Re: trying to be modular

On 12/8/10 11:59 AM, Alan Campin wrote:

And won't the database do the caching already?

On Wed, Dec 8, 2010 at 11:25 AM, Mark Murphy wrote:

The big issue I have with writing my own database caching code is
that I am not keeping my data fresh. If later I update the file,
my internal cache does not get updated, and I can get the wrong
values out. Just saying, you have to be careful with it.

<<SNIP>>

There is little specific and\or explicit caching of database data that I am aware of. An implementation [possibly using triggers to notify of changes] specific to any application would be required, to effect a known caching of row data. The only effective caching I am aware of that is done by the OS, is via the side-effect of the Storage Management and its implementation on the Single Level Storage concept; i.e. when the data from the disk just happens already to be represented in memory when a request for retrieval is made, then the row data is effectively cached, however the database still performed all of the same instructions to retrieve the data that would have been performed even if the data was not yet in memory. While the DB may be tightly coupled with SM for some effects, I can not think of anything special about a simple read of a row that would cause the DB to request of the SM to force the page with that row to remain in memory; i.e. the SM will make that decision independently, based on the activity-level [volatility or read access prevalence] against the data on that page.

Regards, Chuck