Date: Sun, 17 Nov 1996 23:41:18 -0600 (CST) From: Joe Greco <jgreco@brasil.moneng.mei.com> To: dennis@etinc.com (dennis) Cc: isp@freebsd.org Subject: Re: changed to: Frac T3? Message-ID: <199611180541.XAA01714@brasil.moneng.mei.com> In-Reply-To: <199611180207.VAA24485@etinc.com> from "dennis" at Nov 17, 96 09:07:01 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> J. Greco writes...
>
>
> >No, I have a bad habit of routing on low end equipment...
> >
> >But I agree... if a P100 can do 5000pps, what can a PP200 do :-)
>
> Perhaps if Joe would describe this test it would be a good start.
I did, briefly, already.
System: ASUS P/I-P55T2P4, P100 CPU, 16MB RAM, Znyx 314 quad DE21040
ethernet, NE2000 ethernet.
Normal traffic load: 400pps.
trantor% netstat -I ed0 -b 10 | cut -c61-120
input (Total) output
packets errs bytes packets errs bytes colls
103696372 5990 3192031736 104078666 0 3144425590 260844
3438 0 1349683 3443 0 1347715 94
3945 0 1468394 3948 0 1466061 118
3344 0 1234913 3352 0 1233600 95
3478 0 1264837 3483 0 1262795 152
4872 0 1859850 4877 0 1857370 178
3469 0 1277983 3473 0 1275461 145
3649 0 1417299 3653 0 1415398 223
3055 0 1176986 3061 0 1175475 177
4307 0 1678561 4311 0 1676535 188
5142 0 2138517 5147 0 2135524 122
Client 1: ping from an ASUS P100/DE21041 to an ASUS SP3G/DE21041
traversing two of trantor's 'de' interfaces.
hummin# ping -f 206.55.64.254
(now back to trantor)
4764 0 1627137 4769 0 1624997 129
16846 0 2849899 16853 0 2847445 507
21053 0 3050225 21062 0 3048870 574
21920 0 3340778 21926 0 3337597 880
21654 0 3087022 21658 0 3085220 659
Client 2: add a ping from an ASUS SP3G to an ASUS P100/DE21041, same
machines, other direction.
tazenda# ping -f 204.95.172.243
(now back to trantor)
21525 0 3219253 21530 0 3217139 790
20864 0 2902915 20869 0 2900088 604
32442 0 4057497 32447 0 4055258 1664
34969 0 4610889 34974 0 4608259 2409
34630 0 4362398 34631 0 4360598 2072
36529 0 4495033 36537 0 4493280 2285
Ummm.. collisions are getting high. Unfortunately I do not have a
lot more in terms of "fast machines" to hammer on this... right now
I can squeeze it up to about 3900pps, with about 48% idle CPU, but
also running some traffic through the NE2000 interface to get some
other "fast" machines involved.
39173 0 4998716 39178 0 4996554 2666
38726 0 5052015 38730 0 5050544 2791
39024 0 5114201 39031 0 5111263 2873
40068 0 5043406 40073 0 5040965 2736
41134 0 5092456 41143 0 5089235 2677
39956 0 5046334 39961 0 5044636 2608
I would have to go set up a machine or two in order to push past
this. However, considering that the NE2000 was contributing 600pps
to the above numbers, I am assuming that a lot of the CPU was getting
chewed by it.
I am sorry I can't do any faster right now :-) This is a production
network and I just don't have the hardware online to try to beat
on a router right now.
> Anyone have a feel for the avg packet size over a typical backbone
> link? A T3 with an avg packet size of 500 bytes is 21000pps full
> duplex...I suspect the ave packet size may be smaller with lots
> of dialup traffic.....
I tend to see an average of about 350 bytes.
... JG
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611180541.XAA01714>