLE_ARCHIVE_LOGGING (Test 2) - Best Overall

• Fastest practical option: 125 seconds (49% faster than default)
• Lowest redo generation: 953 MB (78% reduction)
• Why it works: DataPump creates tables with NOLOGGING attribute, minimizing redo for bulk inserts
• Production safe: Database remains in ARCHIVELOG mode with backups enabled

2. NOARCHIVELOG Mode (Test 3) - Marginal Benefit

• Slightly faster: 111 seconds (55% faster than default)
• Still generates significant redo: 2,136 MB (51% reduction)
• Why redo is still high: Redo logs are still written for crash recovery, just not archived
• Not recommended: Requires disabling automated backups, losing point-in-time recovery

3. Default Import (Test 1) - Baseline

• Slowest: 245 seconds
• Highest redo: 4,349 MB
• Full logging: All operations generate redo for archiving

Recommendations

✅ Use DISABLE_ARCHIVE_LOGGING (Test 2 approach)

impdp ... TRANSFORM=DISABLE_ARCHIVE_LOGGING:Y

• Best performance/safety balance
• 78% less redo generation
• Maintains backup protection
• No database restart required