[rescue] SunFastEthernet questions

Tue Sep 2 23:23:38 CDT 2003

OK... I found ttcp on the net...  for reference my Solaris for Intel
box (P200MMX) running Solaris 8 is host fury and my U60 is host delenn.

----------------------------- test 1 -----------------------------
delenn (U60) transmitting, fury (Intel) receiving
------------------------------------------------------------------

delenn% ./ttcp -t -s fury
ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> fury
ttcp-t: socket
ttcp-t: connect
ttcp-t: 16777216 bytes in 2.07 real seconds = 7927.94 KB/sec +++
ttcp-t: 2048 I/O calls, msec/call = 1.03, calls/sec = 990.99
ttcp-t: 0.0user 0.1sys 0:02real 6% 0i+0d 0maxrss 0+0pf 688+5csw

fury% ./ttcp -r -s
ttcp-r: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp
ttcp-r: socket
ttcp-r: accept from 10.0.0.11
ttcp-r: 16777216 bytes in 2.07 real seconds = 7926.74 KB/sec +++
ttcp-r: 2108 I/O calls, msec/call = 1.00, calls/sec = 1019.87
ttcp-r: 0.0user 1.2sys 0:02real 61% 0i+0d 0maxrss 0+0pf 762+805csw

----------------------------- test 2 -----------------------------
fury (Intel) transmitting, delenn (U60) receiving
------------------------------------------------------------------

fury% ./ttcp -t -s delenn
ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> delenn
ttcp-t: socket
ttcp-t: connect
ttcp-t: 16777216 bytes in 1.79 real seconds = 9177.40 KB/sec +++
ttcp-t: 2048 I/O calls, msec/call = 0.89, calls/sec = 1147.17
ttcp-t: 0.0user 1.4sys 0:01real 83% 0i+0d 0maxrss 0+0pf 71+111csw

delenn% ./ttcp -r -s
ttcp-r: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp
ttcp-r: socket
ttcp-r: accept from 10.0.0.10
ttcp-r: 16777216 bytes in 1.79 real seconds = 9169.92 KB/sec +++
ttcp-r: 6079 I/O calls, msec/call = 0.30, calls/sec = 3402.34
ttcp-r: 0.0user 0.2sys 0:01real 13% 0i+0d 0maxrss 0+0pf 6053+22csw

So, with ttcp I'm getting 8 to 9 MB/sec on the wire...  which is
getting pretty close to the 10 max theoretical...

My ftp speeds may have not been accurate due to sofware raid 5 and
lowly P200MMX ???

Seems ttcp is showing my net to be pretty speedy and clean... and
everything is in autonegotiate...

For testing I did:

# ndd -set /dev/hme adv_autoneg_cap 0
# ndd -set /dev/hme adv_100T4_cap 0
# ndd -set /dev/hme adv_100fdx_cap 1
# ndd -set /dev/hme adv_100hdx_cap 0
# ndd -set /dev/hme adv_10fdx_cap 0
# ndd -set /dev/hme adv_10hdx_cap 0

on the U60...

Then the 2924XL was saying the port was autonegged to 100baseTX half
duplex... oops !  so I forced it to full, and reran the slower of my
two tests:

----------------------------- test 3 -----------------------------
delenn (U60) transmitting, fury (Intel) receiving
hme forced to 100 full, switched forced to full duplex
------------------------------------------------------------------

delenn# ./ttcp -t -s fury
ttcp-t: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp  -> fury
ttcp-t: socket
ttcp-t: connect
ttcp-t: 16777216 bytes in 1.81 real seconds = 9070.48 KB/sec +++
ttcp-t: 2048 I/O calls, msec/call = 0.90, calls/sec = 1133.81
ttcp-t: 0.0user 0.1sys 0:01real 7% 0i+0d 0maxrss 0+0pf 684+7csw

fury% ./ttcp -r -s
ttcp-r: buflen=8192, nbuf=2048, align=16384/0, port=5001  tcp
ttcp-r: socket
ttcp-r: accept from 10.0.0.11
ttcp-r: 16777216 bytes in 1.80 real seconds = 9078.45 KB/sec +++
ttcp-r: 2064 I/O calls, msec/call = 0.90, calls/sec = 1143.67
ttcp-r: 0.0user 1.3sys 0:01real 72% 0i+0d 0maxrss 0+0pf 801+1002csw

However, running several more iterations resulted in data points
between 8390.30 KB/sec and 9039.55 KB/sec.

Putting everything back at auto and doing severl more delenn->fury
tests I ended up with 7205.18 KB/sec to 9209.22 KB/sec and the
switch was autonegged to 100 full.

Well... I first thought I found something, but then it appears 
I haven't as I've been able to push just as much data at autonegged
speeds.

Using 10 times as many buffers (longer test time hopefully results in
a better overall average of network throughput), I received 8669.85
A few more tests showed less variance between tests...

So... I bumped it up to 100 times as many buffers (much longer test,
and hopefully far less variance)... and I got 8647.25 KB/sec.

Going with forced configs at 100 full, I got 8452.21KB/sec so, at least
in my network autonegotiate performs as well as hard config.

Also, there is not much activity on this network except for about
6 or 7 IMAP connections to fury, and 1 to 2 POP connections to fury
and the occasional www hit to fury... so that traffic should not have
a hit on ttcp results (not in so far as packeg counts... if there was
a cpu spike on the P200MMX intel box (fury) I don't know how that would
effect it...).

So... anyone else have any similar real world measurement... ???

Maybe this is simply network/system/switch dependant.

-- Curt

>Date: Tue, 2 Sep 2003 21:07:24 -0500
>From: "Jonathan C. Patschke" <jp at celestrion.net>
>To: The Rescue List <rescue at sunhelp.org>
>Subject: Re: [rescue] SunFastEthernet questions
>X-Message-Flag: Microsoft Outlook is an insecure piece of crap.  Please 
consider using a different email client.
>X-All-Your-Base: Are Belong To Us
>X-GPG-Keyserver: keys.celestrion.net
>X-GPG-Fingerprint: 143B 8D8C 5AD7 44D4 0143  FE85 288C 370B 67B9 A022
>
>On Tue, 2 Sep 2003, Curtis H. Wilbar Jr. wrote:
>
>> Even if my hme and my autonegotiated switch negotiate at 100baseTX full
>> duplex (and never waver from that), the fact that the hme is in auto
>> negotiate is going to degrade my performance ?
>
>Yes, in my experience.
>
>-- 
>Jonathan Patschke   )  "Gamer weenies...are the sludge at the bottom
>Elgin, TX          (    of the user swamp."           --Gary Nichols
>_______________________________________________
>rescue list - http://www.sunhelp.org/mailman/listinfo/rescue

Curtis Wilbar
Hawk Mountain Networks
rescue at hawkmountain.net

My e-mail is protected against viruses and spam by MailGuardian
                  http://www.mailguardian.net
          Top notch protection at unbelievable prices