In message <005401c284b6$ec68ca40$4e0d4818@ip78>, Robert Carroll
<[EMAIL PROTECTED]> writes
My IP address are both routeable and on the same subnet. I do not believe
that bandwidth limiting is the major cause. Something else is slowing
fproxy down (for non localhost addresses) because I get the slowdown no
matter how I set the bandwidth limiter. What I would like is for all FCP
and FProxy connections to be treaty EXACTLY as if they are from local host
no matter what. If people need to control access this can be done with a
firewall. ssh tunneling is no solution for windows users, unless someone
out there knows how this could be done in windows.
A solution suggested on this list was to run a proxy programme (?Squid)
on the Freenet server, and, presumably, proxy 8888 to a different port
for the other machines to contact. Sounds as if it might work, but my
server has other little jobs and can't spare resources for another
unnecessary daemon. Would be interested to know if it works though?
----- Original Message -----
From: "Roger Hayter" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Tuesday, November 05, 2002 2:24 AM
Subject: Re: [freenet-support] Tired of my complaining yet? If not, read!
In message <002201c28492$b6cd6180$4e0d4818@ip78>, Robert Carroll
<[EMAIL PROTECTED]> writes
I just want to spout off about my node's performance, or lack of it.
I'm running a permanent node on a server that has other things to do
besides spend every cpu cycle on a java app. Also, like alot of node
operators, I don't use the machine directly, but instead access fproxy
and fcp from another machine. Sometimes it can take SEVERAL minutes
just to call up the web interface! Even when I do finally get through,
alot of content is simply not found, no doubt due to overloaded permanent
nodes. Here are some suggestions for the developers:
As a matter of interest, our your other machines on private or routeable
IP addresses? I gather the bandwidth limits do not now apply to
machines on the same private subnet. This does not help those of us
using routeable IP addresses, and I wondered if the speed of contacting
fproxy was increased in practice. Lynx on the Freenet server certainly
seems to work faster, but it is less than useful for graphics,
obviously.
Transient nodes should be your lowest priority right now. Attention
should be shifted to the permanent nodes because the health of the
network depends on them. That means that CPU utilization needs to be
reduced drastically. Performance for FCP and FProxy need to get the
HIGHEST priority when the node is processing. If that means that other
transactions have to be put on hold in order to process a local
request, then so be it. Permanent node operators should not be
penalized for serving the Freenet community.
See above, we need to know if bandwidth limiting is a major factor in
Fproxy performance.
Do you have the strange problem that my permanent node has, that it
loses all the contacts from its routeing table because it cannot
successfully contact them, although a transient node close by can easily
contact the same nodes > 50% of the time? If there is a solution to
this that does not severely impact on incoming requests (or even if it
does!) it would make running a permanent node less frustrating.
More should be done to encourage people to run permanent nodes.
Requests could be prioritized by each node according to the requesters
responsiveness and availability to that node. Of course some minimum
level of resources need to be dedicated to slower nodes to prevent the
network from fragmenting. To help those who are firewalled or using
NAT (who have trouble receiving inbound connections, but not
establishing outbound connections), nodes with a permanent IP could
also serve as helpers, by listening on behalf of a firewalled or NAT
node. This would only be required for establishing connections, NOT
for acting as a proxy for the data. Bandwidth requirements would be
minor and would allow alot of people to become permanent nodes.
Sounds interesting, the firewalled node would maintain a permanent
connection to the "helper" node?
--
Roger Hayter
_______________________________________________
support mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/support
_______________________________________________
support mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/support
--
Roger Hayter
_______________________________________________
support mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/support
