I have the situation that I need to transfer files (~15-20GB each) across a transcontinental WAN, ~220ms latency, and I have packet loss. period. I know from empirical tests I get at max about 47Mbit/sec at the destination, but with a transmit rate of 45Mbit/s I still get at minimum 9-13 NAKs per section (and it's a given as I can't even run TFMCC as that just goes to a crawl), and with a transmit speed of 50Mbit/s I get about 400 NAKs per section.
Now my questions:
1) After a over night transfer, the transfer was aborted, with still doing ~10NAKs per sections.
- Does that mean 'cause of the guaranteed losses/errors I won't be able to make a succesful transfer?
2) does uftp (server) only resend the NAKs, or does it resend the whole section?
ie. is it efficient to not do a full transfer where it only could've sent a partial transfer?
I got the idea a full transfer was done everytime :(
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
A few losses shouldn't mean a failed transfer. 10 NAKs in a section is roughly a 0.1% loss rate which is not high. UFTP was designed to be tolerant of loss and latency. When there are NAKs, only the missing packets are retransmitted, not the whole section.
If your transfers are failing after a long running session, take a look at the logs on both sides to find out why the session failed. Did the sender reach its max transmission time? Did the receiver time out due to lack of incoming data?
You might also want to check what the calculated group round trip time (GRTT) is when the session starts. If it's too low, you could be timing out faster than if the GRTT accurately reflects the latency. You can use the -R option to put upper and lower bounds on the GRTT which can help with issues like this.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Goodday,
I have the situation that I need to transfer files (~15-20GB each) across a transcontinental WAN, ~220ms latency, and I have packet loss. period. I know from empirical tests I get at max about 47Mbit/sec at the destination, but with a transmit rate of 45Mbit/s I still get at minimum 9-13 NAKs per section (and it's a given as I can't even run TFMCC as that just goes to a crawl), and with a transmit speed of 50Mbit/s I get about 400 NAKs per section.
Now my questions:
1) After a over night transfer, the transfer was aborted, with still doing ~10NAKs per sections.
- Does that mean 'cause of the guaranteed losses/errors I won't be able to make a succesful transfer?
2) does uftp (server) only resend the NAKs, or does it resend the whole section?
ie. is it efficient to not do a full transfer where it only could've sent a partial transfer?
I got the idea a full transfer was done everytime :(
A few losses shouldn't mean a failed transfer. 10 NAKs in a section is roughly a 0.1% loss rate which is not high. UFTP was designed to be tolerant of loss and latency. When there are NAKs, only the missing packets are retransmitted, not the whole section.
If your transfers are failing after a long running session, take a look at the logs on both sides to find out why the session failed. Did the sender reach its max transmission time? Did the receiver time out due to lack of incoming data?
You might also want to check what the calculated group round trip time (GRTT) is when the session starts. If it's too low, you could be timing out faster than if the GRTT accurately reflects the latency. You can use the -R option to put upper and lower bounds on the GRTT which can help with issues like this.