[freenet-support] Freenet 0.7 build 1009

[Matthew Toseland]

Freenet 0.7 build 1009 is now available. Please upgrade. Please test,
and report bugs. This build is an experiment in load limiting. It will
in the short term yield a significant advantage to nodes running it;
what we want to determine is whether this is sustainable once the build
becomes mandatory on december 31.

Technical details: We have an AIMD window, this represents the number of
requests that can be in flight at once of any type. It is increase
additively (Additive Increase) when a request completes without timing
out or reporting an overload, and decreased multiplicatively (MD) when a
request times out. We divide the round trip time (average time it takes
to complete a request) by the window, to get the interval between
sending requests. This is based on TCP/IP. What this build does is it
multiplies the window from above by the number of connected, routable
peers. This is on the basis that TCP assumes a single connection - a
single channel from one host to another host - whereas freenet is
dealing with requests which go all over the network; it seemed to be
worth trying, especially as there seems to be serious underload on the
network at present. So lets see how it goes!
-- next part --
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: 

[freenet-support] Freenet 0.7 build 1009

[Matthew Toseland]

Freenet 0.7 build 1009 is now available. Please upgrade. Please test,
and report bugs. This build is an experiment in load limiting. It will
in the short term yield a significant advantage to nodes running it;
what we want to determine is whether this is sustainable once the build
becomes mandatory on december 31.

Technical details: We have an AIMD window, this represents the number of
requests that can be in flight at once of any type. It is increase
additively (Additive Increase) when a request completes without timing
out or reporting an overload, and decreased multiplicatively (MD) when a
request times out. We divide the round trip time (average time it takes
to complete a request) by the window, to get the interval between
sending requests. This is based on TCP/IP. What this build does is it
multiplies the window from above by the number of connected, routable
peers. This is on the basis that TCP assumes a single connection - a
single channel from one host to another host - whereas freenet is
dealing with requests which go all over the network; it seemed to be
worth trying, especially as there seems to be serious underload on the
network at present. So lets see how it goes!


signature.asc
Description: Digital signature
___
Support mailing list
Support@freenetproject.org
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]