Benchmark MySQL's UPSERT

• ON DUPLICATE KEY UPDATE
• REPLACE INTO

CREATE TABLE IF NOT EXISTS test_table (id INT PRIMARY KEY, data VARCHAR(20));

n	ON DUPLICATE KEY UPDATE Time Taken	REPLACE INTO Time Taken
1	2.525ms	1.267ms
10	2.8037ms	17.238ms
100	27.2609ms	139.3891ms
500	105.1636ms	706.9529ms
1000	217.6239ms	1.2950787s
5000	1.235172s	6.8721067s
10000	2.3808408s	12.8781352s
50000	11.9876923s	3m2.1582408s

Reason for Performance Difference

• ON DUPLICATE KEY UPDATE only updates existing rows on conflict, which minimizes locking and reduces index operations.
• REPLACE INTO first deletes any existing row and then inserts a new one, causing additional I/O overhead, index adjustments, and potentially larger locks, which slow down performance, particularly as the data volume grows.

Conclusion

• ON DUPLICATE KEY UPDATE is generally more efficient than REPLACE INTO for large-scale operations, especially when avoiding unnecessary deletions and insertions.