Basically, 2nd-phase of prepare is slow. It should be random read IO bound to applying transaction log.
I have changed polling period of buf_read_recv_pages(), and increased IO threads.
They may be helpful for the performance of the 2nd-phase when we use based on RAID storage.
Basically, 2nd-phase of prepare is slow. It should be random read IO bound to applying transaction log.
I have changed polling period of buf_read_ recv_pages( ), and increased IO threads.
They may be helpful for the performance of the 2nd-phase when we use based on RAID storage.