Arnaldo, Well done although I think you undersell yourself a little as we had this performance a while ago and then it just dropped temporarily while you improved the code!
I will test later today on our emulation network with loss and delay introduced and post the results back. Regards, Ian On 23/08/05, Arnaldo Carvalho de Melo <[EMAIL PROTECTED]> wrote: > Hi, > > I'm pleased to show some promising preliminary, limited results > with the Linux kernel DCCP + CCID3 implementation available at: > > http://www.kernel.org/git/?p=linux/kernel/git/acme/net-2.6.14.git > > Server is an AMD Athlon 2000+ machine with a VIA Rhine II > ethernet card and the client is a Pentium IV with an Intel PRO/100 > card, no packet loss was explicitely induced. > > [EMAIL PROTECTED] ~]# ~acme/ttcp.dccp -c -l1424 -n1000000 -t 192.168.1.2 > ttcp-t: buflen=1424, nbuf=1000000, align=16384/+0, port=5001 dccp -> > 192.168.1.2 > ttcp-t: socket > ttcp-t: connect > ttcp-t: 1424000000 bytes in 169.59 real seconds = 8200.10 KB/sec +++ > ttcp-t: 1000000 I/O calls, msec/call = 0.17, calls/sec = 5896.70 > ttcp-t: 0.0user 0.7sys 2:49real 0% 0i+0d 0maxrss 0+2pf 47488+0csw > [EMAIL PROTECTED] ~]# ~acme/ttcp.dccp -u -l1424 -n1000000 -t 192.168.1.2 > ttcp-t: buflen=1424, nbuf=1000000, align=16384/+0, port=5001 udp -> > 192.168.1.2 > ttcp-t: socket > ttcp-t: 1424000000 bytes in 119.24 real seconds = 11662.69 KB/sec +++ > ttcp-t: 1000006 I/O calls, msec/call = 0.12, calls/sec = 8386.71 > ttcp-t: 0.0user 4.3sys 1:59real 3% 0i+0d 0maxrss 0+2pf 31246+1csw > [EMAIL PROTECTED] ~]# ~acme/ttcp.dccp -t -l1424 -n1000000 -t 192.168.1.2 > ttcp-t: buflen=1424, nbuf=1000000, align=16384/+0, port=5001 tcp -> > 192.168.1.2 > ttcp-t: socket > ttcp-t: connect > ttcp-t: 1424000000 bytes in 135.87 real seconds = 10234.64 KB/sec +++ > ttcp-t: 1000000 I/O calls, msec/call = 0.14, calls/sec = 7359.74 > ttcp-t: 0.1user 4.8sys 2:15real 3% 0i+0d 0maxrss 0+2pf 17955+9csw > [EMAIL PROTECTED] ~]# > > Interestingly packet loss was observed only on the UDP run: > > ttcp-r: 1423601280 bytes in 119.25 real seconds = 11658.33 KB/sec +++ > > Why 1424 bytes? Its the biggest packet size accepted by DCCP in > this test setup using the current unfinished dccp PMTU discovery routines. > > Implementation seems to be solid, no memory leaks or crashes > detected in the limited number of tests the current code received, but > I'm pretty satisfied with the stability of what we have so far. > > There are only 7 changesets not merged by David Miller for > inclusion in Linux 2.6.14, aka today's work :-) > > Now back to testing and collecting data for OSTRA callgraphs and > plottings, also I'll add support in dccp_diag to report RTT, etc thru > the ss utility in iproute2. > > - Arnaldo > > - Arnaldo > - > To unsubscribe from this list: send the line "unsubscribe dccp" in > the body of a message to [EMAIL PROTECTED] > More majordomo info at http://vger.kernel.org/majordomo-info.html > - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
