From: Brennan_Murphy@NAI.com
Date: Thu Jan 09 2003 - 16:38:02 GMT-3
It'd be nice to have this in a spreadsheet with
the ability to convert to frame relay.
Anyone out there going to bite on this query
or should I take it to NANOG? :)
Thanks,
BM
-----Original Message-----
From: Chris Home [mailto:clarson52@comcast.net]
Sent: Wednesday, January 08, 2003 8:40 PM
To: Brennan_Murphy@NAI.com; ccielab@groupstudy.com
Subject: Re: Burst with Frame vs ATM
I believe with ATM it is actually a formula that describes the amount of
time that you can burst to PCR. So even if all bandwidth is available you
can only burst to PCR for a certain number of seconds. This sounds like the
leaky bucket you are referring to.
I can try to dig out the formula.
My understanding of it is basically this:
If you have a SCR of say 20 mbs and a burst of 50mbs you have to transmit
under 20 for enough time to fill the bucket with 50 megs worth of burst
tokens. When you have bursts of traffic it can be accomodated for only as
long as there are tokens iin the bucket. After that you will only get your
SCR and traffic above 20 will get dropped.
I forget the formula to figure out just how long you can operate a burst
speed. It is a simple formula but I cannot seem to pull it out of my head at
the momemet. It is usually not that long though if I remember right. When we
had a 10 meg (maybe it was 20) scr and peak of 50 the total time we could
use the 50 burst rate was under 3 seconds IIRC.
Maybe someone else will comment and provide more info and the formula.
----- Original Message -----
From: <Brennan_Murphy@NAI.com>
To: <ccielab@groupstudy.com>
Sent: Wednesday, January 08, 2003 2:23 PM
Subject: OT: Burst with Frame vs ATM
> Hi,
>
> I'm trying to understand the difference between Frame Relay and
> ATM when it comes to maximum throughput. Suppose I
> currently have a Frame circuit at 256K CIR with burst to 512K.
> The Frame vendor claims that you can almost always achieve
> the burst throughput....or at least consistently around 450K.
> This seems to bear out under ttcp testing.
>
> But if you migrate from Frame to ATM, you aren't necessarily
> getting the same throughput, right? I seem to recall ATM follows
> a leaky bucket algorithm so that a scr/pcr of 256/512 would
> actually give you less total bandwidth. Is this correct and
> is there an easy way to determine how to translate from
> frame to ATM so that bandwidth is roughly the same.
> I do understand that bursting in Frame is not guaranteed
> but most vendors do give you an idea of what to expect.
>
> I think I've got a decent understanding of converting Frame
> protocol overhead into ATM protocol overhead. It's the burst
> piece that is confusing me.
>
> I've looked around for a document on this subject but can't find
> one. I'm more used to Frame than ATM... It would be pretty
> cool if there were a simple Frame to ATM conversion calculator
> out there somewhere.
>
> Anyone have any recommendations?
>
> Thanks!
>
> -BM
> .
.
This archive was generated by hypermail 2.1.4 : Sat Feb 01 2003 - 07:33:46 GMT-3