----- Forwarded message from kai ----- To: [EMAIL PROTECTED] Date: Sat, 23 Nov 2002 16:06:31 +0100 Subject: Transmit timestamps to fast? X-Mailer: Mutt (Linux 2.4.19 i586, Debian, SusE go home!)
Hi, im computer science student and writing software for bottleneck bandwidth measurement (rrbw) for a research project in adaptive applications. For potential Bandwidth calculation (send Bandwidth) I need the correct transmit timestamp of packets, measured at localhost (where the packet is produced). I've done some measurement... pls see debug log.. When I calculate the potential (sent) bandwidth from the first and last packet in a bunch. (delta time from first to last packet)/packet size. I get Bandwidths which are much higher as the physical network can handle (100Mbit-FD). This vary from nic to nic. This log is measured with a AMD PCnet Family NIC (26.4 Mbyte/s). A 3Com 3c95 NIC gives transmit bandwidth up to 160 Mbyte/s). Strange %-) Why do I get so small deltas from the timestamps? Is this normaly, or maybe a bug?. Maybe the nic buffers the packets?? Can somebody give me a hint why this is so, and how can I get correct timestamps (Timestamp, when the packet is on wire)? It would be great If somebody could help me :) This effect is the same at winpcap 3.0alpha1 - alpha4. Im using Win2000 SP2 on a athlon 1,3G. [root.nmaPacketReceiver, receiverThread] packet: 1038062241:30641 (52428) ID:25755 [root.nmaCalcMetric, receiverThread] receive(25756) [root.nmaCalcMetric, receiverThread] packets in map: 5000 packets in list: 5000 [root.nmaCalcMetric, receiverThread calcPerfSample] np:(25756) [root.nmaCalcMetric, receiverThread calcPerfSample] bp:(25751) [root.nmaCalcMetric, receiverThread calcPerfSample] id(25751) timings: arr=1038062241:26354 transm=1038062239:178061 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] id(25752) timings: arr=1038062241:27426 transm=1038062239:178147 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] id(25753) timings: arr=1038062241:28498 transm=1038062239:178213 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] id(25754) timings: arr=1038062241:29569 transm=1038062239:178274 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] id(25755) timings: arr=1038062241:30641 transm=1038062239:178337 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] id(25756) timings: arr=1038062241:31727 transm=1038062239:178401 [root.nmaCalcMetric, receiverThread calcPerfSample] len: 1500 [root.nmaCalcMetric, receiverThread calcPerfSample] Packet: cnt=6 trans_size=9000 arr_size=9000 Deltas: arr=0.005373, transmit=0.000340 Metrics: RRBW=13400335.008375 1.675042 (MB/s), PBW=211764705.882353 26.470588 /MB/s) fyi. The receive timestamps are measured at the destination host, where this packet is send to. This arrivial timings are sent back via socket to sent host... Thx in advance, Kai Olbrich -- Windows has detected that your mouse pointer has moved. Your computer will have to restart for this change to take effect. Do you wish to restart now (Y/N)? ----- End forwarded message ----- -- Windows has detected that your mouse pointer has moved. Your computer will have to restart for this change to take effect. Do you wish to restart now (Y/N)? ================================================================== This is the WinPcap users list. It is archived at http://www.mail-archive.com/[email protected]/ To unsubscribe use mailto: [EMAIL PROTECTED]?body=unsubscribe ==================================================================
