frame-relay traffic shaping and BE calculation

[Philip Moore]

Hello All

This is my first question to this group, so hello all. I am studying for CCIE, have just completed a 1 month bootcamp as a 'learning accelerator' and now preparing for CCIE written next week.
I had a question on frame-relay traffic shaping. The question is not concerned with implementation, only the calculation.

The values I have are:

physical line speed is T1 (i.e. 1.544mbps)
cir 768kbps
tc is default (125ms)

task is to calculate be and bc values

bc is calculated as follows:
cir * 125ms / 1000
768000 * 125 /1000 = 96000

for be, scenario states that be can be up to line rate for periods of up to 30 secs. my notes give the calculation as:

(AR - CIR) * TC / 1000
(1544000 - 768000) * 125 / 1000 = 97000

The answer given is Be = 46320000. I have no idea how they arrived at this figure.


Can somebody explain please?

[Scott Morris]

Someone's on *****...

An interesting math note though, the 46320000 divided by the magical 30 seconds they want yields a value of 1544000 or the "speed" of your T1 line. However, that just shows a lack of understanding about what gives there.

So the big part is that the "30 seconds" is really not workable. Because your router doesn't pay attention for that long to check whether something is matching or not. The whole concept about a "Tc" (Time Cycle or Time per Cycle) is that you have a FINITE measurable amount of time to be playing in. And 30 seconds is not an option!

So in your default 125 milliseconds, you have a finite amount of bits to play with period. This is based on your physical line speed. So let's work with that.

If your line speed is 1544000 (let's call it 1536000 because Cisco lies to you about that extra 8k). So you can send a MAXIMUM of 1,536,000 bits per ONE SECOND. You cannot send more than that. Period, end of story. So if your time cycle is 125ms (1/8 of a second), you can only send 192,000 bits per time cycle. If you want more, you'll need to buy a faster line. (Sorry, we aren't allowed to break the laws of physics no matter how cool our configurations are!)

Now we get to decide what we WANT to do (enter the world of CIR, or our "target" rate). CIR is our Bc, or Burst per Cycle, that we are measuring. So if you want to send 768000 bits per SECOND, then that mathematically equates to 96,000 bits per 1/8 second time cycle as you have measured out very nicely. Now... The hard part.

If our MAX was 192,000 bits per Tc, and you just allowed 92,000 of those, how many are leftover? 92,000. Now, your Be (Excess Burse PER TIME CYCLE) can be anywhere between 0 and 92000, depending on what you are looking to do. If you have other PVCs or other types of traffic to shape, you may want to wait until you have all of your "targets" (CIR) laid out. Because ALL TOGETHER, they cannot exceed your theorhetical maximum.

Well, let's just say that if you do try to do that, and the router config allows it, what you will end up seeing is lots of OUTPUT DROPS on your interface because no matter how cool you are, you can't beat physics.

So even if we thought about a 30 second timer, you'd take your differential there and multiply it by 30. But the router doesn't leave things that long. The count resets every Tc value. Good news is that you have a queue, but if your shaper things it will fit on the line, it's not queued, it'll be in the output drop section.

In the end, math rules.


HTH,
Scott

[Keith Barker]

I would agree that Be would be 97k

Just for fun, I did some options, like taking the Be, times 8 (for 8 time slices) times 30 seconds, based on the wording, but nothing I saw would reach that number they gave.
A 46Mbit Be on a T1 would be a bit (or a lot of bits) much.
Perhaps it is a major typo.


Best wishes,
Keith

[Philip Moore]

Dear All

Thanks for your detailed responses.
Yes I would agree that keeping within the laws of physics is always a good thing. I guess it's good that these questions with false answers appear to sanity check our thinking.
Much appreciated, cheerio


Phil