On Thu December 18 2008, Michael S. Zick wrote: > > You would probably be better off with tipc: > http://tipc.sourceforge.net > Well, I mentioned it, and it is a slow weekend - - - Besides, haven't done anything with tipc in nearly 3 decades ;) = = = = Hardware: C2Q (Q9300) @ 2.5Ghz (probably typical of a server) Host: 64bit-Gentoo, 2.6.27.10-vs2.3.0.36.2 + tipc-1.7.6 + tipc_utils-1.1.8 Guest: 32bit-Gentoo-i686 Network: dummy0 = = = = The good: The tipc sources do not conflict with the vserver patch. Just overwrite net/tipc with the project's tarball for .27 (The kernel main-stream isn't even close to current.) The bad: The tipc sources are probably not context aware. I don't think that really makes a difference in this case, could be wrong. Build and install stuff - the usual fun, plus build tools in 32bit also. Interfaces do not have addresses, only the machine node has a "name" (number). Probably just using the context number for the "name" would do fine. = = = = The included benchmark program (guest is server, host is client here): wolf-c2q benchmark # ./client_tipc ****** TIPC benchmark client started ****** Client master: Starting Latency Benchmark Exchanging 10240 messages of size 64 octets (burst size 10240) ... took 71 ms (round-trip avg/msg: 6 us) Exchanging 2560 messages of size 256 octets (burst size 2560) ... took 17 ms (round-trip avg/msg: 6 us) Exchanging 640 messages of size 1024 octets (burst size 640) ... took 4 ms (round-trip avg/msg: 6 us) Exchanging 160 messages of size 4096 octets (burst size 160) ... took 1 ms (round-trip avg/msg: 6 us) Exchanging 40 messages of size 16384 octets (burst size 40) ... took 0 ms (round-trip avg/msg: 0 us) Exchanging 10 messages of size 65536 octets (burst size 10) ... took 0 ms (round-trip avg/msg: 0 us) Client master: Completed Latency Benchmark Client master: Starting Throughput Benchmark Exchanging 8*10240 messages of size 64 octets (burst size 2048) ... took 163 ms (avg 251288 msg/s/dir, 128659456 bits/s/dir) avg execution time (send+receive) 0 us/msg Exchanging 8*2560 messages of size 256 octets (burst size 512) ... took 47 ms (avg 217872 msg/s/dir, 446201856 bits/s/dir) avg execution time (send+receive) 1 us/msg Exchanging 8*640 messages of size 1024 octets (burst size 128) ... took 9 ms (avg 284444 msg/s/dir, 2330165248 bits/s/dir) avg execution time (send+receive) 0 us/msg Exchanging 8*160 messages of size 4096 octets (burst size 32) ... took 4 ms (avg 160000 msg/s/dir, 5242880000 bits/s/dir) avg execution time (send+receive) 1 us/msg Exchanging 8*40 messages of size 16384 octets (burst size 8) ... took 11 ms (avg 14545 msg/s/dir, 1906442240 bits/s/dir) avg execution time (send+receive) 17 us/msg Exchanging 8*10 messages of size 65536 octets (burst size 2) ... took 2 ms (avg 20000 msg/s/dir, 10485760000 bits/s/dir) avg execution time (send+receive) 12 us/msg Client master: Completed Throughput Benchmark ****** TIPC benchmark client finished ****** = = = = Interesting. That should get a message to the host and back. ;) Note: The origins of tipc at Ericsson was to control networks of telephone switching machines in "near real time" - initially to implement a "least cost routing" switching system. Goal: Complete the telephone connection dialed within 4 seconds Nationally, 10 seconds Inter-nationally. Do I write as if I had something to do with it back in '82? ;) Mike