We can't get the network transmission [to exceed] 2-3 % [of nameplate]
What does iperf report? Have you tried a link local subnet?
1000Mbps
Have you tried Send Unreliable Datagram(UD) with immediate (or multicast)?
-----Original Message----- From: Brent Kimberley Sent: Wednesday, February 24, 2016 12:28 PM To: 'richard.k.evans@nasa.gov' Cc: udpcast@udpcast.linux.lu Subject: RE: Udpcast Digest, Vol 123, Issue 1
Hi Richard, Your performance on a general purpose machine should exceed the following.
384mbps @9000 395mbps @4096 352mbps @1456 312mbps @1024 193mbps @256 63mbps @64
Best Regards, Brent
<snippet> @REM cd %TEMP% @REM fsutil file createnew test1.tx 2199023255552 @REM dir test1.tx
@REM cd %TEMP% .\udp-sender.exe -b <blocksize> --autostart 5 --stat-period 2000 --max-bitrate 1000m --async --fec 8x8 --mcast-rdv-addr 127.0.0.1 --mcast-data-addr 127.0.0.1 --interface 127.0.0.1 --file test1.tx
@REM cd %TEMP% .\udp-receiver.exe --file test.<blocksize>.rx --nosync --mcast-rdv-addr 127.0.0.1 --stat-period 2000 --interface 127.0.0.1
@REM del test*.* udp*.exe </snippet>
< receiver observations> UDP receiver for test.64.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K ( 63.23 Mbps)) Transfer complete.
UDP receiver for test.256.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (193.83 Mbps)) Transfer complete.
UDP receiver for test.1024.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (312.40 Mbps) Transfer complete.
UDP receiver for test.4096.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (395.57 Mbps)) Transfer complete.
UDP receiver for test.9000.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (384.52 Mbps) Transfer complete.
UDP receiver for test.1456.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (352.54 Mbps) Transfer complete. </receiver observations>
Message: 1 Date: Thu, 18 Feb 2016 18:39:30 +0000 From: "Evans, Richard K. (GRC-H000)" richard.k.evans@nasa.gov To: "udpcast@udpcast.linux.lu" udpcast@udpcast.linux.lu Cc: "Evans, Richard K. (GRC-H000)" richard.k.evans@nasa.gov Subject: [Udpcast] maximum asynchronous speed and Jumbo Frames Message-ID: DC3FB55EE7FEFD409A6FC1EA00D50D0B0F5C4510@NDJSMBX203.ndc.nasa.gov Content-Type: text/plain; charset="iso-8859-1"
Hello, I am using the UDPCast in asynchronous mode to transmit large data files through a Data Security Diode (mono-directional network).
Here is the command we use to transmit:
Udp-sender.exe --async --fec 8x8 --max-bitrate 1000m --mcast-rdv-addr 224.2.2.1 --mcast-data-addr 224.2.2.2 --interface <ipTx> --file myfile.dat
The Receive command I use is:
udp-receiver.exe --nosync --mcast-rdv-addr 224.2.2.1 --interface <ipRx> --f myfile.dat
Overall, this works well but we have noticed that we can't get the network transmission speed up above 2-3 % (22Mbps) of the Gigabit Ethernet adapter's full capability (1000Mbps). We also know that we are not processor limited, nor are we disk speed limited or anything else we can think of.
I know that the asynchronous mode application is for satellite transmissions that are understandably slower than what a file transfer purely over Gibabit Ethernet can do, so I am writing here to ask if any one has any thoughts or insights about what might be the core issue that keeps the process running so slow.
Fwiw, the Data Security Diode is a Canary GT-10SD, but for tuning UDPCast, we bypass the diode with a direct cable connection between Tx and Rx machines. So the performance of the system I am describing is without the diode in place.
Thanks in advance, folks. -Rich
- Richard Evans, NASA GRC - Plum Brook Station ? http://www.linkedin.com/in/rkevans ???216.973.4398
------------------------------
_______________________________________________ Udpcast mailing list Udpcast@udpcast.linux.lu https://www.udpcast.linux.lu/mailman/listinfo/udpcast
End of Udpcast Digest, Vol 123, Issue 1 *************************************** THIS MESSAGE IS FOR THE USE OF THE INTENDED RECIPIENT(S) ONLY AND MAY CONTAIN INFORMATION THAT IS PRIVILEGED, PROPRIETARY, CONFIDENTIAL, AND/OR EXEMPT FROM DISCLOSURE UNDER ANY RELEVANT PRIVACY LEGISLATION. No rights to any privilege have been waived. If you are not the intended recipient, you are hereby notified that any review, retransmission, dissemination, distribution, copying, conversion to hard copy, taking of action in reliance on or other use of this communication is strictly prohibited. If you are not the intended recipient and have received this message in error, please notify me by return e-mail and delete or destroy all copies of this message.
Hi Brent, Thanks for the quick reply and for the really great insights.
Interesting! We never tried to use the BLOCKSIZE parameter because I figured that the default value of 1456 (which is stated to also be the maximum) would be inarguably the best value for speed. For reasons I don't fully understand yet, I clearly was wrong. It turns out that with BLOCKSIZE =300 we are seeing ~170mbps - versus the default value providing only ~22mps. Any thoughts on why that is the case?
Also, I noticed that all your tests were done on the 127.0.0.1 (software loopback) IP addr. My understanding is that that doesn't actually transmit anything using the actual network adapter and so it is not representative of a real through-put. Can you think of anything specific that would cause the network adapter itself to be a bottleneck? I know the Windows MTU is set to 1500 bytes by default. I'm not so savvy that I want to go monkeying around with that without someone explaining to me why it would help to do so.
So, with the BLOCKSIZE=300 I can hit ~170mps... but you're saying I should be able to achieve as much as ~350mps.. I would really love to achieve this. Any additional help you (or anyone) can provide is greatly appreciated.
Thanks! -Rich
-----Original Message----- From: Brent Kimberley [mailto:Brent.Kimberley@Durham.ca] Sent: Wednesday, February 24, 2016 1:52 PM To: Evans, Richard K. (GRC-H000) Cc: 'udpcast@udpcast.linux.lu' Subject: RE: Udpcast Digest, Vol 123, Issue 1
We can't get the network transmission [to exceed] 2-3 % [of nameplate]
What does iperf report? Have you tried a link local subnet?
1000Mbps
Have you tried Send Unreliable Datagram(UD) with immediate (or multicast)?
-----Original Message----- From: Brent Kimberley Sent: Wednesday, February 24, 2016 12:28 PM To: 'richard.k.evans@nasa.gov' Cc: udpcast@udpcast.linux.lu Subject: RE: Udpcast Digest, Vol 123, Issue 1
Hi Richard, Your performance on a general purpose machine should exceed the following.
384mbps @9000 395mbps @4096 352mbps @1456 312mbps @1024 193mbps @256 63mbps @64
Best Regards, Brent
<snippet> @REM cd %TEMP% @REM fsutil file createnew test1.tx 2199023255552 @REM dir test1.tx
@REM cd %TEMP% .\udp-sender.exe -b <blocksize> --autostart 5 --stat-period 2000 --max-bitrate 1000m --async --fec 8x8 --mcast-rdv-addr 127.0.0.1 --mcast-data-addr 127.0.0.1 --interface 127.0.0.1 --file test1.tx
@REM cd %TEMP% .\udp-receiver.exe --file test.<blocksize>.rx --nosync --mcast-rdv-addr 127.0.0.1 --stat-period 2000 --interface 127.0.0.1
@REM del test*.* udp*.exe </snippet>
< receiver observations> UDP receiver for test.64.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K ( 63.23 Mbps)) Transfer complete.
UDP receiver for test.256.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (193.83 Mbps)) Transfer complete.
UDP receiver for test.1024.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (312.40 Mbps) Transfer complete.
UDP receiver for test.4096.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (395.57 Mbps)) Transfer complete.
UDP receiver for test.9000.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (384.52 Mbps) Transfer complete.
UDP receiver for test.1456.rx at 127.0.0.1 on Software Loopback Interface 1 Connected as #0 to 127.0.0.1 bytes= 2 097 152K (352.54 Mbps) Transfer complete. </receiver observations>
Message: 1 Date: Thu, 18 Feb 2016 18:39:30 +0000 From: "Evans, Richard K. (GRC-H000)" richard.k.evans@nasa.gov To: "udpcast@udpcast.linux.lu" udpcast@udpcast.linux.lu Cc: "Evans, Richard K. (GRC-H000)" richard.k.evans@nasa.gov Subject: [Udpcast] maximum asynchronous speed and Jumbo Frames Message-ID: DC3FB55EE7FEFD409A6FC1EA00D50D0B0F5C4510@NDJSMBX203.ndc.nasa.gov Content-Type: text/plain; charset="iso-8859-1"
Hello, I am using the UDPCast in asynchronous mode to transmit large data files through a Data Security Diode (mono-directional network).
Here is the command we use to transmit:
Udp-sender.exe --async --fec 8x8 --max-bitrate 1000m --mcast-rdv-addr 224.2.2.1 --mcast-data-addr 224.2.2.2 --interface <ipTx> --file myfile.dat
The Receive command I use is:
udp-receiver.exe --nosync --mcast-rdv-addr 224.2.2.1 --interface <ipRx> --f myfile.dat
Overall, this works well but we have noticed that we can't get the network transmission speed up above 2-3 % (22Mbps) of the Gigabit Ethernet adapter's full capability (1000Mbps). We also know that we are not processor limited, nor are we disk speed limited or anything else we can think of.
I know that the asynchronous mode application is for satellite transmissions that are understandably slower than what a file transfer purely over Gibabit Ethernet can do, so I am writing here to ask if any one has any thoughts or insights about what might be the core issue that keeps the process running so slow.
Fwiw, the Data Security Diode is a Canary GT-10SD, but for tuning UDPCast, we bypass the diode with a direct cable connection between Tx and Rx machines. So the performance of the system I am describing is without the diode in place.
Thanks in advance, folks. -Rich
- Richard Evans, NASA GRC - Plum Brook Station ? http://www.linkedin.com/in/rkevans ???216.973.4398
------------------------------
_______________________________________________ Udpcast mailing list Udpcast@udpcast.linux.lu https://www.udpcast.linux.lu/mailman/listinfo/udpcast
End of Udpcast Digest, Vol 123, Issue 1 *************************************** THIS MESSAGE IS FOR THE USE OF THE INTENDED RECIPIENT(S) ONLY AND MAY CONTAIN INFORMATION THAT IS PRIVILEGED, PROPRIETARY, CONFIDENTIAL, AND/OR EXEMPT FROM DISCLOSURE UNDER ANY RELEVANT PRIVACY LEGISLATION. No rights to any privilege have been waived. If you are not the intended recipient, you are hereby notified that any review, retransmission, dissemination, distribution, copying, conversion to hard copy, taking of action in reliance on or other use of this communication is strictly prohibited. If you are not the intended recipient and have received this message in error, please notify me by return e-mail and delete or destroy all copies of this message.
-
Brent Kimberley -
Evans, Richard K. (GRC-H000)