expire info in history when using CNFS

[Fabien Tassin]

According to Russ Allbery:
> 
> Fabien Tassin <fta at sofaraway.org> writes:
> > According to Russ Allbery:
> 
> >> You should be able to cut back on your innfeed memory usage quite a
> >> bit.  Mine are 16MB, 6MB, and 6MB.  Break out binaries into a separate
> >> innfeed process and reduce it's max-queue-size; the default of 25 is
> >> way too long for large binaries.  Instead, set the queue size for that
> >> innfeed to about 5.  Also make sure you're using min-queue-connection.
> >> (I wonder if min-queue-connection should even be an option, or if we
> >> should just always turn it on unconditionally.)
> 
> > I'm already doing all of this except that my max-queue-size is much
> > higher than that. Making it lower limits my output rate and I'm then no
> > longer able to full-feed leaf (or slow) peers, even with 25 (the
> > default).
> 
> That doesn't make any sense to me... even five binary articles are already
> going to saturate network buffers, and I don't see why offering fewer at a
> time would reduce your throughput.  What am I missing?

I don't know either but this is something I've observed while playing with
innfeed parameters (number of queues, queue sizes, number of checks and
dynamic methods).

-- 
Fabien Tassin -+- fta at sofaraway.org