Re: A Transport Protocol Without ACK

["Matt Wakeley" <matt_wakeley@agilent.com>]

Y P Cheng wrote:

> The fibre
> channel adapters make reliable delivery, lost packet detection, and
> retransmission without TCP/IP.

No, they don't.  That is the fallacy of fibre channel. Especially class 2 is
thought to be "reliable".  It's simply not true.  The FCP-2 group has been
working on this for a couple years now, and it's still not finished.  Problems
really arise when trying to address "out of order" frame delivery.

Class 3 (the most common implementation of FC) is "connectionless" - you ship a
frame and hope it gets there.

Class 2 (with its ACKs) over advertises EE credit to each node that connects to
it.  It ultimately falls back to BB credit, which causes a pile up of frames
from the receiving node all the way through the fabric.  And with this model,
you only know that the frame got to where it's going.  If you don't get the ACK
back, you don't know if it got there or not.  And there is no mechanism to
request retransmission of a particular frame.

-Matt