Gnutella Forums - ultrapeer problem.

Gnutella Forums (https://www.gnutellaforums.com/)

- LimeWire Beta Archives (https://www.gnutellaforums.com/limewire-beta-archives/)

- - ultrapeer problem. (https://www.gnutellaforums.com/limewire-beta-archives/6292-ultrapeer-problem.html)

ultrapeer problem.

i configured limewire to have 10 connections up, everything is fine until it find another ultranode. as soon as it connects to that ultranode it disconnect from all other hosts. and that those ultranodes usaly have only 1-3 other hosts connected...
i'm running WinXP..

I experienced the same. - I guess the reason is, that there aren't enough ultra-peers yet, so the connection betwixt the different ultra-peers is almost none existent yet.

by design (I think)

I think it is designed to be this way.

When you connect to an Ultranode, you automaticly drop all other connections, since they are not needed.

Now, the ultranode you connected to, was probably confgiured to connect to only 3 other hosts.

I do not think this has anything to do with how many Ultranodes are out there.

hmm..could be...
ok, but i'm acting as ultranode too and i should see others connect to me..
and shouldn't limewire how many clients can that ultranode see?

Slight problem with the ultrapeer functionality.. When connecting normally it drops the other peer connections when it finds an ultrapeer.. But, if that ultrapeer should go down and it tries to connect to normal peers again, it only seems to try to connect to 1 at a time instead of however many are specified. The only way I could get around it was to disconnect and reconnect.

DubStar: havnt had that problem yet so I dont know what could be causing that...

But the Ultrapeer functionality is acting a little weird. I'll connect and maybe at first it'll connect to the max num of regular peers, ok fine. Then after a few mins, it drops and etc it'll find an Ultrapeer to connect to. So it drops all other connections and stays connected to the Ultrapeer. Then everything is cool and it'll drop the Ultrapeer connection and repeat the process I just described (I can actually sit and watch it do this in 2 minutes time atleast 2 times). Not really sure what is going on. Another thing is when I am connected to a Ultrapeer and execute a search the Input on the connection is like ~5-9 KB/s until I pretty much get all of the results back. Is the throughput suppose to be that high? I mean thats even when it indicates it has 1 hosts (assuming 1 other Ultrapeer connected to it?). I was thinking maybe it dropped the connection b/c of bandwidth usage b/c you guys implemented the new upload options and they were not optimize correctly. I have cable access and 6 sim. uploads going @ >4KB/s lags my ***. So it drops connections and gets everything in a mess. But anyways after that first search it drops the connection and repeats with the connection situation I was talking about in the above. But as far as not searching for anything the I/O is very low ~0.05 KB/s at max. So what are your ideas Adam?

I should point out it doesnt drop connections continuously, just more than it should.

What is the significance of the "Hosts" count when you're connected to an Ultrapeer? Does that mean how many Ultrapeers you're connected to?

I'm going to break down what Adam is saying for the people who really dont know what he is talking about. (I'm trying to be helpful not sarcastic.)

Quote:

Originally posted by afisk
When you connect to an UltraPeer, you do automatically drop the rest of your connections. This is the desired behavior, as the peer is acting as your proxy and is shielding you from all of the traffic on the network, except for queries for files that you have, which are forwarded to you.

This means you are connected to ONE peer that handles queries that would overwise pass through and consume network bandwidth. That peer caches what you are sharing so when someone searches for something (sends a query our on the network) that would go to that ONE peer and stop and if there is a match of something you are sharing it sends the information back so that host can see that you're sharing that file. Then of course after that he can connect to you and download the file.

Quote:

Also, I should mention that being a client peer or an UltraPeer does not necessarily reflect the power of your machine or your bandwidth -- most users who have plenty powerul enough machines and plenty of bandwidth will remain client peers most of the time.

self-explanatory.

Quote:

I admit that the switch to becoming a client peer seems a little bit wierd at first, as your connectivity (and hence network reach) appears to have dropped. Once there are more UltraPeers on the network, however, it should not matter at all. This is because each forwarding of a query to another layer out on the network will become much more significant, as each "hop" will effectively query a larger number of hosts/files than before. This is because a hop to an UltraPeer could mean searching through 50 hosts connected directly to that UltraPeer, which means your searching far more files that a hop to an old peer who might only have 4 hosts connected to it, and therefore far less files.

This means now with Ultrapeers the network will look like this.

XXXX
XXXX (Host X = Ultrapeer)
XXXX
/ | \
A B C (Hosts A, B, and C = Clientpeers)

meaning one Ultrapeer could have 50 hosts connected to it instead of the old design

A B
\ /
/ \
C D

where the max number of hosts connected are 4 and they produce X amount of traffic for a specifc query. As where when connected to an Ultrapeer it produces the same X amount of traffic with 100 times more clients connected to it for a specific query.

So in other simplier terms when connecting to an Ultrapeer its like that ONE peer is sharing all files of those 50 hosts instead of 50 individual hosts sharing their own files. Because you query Host Z for X and Host Z replies telling you he has X available for download. So you query Ultrapeer and it replies telling you Host Z has X available for download. Essentially cutting down extremely on Gnutella network traffic. Hope that paints a picture.

Quote:

I hope that makes some sense. So, there is a timing issue here. As more LimeWire 1.9 and aboves get out there, you should see better search results even as the total messages past are far fewer than before.

Gnutella is really going to kick *** when LimeWire 1.9 is released full stream and Bearshare does the same (I think Vinnie was or is going to implement Supernodes or at least the new "Query Routing Protocol")

One last question....Adam what is the LimeWire's comments on the Network History Size graph on LimeWire.com? About its up and downs?

oh yeah and I love networking (just felt like saying that :P).