Hi Folks,
Today I did tests with Ian McDonald, using patched netcat, with
which we chatted about the state of the Linux DCCP stack over a DCCP
connection from Brazil to New Zealand, with Ian's machine behind a DSL
router that does NAT and doesn't know squat about DCCP, with the machines
being 18 hops away, exciting, huh? :-)
We also did a ttcp session transmitting 5000 256 byte packets, I
used my latest DCCP tree, that has fixes for 32 bit math on ELAPSED_TIME,
TIMESTAMP, send rate, etc as well with other important fixes, Ian used the
code in upstream Linus Torvalds's git tree, that has bugs fixed in the tree
I used, but even so:
My side:
[EMAIL PROTECTED] acme]# ~acme/ttcp_new.dccp -c -l256 -r
ttcp-r: buflen=256, nbuf=2048, align=16384/+0, port=5001 dccp
ttcp-r: socket
ttcp-r: accept from 222.153.242.9
ttcp-r: 496128 bytes in 24.80 real seconds = 19.54 KB/sec +++
ttcp-r: 1939 I/O calls, msec/call = 13.10, calls/sec = 78.19
ttcp-r: 0.0user 0.0sys 0:24real 0% 0i+0d 0maxrss 0+1pf 1939+0csw
[EMAIL PROTECTED] acme]#
His side:
ttcp-t: buflen=256, nbuf=5000, align=16384/+0, port=5001 dccp -> 201.11.59.173
ttcp-t: socket
ttcp-t: connect
ttcp-t: 1280000 bytes in 23.76 real seconds = 52.61 KB/sec +++
ttcp-t: 5000 I/O calls, msec/call = 4.87, calls/sec = 210.44
ttcp-t: 0.0user 0.0sys 0:23real 0% 0i+0d 0maxrss 0+1pf 5044+0csw
Yes, the results are not perfect (cheap shot), we have problems
with checksum in some cases (we disabled checksum verification in our
kernels, this may be related to some packet munging on middle boxes, we're
not sure at this point) but we're getting there :-)
A dump of this historical (at least for me :-) ) session is
available at:
http://oops.ghostprotocols.net:81/dccp/tcpdump_first_ttcp_dccp_session_BR_NZ.gz
And if you have problems with the above site:
http://vger.kernel.org/~acme/tcpdump_first_ttcp_dccp_session_BR_NZ.gz
I'll be doing more tests with other machines in other parts of the
world as I find people to test with me, but so far it doesn't seems like
DCCP packets aren't like birds going thru jet stream engines, oops
firewalls/nat boxes like someone said recently 8)
I'm not aware of other tests like this, are you? Anyway, I'm more
than happy with the state of this beast!
- Arnaldo
-
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
