[nsp] Real IP troughput over ATM

[sthaug at nethelp.no]

> Realistically you're getting pretty close to the max....  There is IP, TCP 
> overhead too.  The packets/sec number is not IP.  That's down at the 
> interface level already so it's really ATM frames methinks.  i never pay 
> much attention to those numbers because they have been known to be wrong 
> and are most certainly rounded off.
> 
> Your throughput will also depend on TCP window sizes and various other 
> things.  A single stream has a hard time maxing out the link for a lot of 
> reasons.

Also, he has not accounted for SDH overhead. If you remove SDH overhead,
your STM-1 has 149.76 Mbps available for ATM. Remove ATM cell header,
and you're down to 149.76*48/53 = 135.63 Mbps - and this is before any
AAL5 padding and CRC.

Steinar Haug, Nethelp consulting, sthaug at nethelp.no


> 
> --On Thursday, November 06, 2003 17:47 -0300 "Noriega, Alejandro" 
> <ANoriega at prima.com.ar> wrote:
> 
> >> Hello folks,
> >> 	I have been reading all sorts of documents about real IP
> >> troughput over ATM link but I didn't find anything useful.
> >> The last week my company bought 130Mbps IP over an OC3 link.
> >> I was testing the link, only with downstream traffic, and the link
> >> never reaches more than 125Mbps (that was the info that the 'input
> >> rate' of "show int ATM2/0" showed me. I'm not sure if that rate is IP
> >> or ATM.)
> >>
> >> OC3 = 155.520Mbps
> >> ATM AAL5 = 1 cell = 53bytes (48 payload + 5 header)
> >>
> >> I am not considering the bytes cosumed by the AAL5 cell trailer:
> >> Should the real IP troughput over an ATM link be ~
> >> 155.520/(53*8)*(48*8)=140.850 ?.
> >> Well, if my calculation is right my link SHOULD have ~143.542 of
> >> downstream to get 130Mbps IP, right? (143.54/(53*8)*(48*8)=130)
> >>
> >> My questions are:
> >> lima10#sh int atm2/0
> >> ATM2/0 is up, line protocol is up
> >>   Hardware is ENHANCED ATM PA
> >>   Description: Conexion: TeCo Fibra oscura - ID: 5601631
> >>   MTU 4470 bytes, sub MTU 4470, BW 155000 Kbit, DLY 80 usec,
> >>      reliability 255/255, txload 88/255, rxload 187/255
> >>   Encapsulation ATM, loopback not set
> >>   Encapsulation(s): AAL5
> >>   4095 maximum active VCs, 1 current VCCs
> >>   VC Auto Creation Disabled.
> >>   VC idle disconnect time: 300 seconds
> >>   0 carrier transitions
> >>   Last input 00:00:01, output 00:00:00, output hang never
> >>   Last clearing of "show interface" counters 02:19:19
> >>   Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops:
> >> 0
> >>   Queueing strategy: Per VC Queueing
> >>   5 minute input rate 113886000 bits/sec, 27138 packets/sec <--- This
> >> info talks about packets so I assume it's IP traffic, right?
> >>   5 minute output rate 53787000 bits/sec, 17195 packets/sec
> >>      224971412 packets input, 2850386068 bytes, 0 no buffer
> >>      Received 0 broadcasts, 0 runts, 0 giants, 0 throttles
> >>      0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort
> >>      138570460 packets output, 2615814337 bytes, 0 underruns
> >>      0 output errors, 0 collisions, 0 interface resets
> >>      0 output buffer failures, 0 output buffers swapped out
> >>
> >> If the input bits/sec referrs to IP traffic, in order to get my real
> >> throughput over the link I should sum the ATM overhead.
> >> So if the 'show int' shows me a peak of 125Mbps, my "real" traffic
> >> SHOULD be 125/(48*8)*(53*8)=138.
> >> Anyone has a PA-A3-OC3SMI handling more traffic?
> >>
> >> Thanks in advance.
> >>
> >> Regards,
> >> Alejandro.
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
> 
> 
> 
> --
> GPG/PGP --> 0xE736BD7E 5144 6A2D 977A 6651 DFBE 1462 E351 88B9 E736 BD7E 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>