What difference does checkpointing makes in log based recovery?

This is more of a big data question(GFS related). I'm not getting how checkpointing help? Can you share 1 example(which could be DBMS example).

Say these are the operations on the database.

Time| Operations

0 start T1

1 read A

2 write B=B+20

3 read B

4 write A=A+40

5 Commit T1

6 start T2

7 write C=C+A

8 Write A=A+B

9 Read B

10 Commit T2

I don't get what checkpointing helps in replaying effort here? Can you clarify me?

---

1 Answers