Re: [ADMIN] pgbench Windows outperforms Debian in Write tests while Debian is faster in SELECT only tests

> Von: Tom Lane

> Gesendet: Montag, 29. Mai 2017 18:17

>

> OK, so open_datasync and fdatasync are nonfunctional on Windows, which is

> unsurprising.  The other cases are evidently doing *something*, but the amount

> of overhead is variable from one sync method to another, which is likewise

> unsurprising.  The whole reason why we have these options at all is that some

> systems handle some of them better than others.  In this case, I think the

> relevant conclusion is that correct configuration for Debian is to use

> open_datasync or fdatasync, while correct configuration for Windows is to use

> fsync or fsync_writethrough, and when you compare those two configurations,

> yes Windows is slower.

Excellent answer, thanks.

Correct me if I'm wrong: I understand that synchronous_commit = off in combination with fdatasync (Debian) or fsync_writethrough (Windows) is „safe“ (for regular interactive desktop applications).

These options gave the following results:

                         Windows        Debian

40 clients READ/WRITE       5400          7400  

1 client READ/WRITE          400           390

Single client READ/WRITE tests are on now the same level. In the multi client test, Debian is performing significantly better than Windows.

>

> If you're a paranoid sort you would want to do plug-pull testing to ensure that

> your selected setting actually does prevent data corruption during a system

> crash.

That’s certainly an option before we take this into production.

Regards Klaus