Cell Relay Archive

Cell Relay Retreat>List Archive>month:1997-Apr> msg00319



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: Cell Loss Statistics

  • From: "Eddie Hoffman" <edwinhof@ix.netcom.com>
  • Date: 28 Apr 1997 21:08:41 GMT

Cell loss is dependent on congestion and the vendors implementation of flow
control
in some switches as must as 1% cell loss can occur and 70%-90% loading.
This may
not seem to be much until you consider a 1500 byte Ethernet packet. If 1%
of 1500 is lost
then 15 cells are lost causing EPD (early packet discard) to throw out the
whole packet
and therefore a re-transmission etc. etc.

ER (Explicit Rate) and EFCI cannot control this cell loss. a new "credit"
based flow control
system QFC (Quantum Flow Control) will go a long way to preventing cell low
but cannot
completely stop it. Digital's FLOWMaster (credit based) can and does
prevent cell loss in
ABR applications, but nothing can prevent cell loss caused by
oversubscription of CBR
or VBR vc's.

Regards


Jim Holland <hollandj@bellsouth.net> wrote in article
<01bc510d$ceacff40$de0835cf@jims.bnr.ca>...
> Anyone have any idea where I can find an analysis on the characteristics
> of cell loss? I'm wondering if losing a cell is a "once in a blue moon"
> event
> or if when one cell is lost, the probability of losing other cells is
more
> probable.
> 
> Any ideas recommendations?
> 
> Jim Holland
>