SixXS::Sunset 2017-06-06

TCP losses
[de] Shadow Hawkins on Thursday, 14 October 2010 12:51:23
Hi, I use aiccu with ayiya Tunnel on Win7/64bit. When I download from newsgroups I experience a high amount of Upload from aiccu.exe. When I checked with wireshark I got loads of: TCP: Previous segment lost (common at capture start) TCP: Fast retransmission (suspected) TCP: Zero window TCP: Out-Of-Order segment Friend suggested updating system time. I did so but no change. Any ideas? Is this common for ayiya? Could port-forwarding help? I did none so far as the tunnel works apparently fine.
TCP losses
[ch] Jeroen Massar SixXS Staff on Thursday, 14 October 2010 13:13:47
AYIYA does not use TCP, it runs over UDP. Can you provide the wireshark packet capture in a file?
TCP losses
[de] Shadow Hawkins on Thursday, 14 October 2010 14:22:45
file I hope I didn't offer sensible information. I cut the file after 1MB.
TCP losses
[de] Shadow Hawkins on Thursday, 14 October 2010 14:25:35
TCP losses
[ch] Jeroen Massar SixXS Staff on Thursday, 14 October 2010 15:12:45
I don't see anything weird in there. As for having a lot of 'upload', note that AYIYA has quite a bit of overhead as it signs every packet, as such, simple TCP ACKs become a bit larger than normal, but still remain one packet, the size of course will be larger and if you count by that, indeed the upload will be quite a bit more.
TCP losses
[de] Shadow Hawkins on Thursday, 14 October 2010 15:26:39
Actually I am not uploading anything but downloading with about 1.9Mbyte/s. Aiccu is then uploading about 100 kByte/s of whatever which is quite annoying but if its just "normal" I can cope with it.
TCP losses
[ch] Jeroen Massar SixXS Staff on Thursday, 14 October 2010 15:51:58
The 'upload' or better said 'upstream' are basically just TCP ACKs. This is also the reason why having a 100/1 link does not make sense while a 20/4 is already a much better ratio as otherwise you fill the upstream with just ACKing the TCP packets. UDP has a partial advantage there as a file transfer protocol, one could put a offset number in it and just stream all packets to disk, then when the sender indicates 'done', just ask for all the chunks that are missing. Just depends on what one is really trying to accomplish of course, and actually in the case of NNTP the overhead is in the part of uuencoding which blows the binary files up a lot.
TCP losses
[de] Shadow Hawkins on Thursday, 14 October 2010 16:27:15
Okay thanks for this information. Will this be better with a Heartbeat Tunnel?
TCP losses
[ch] Jeroen Massar SixXS Staff on Thursday, 14 October 2010 16:37:56
Heartbeat tunnels use protocol 41, they can't cross a NAT and they are technically less secure as packets are not signed. As they don't have those properties they do have a lot less overhead and thus are a bit faster. See the FAQ for the tunnel comparison for the overview.

Please note Posting is only allowed when you are logged in.

Static Sunset Edition of SixXS
©2001-2017 SixXS - IPv6 Deployment & Tunnel Broker