From my Qube2:

ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> 
192.168.1.32
ttcp-t: socket
ttcp-t: connect
ttcp-t: 858750976 bytes in 711.51 real seconds = 1178.66 KB/sec +++
ttcp-t: 104828 I/O calls, msec/call = 6.95, calls/sec = 147.33
ttcp-t: 0.2user 23.3sys 11:51real 3% 0i+0d 0maxrss 0+182222pf 297588+1663csw

      Performance is the same on a smaller (but still big) file, as one 
would expect:

ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> 
192.168.1.32
ttcp-t: socket
ttcp-t: connect
ttcp-t: 48201728 bytes in 40.60 real seconds = 1159.47 KB/sec +++
ttcp-t: 5884 I/O calls, msec/call = 7.07, calls/sec = 144.93
ttcp-t: 0.0user 1.2sys 0:40real 3% 0i+0d 0maxrss 0+11654pf 17196+16csw

      If I send the output of ttcp to /dev/null, I get this:

ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> 
192.168.1.32
ttcp-t: socket
ttcp-t: connect
ttcp-t: 48201728 bytes in 14.81 real seconds = 3178.04 KB/sec +++
ttcp-t: 5884 I/O calls, msec/call = 2.58, calls/sec = 397.26
ttcp-t: 0.0user 1.2sys 0:14real 8% 0i+0d 0maxrss 0+11482pf 15269+240csw
bash-2.05b# ttcp -t 192.168.1.32 < /home/cvk/pcap/chan6.pcap2

      It's running 2.0_RC4 from a couple of weeks ago:

bash-2.05# uname -a
NetBSD tokyo 2.0_RC4 NetBSD 2.0_RC4 (TOKYO) #2: Sat Oct 16 02:55:49 UTC 
2004  root@tokyo:/usr/src/sys/arch/cobalt/compile/TOKYO cobalt

--
c v k @ z y b x . c o m

Andy Ruhl wrote:
> On Sat, 06 Nov 2004 22:30:27 +0100, Colin Raven <duiker@haggis.nl> wrote:
> 
>>
>>
>>
>>---------- Forwarded message ----------
>>From: Colin Raven <duiker@haggis.nl>
>>To: Andy Ruhl <acruhl@gmail.com>
>>Date: Sat, 06 Nov 2004 22:29:16 +0100
>>Subject: Re: Binding more than one IP to a NIC
>>Andy Ruhl wrote:
>>
>>
>>>On Sat, 6 Nov 2004 15:56:07 -0500, Justin Newcomer <liquidice5@gmail.com> wrote:
>>>
>>>
>>>
>>>>i dont know what you consider fixed, but in 1.6.1 with apache 2 i get
>>>>about 4-8 MB/s out of my qube
>>>>
>>>>
>>>
>>>That's quite surprising. I just ran a test again a little earlier
>>>because I didn't know if I ran it after going to 2.0-release. Here it
>>>is:
>>>
>>>ttcp-t: 733945856 bytes in 563.28 real seconds = 1272.45 KB/sec +++
>>>
>>>The above result was with the Qube.
>>>
>>>That same test took 81 seconds on a FreeBSD box. The source in both
>>>cases was NetBSD/i386.
>>>
>>>
>>>
>>>
>>
>>This might well be useful down the road aways.......
>>How did you run this test? Did you download a .bin file from a known
>>location on two different hosts?
>>How did you output those statistics you quoted above?
>>I'd like to do (and document how to) this with my qube2 which seems
>>kinda quick at downloading over an 8Mps SDSL connection, but I don't
>>know how to acquire definitive statistics as you showed us.
>>-Colin
> 
> 
> /usr/pkgsrc/benchmarks/ttcp
> 
> I catted a movie file first to my FreeBSD box, which took 81 seconds.
> I then did it to the Qube and those were the results I got, 500 and
> some seconds.
> 
> I'm not doing anything fancy. I was told before by one of the cobalt
> developers that there is some bug on the PCI bus limiting bandwidth.
> If you're doing better than that, I'd like to know about it.
> 
> I'm on a Qube2 BTW.
> 
> Andy