[rrd-developers] rrd cache work
kevin brintnall
kbrint at rufus.net
Tue Sep 2 23:41:23 CEST 2008
On Tue, Sep 02, 2008 at 11:34:37PM +0200, Tobias Oetiker wrote:
> I think the key here is to document what happens, so if the library
> opens a connection and pools it, this should be explained in the
> docs.
Sure.. If we do that, would it make sense to export rrdc_disconnect to
the other language bindings? In case a long-lived process knew it would
not need the socket cached.
Most of my applications will benefit from the cached connection, but there
may be some that don't.
(Also note, I am not proposing that we cache more than a single socket...
No "pool", per se).
> Not sure how this is applicable here, but an idea I have been toying
> with recently is that the application could throttle itself based on how
> long it takes to send a block of data to the target without throttling
> and then starts backing of from this a little. So whenever the system is
> not able to keep up, the sender will automatically throttle itself a
> little more.
Either way, I think this is an internal implementation detail that we can
add later... no effect on API.
--
kevin brintnall =~ /kbrint at rufus.net/
More information about the rrd-developers
mailing list