[tahoe-dev] Selecting optimal FEC parameters (also, idea for new peer selection policy)

Zooko O'Whielacronx zookog at gmail.com
Thu Aug 20 21:50:57 UTC 2009


On Wed, Aug 19, 2009 at 3:24 PM, Brian Warner<warner at lothar.com> wrote:
>
> Only if your downstream pipe is faster than the sum of the servers'
> upstream (but only if those servers aren't doing anything else with
> their upstream bandwidth). American DSL/cablemodem lines seem to mostly
> have an 8-to-1 ratio between down and up. So if you assume a homogenous
> mix of nodes, you don't get too much benefit from k>8.

Since we stream the file from all servers in lock-step parallel, your
download bandwidth is equal to the bandwidth of your worst server
times the number of servers (ignoring the possibility of
multi-share-per-server).  Here's one report, where I went from k=4 to
k=13 and was then able to happily stream Master of Puppets (120 KB/s).

http://allmydata.org/pipermail/tahoe-dev/2009-April/001554.html

Regards,

Zooko



More information about the tahoe-dev mailing list