[tahoe-dev] Perf-related architecture question

[Zooko O'Whielacronx]

On Wed, Jul 21, 2010 at 12:41 PM, Brian Warner <warner at lothar.com> wrote:
>
> I haven't yet decided if the new-downloader will pipeline block reads or
> not

Ticket #1110.

My preference at this point is to get Brian's New Downloader into
trunk as soon as possible, provided:

1. the new design doesn't cause any significant regressions vs. 1.7.1
for any current users

2. some basic performance measurements to verify that the new design
is not slower for some typical download (maybe just someone manually
comparing the Recent Uploads and Downloads reports with 1.7.1 and with
your patches)

3. code review

I don't think putting off #1110 would cause there to be any
regression, so I'm happy to put it off til New Downloader is in trunk.

http://tahoe-lafs.org/trac/tahoe-lafs/ticket/1110# pipeline download
blocks for better performance

> It would bring us back to wanting some better bandwidth-control
> tools, though, so that your home DSL line remains useful for other
> applications while Tahoe is pushing or pulling hard.

That's a good point. #1110 might be a regression in terms of
network-niceness. Maybe it will have to come with a knob to control
the pipeline depth just for that reason, if it doesn't instead come
with some other network-friendiness features (such as the awesome μTP
that Greg Hazel has offered).

Regards,

Zooko