On Fri, Dec 7, 2012 at 3:28 PM, Keith Chew <keith.c...@gmail.com> wrote: > > We can see the times taken to do the syscalls are fast, so what is > causing the delay between the seek and write? I had a look at the > code, and found the method seekAndWrite(), which looks pretty straight > forward. The only time it could be slow is when osWrite() return -1. > Any ideas? > > Please advise if I should still try to get vfs=unix-none to work.
Looks like I have managed to pin point the problem area. Here are the results: ------- Config 1: No latency ------- PRAGMA journal_mode = OFF; PRAGMA locking_mode = EXCLUSIVE; ------- Config 2: Latency observed ------- PRAGMA journal_mode = WAL; PRAGMA locking_mode = EXCLUSIVE; ------- Config 3: Latency observed ------- PRAGMA journal_mode = OFF; PRAGMA locking_mode = NORMAL; ------- >From the above, we can see that: - WAL itself is not the issue, because config 3 shows the latency without WAL - Speed of writing to disk is not the issue, because of config 1 & 3 What causes the latency is the common thing between WAL and locking_mode=normal, and just briefly scanning the code, it looks like they both lock files before writing to disk. I am happy to try out any suggestions to pin point this even further. Any guidance on where I should next look at would be greatly appreciated! I am happy to change or comment out sections of the code just to prove things... Regards Keith _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
