RE: [RPRWG] RPR Perf: simulation models
Hi Pinar,
Please see my answers below. I hope they address
the questions you have.
Adisak
-----Original Message-----
From: Pinar Yilmaz [mailto:pinar@xxxxxxxxx]
Sent: Thursday, October 11, 2001 11:22 AM
To: Adisak Mekkittikul
Cc: stds-802-17@xxxxxxxx
Subject: RE: [RPRWG] RPR Perf: simulation models
Hi Adisak,
I have a few more questions about the simulation models that will be
posted in the RPRWG web site.
- Could you provide the binaries in Opnet 8.0.C format? The current
models are in Opnet 7.0.B format and cannot be used with 8.0.C.
>>Adisak: We will work with OPNET to see how much it take to upgrade.
If time permit we will do it in this release, otherwise we will
include in the third release.
- In the documentation to be provided, could you describe the
interactions between MAC and MAC Client in some fashion similar to the
proposal description?
>>Adisak: This is a very good input. We will include in the document as much
as
detail of the interface in case that users want to create a
different
client model (FIFO, CoS, etc. But a complete document won't be ready
soon (we also have product development responsibility). We'll try to
get a rough draft out ASAP.
- Are you planning to provide a simple (non-VoQ) client to interface
with the RPR MAC?
>>Adisak: Well, if we have enough request and can steal time from writing
code.
- I tried to connect a raw packet generator node with an Ethernet
interface to the MAC Client via MAC Client's Ethernet interface. "Decap"
process complains about not being able to read fields in the packet.
Ethernet payload is just a stream of bytes with no fields (hence raw
packet). What kind of payload does the MAC Client expect? If raw packet
generators can be used without the whole TCP-UDP/IP stack the
simulations will run faster.
>>Adisak: I'll consult experts and see if there is any trick you can do.
Thanks,
Pinar Yilmaz
-------- Original Message --------
Subject: RE: [RPRWG] RPR Perf: simulation models
Date: Fri, 28 Sep 2001 12:51:28 -0700
From: "Mannix O'Connor" <mannix@xxxxxxxxxxxxxx>
To: "'Pinar Yilmaz'" <pyilmaz@xxxxxxxxx>
CC: stds-802-17@xxxxxxxx
Pinar -
We are making the model more user friendly as a result of your
suggestions
and will release a new version.
Please bear in mind that this model is for the purpose of evaluating the
performance of the RPR MAC BW Management (Fairness) algorithm only. The
model does not address packet format or topology discovery and
protection.
If there is a need to address these by simulation, rather than
analytically,
then we can discuss that separately.
Please see below for specific answers to your questions.
Our update plan includes:
· Frame Format
· Most recent RCM message format
· Clean-up OPNET default values such as queue sizes
· Include per destination PAUSE as presented in the Sept. meeting
Specific to your questions, here is what I got from the engineers.
· crr is the same as RCF
· Stack bit is not used (our error in copying from the
other model)
· CFI is an Ethernet .1Q field
· The default rate is 0 (best effort) and the weight is 1
(equal fairness). The unit of the rate is bit per second.
· The out of band fields are used for debugging purposes
and do not affect the bandwidth
· We currently don't have a raw packet interface. The
only interface in the model is Ethernet but you can add other
interfaces easily. It is an OPNET standard interface.
This time we will make an effort to write-up a simple
how-to-run description to be given along with the model.
Adisak