At 23/08/2002 03:34 PM, you wrote:

>I plan to use InterBase as my data server. Early testing with WebHub using
>the BDE and a Paradox database indicates a slight increase in response time
>when different data are requested on two alternating data queries, but have
>no idea if that approach actually defeated caching. To run this back to
>ground truth requires knowing how caching works, how many copies of data can
>be cached, etc. Better would be to just turn off caching.

Did you try by assigning a database cache of 1 page? The cache size can be
assigned for the server(ibconfig file) or at the database level(using gfix).

>I'm still not certain of the relationship of caching to the "writes" that
>you mentioned below.
>
>If you can point me to a source where caching, synch/asynch writes and
>forced writes are explained I would be grateful.

Take a look here: http://www.ibphoenix.com/ibp_config.html

>Could you also pease confirm if "asynchronous writes" are the same as
>"forced writes".

Forced Writes ON will write the data to disk as soon as there's a
change. Forced Writes OFF, by definition, will write the data to disk when
the system is idle, but under Windows this is performed only when all of
the connections are disconnected. Your database could start with 3MB in
size and insert 3GB and never see the physical size increase until all
connections are disconnected. To my knowledge, the problem doesn't exist
on the other platforms(OS).

>Does Win2k Pro support Forced Writes?

Yes.

>What I get from your message is that
>Forced Writes are equivalent to synchronous writes, and disabling forced
>writes amounts to setting asynchronous writes. If you could connect this
>terminology to the description of caching that I gave above, that would be
>very helpful.

Your analogy is correct.


Daniel Rail
Senior System Engineer
ACCRA Group Inc. (www.accra.ca)
ACCRA Med Software Inc. (www.accramed.ca)