Cell Relay Archive

[Charles Duncan <duncan@nebo.cs.byu.edu>]

Bernd Reuther wrote:
> 
> --
> Hi !!
> 
>     I am using The NIST ATM Network Simulator(version 2) for simulating
> my lab environment .There are some Input information for the ATM switch,
> which has to be enetred ,before running the simulator.But there is no
> info.in the manual for these factors,I shall be thankful if someone can
> help me :
> 
> ATM Switch :Input parameters ;
> # 1(EFCI), 2(NIST), 3(735R1): 2
>         # MACR Additive Increase Rate (Mbits/s): 0.5
>         # Target Rate (Mbits/s): 142
>         # Explicit Reduction Factor (ERF): 0.9375
>         # Measurement Interval in cells (N): 100
>         # Congestion Tolerance in cells (tau): 0
>         # Initial MACR rate (Mbits/s): 1.492
>         # VC Separator: 0.875
>         # Average Factor (AV): 0.0625
>         # DQT (Cells): 2000
>         # Major Reduction Factor (MRF): 0.95
>         # Down Pressure Factor (DPF): 0.875
>          #Congestion (True/False): 1
>          #% cell drop: 0
> 
> In the above there is shown some input values given in the sample
> configfile of the simulator.
> 

These values are what the ER ABR flow control algorithm uses to govern
its behaviour.  2(NIST) is NIST's version of ER and 3(735R1) is NIST's
combination of EFCI and ER.  The next 11 parameters are used to govern
the way these two algorithms react to congestion.  To really be able to
play with these and optimize them for your network configuration, you
need a really good understanding of ER and how it works (thus, IMHO, one
of the biggest problems with ER).  

The ones shown here are generic and commonly used.  We've modified some
of them to help increase the responsivness of the algorithm (at the risk
of increasing cell loss) by lowering N to 10 and increasing ERF to .95. 
Also, we increased the Initial MACR (Mean Allowed Cell Rate) to 142
since the BTE restrains the initital cell rate to 7.95 Mbits/sec anyway
(at least in our configurations).  

I don't claim to have a complete understanding of ER, but these setting
have worked for us on our configurations.

Charles

-- 

 Charles R Duncan         email: duncan@cs.byu.edu
 CS Department            http://nebo.cs.byu.edu/~duncan/
 Brigham Young University