>> It would be also interesting if "reads" exposed in various places (isql,
>> monitoring tables, trace api) are pure physical reads from disk and/or
>> fetched from the OS file system cache.
>>
>> I guess, it could be also from the file system cache, because I've seen
>> situations where execution time is larger for the same number of
>> physical reads for a newly established connection, which then decreases
>> for sub-sequent execution of the same statement with an equal number of
>> reads.
>
>      There is no way to know if OS resolved read request from cache or from 
> disk.

Thought so that Firebird can't differ here. Just speaking out loud that 
if execution time for the same statement is a fraction with the same 
number of physical reads, then data is probably coming from the file 
system cache instead of the disk. A perfect use case for the Trace API 
to detect such things.

> Even if application is used so called "direct IO" read request could be 
> resolved
> from file cache - if some other application reads same file using file cache 
> few
> moments before.


Yup.

Thanks,
Thomas

------------------------------------------------------------------------------
RSA(R) Conference 2012
Save $700 by Nov 18
Register now
http://p.sf.net/sfu/rsa-sfdev2dev1
Firebird-Devel mailing list, web interface at 
https://lists.sourceforge.net/lists/listinfo/firebird-devel

Reply via email to