The default UDT parameters are tuned for very high speed networks. In very
low bandwidth environment, it may send too many packets out before it
detects the available bandwidth, thus it cause many packet loss. We are
working on automatic parameter tuning on this issue.
Thanks for the report on your interesting experiment.
On Fri, 9 Nov 2007, hongsion wrote:
> hi UDT fellows,
>
> I just compared TCP and UDT between america and china. I transfer a file, which size if 3.46M.
> 1) I install a ftp server (ServU) in china. and put file from america to china. and I found it takes about 73 seconds to finish it. and I use Dumeter to monitor the network flow. and during this period, the actual data transfered to ftp server is 3.46M, that mean TCP didn't transfer some duplicated data.
> 2)Then i setup sendfile in america. and recvfile in china. two PC are same with testing case 1). this time UDT use only 53 seconds to finish the work. But Dumeter displays that UDT has transferd 5.1M bytes to PC in china. that mean UDT is faster than TCP, but it transfered some duplicated data.
>
> I think if many people use UDT same time, it will waste more network bandwidth.
>
>
> regards,
>
>
>
|
