(2014/01/29 15:51), Tom Lane wrote: > KONDO Mitsumasa <kondo.mitsumasa@lab.ntt.co.jp> writes: >> By the way, latest pg_stat_statement might affect performance in Windows system. >> Because it uses fflush() system call every creating new entry in >> pg_stat_statements, and it calls many fread() to warm file cache. > This statement doesn't seem to have much to do with the patch as > committed. There are no fflush calls, and no notion of warming the > file cache either.
Oh, all right.
> We do assume that the OS is smart enough to keep > a frequently-read file in cache ... is Windows too stupid for that?
I don't know about it. But I think Windows cache feature is stupid. It seems to always write/read data to/from disk, nevertheless having large memory... I'd like to know test result on Windows, if we can...
I am quite certain the Windows cache is *not* that stupid, and hasn't been since the Windows 3.1 days.
If you want to make certain, set FILE_ATTRIBUTE_TEMPORARY when the file is opened, then it will work really hard not to write it to disk - harder than most Linux fs'es do AFAIK. This should of course only be done if we don't mind it going away :)
As per port/open.c, pg will set this when O_SHORT_LIVED is specified. But AFAICT, we don't actually use this *anywhere* in the backend? Perhaps we should for this - and also for the pgstat files?
(I may have missed a codepath, only had a minute to do some greping)
