At 10:58 AM 15-10-96 -0400, Avi Freedman wrote:
This is based on my feeling that HDLC is more efficent than ATM signaling/framing.
Agreed... But it's not a feeling, it's a fact :)
--bill
Avi
While this is certainly true, HDLC is a point-to-point protocol and not a network protocol, like frame relay, SMDS or ATM. And HDLC itself isn't quite enough, IMHO, you really need PPP. And the efficiency lost to ATM is not 40% as often claimed on this list, but rather it is 12% less efficient than PPP for TCP. 10% is the cell header overhead and 2% is due to modulo 48 padding, given actual traffic at FIX West as measured by kc at NLANR. --Kent We return to our regularly scheduled ATM tweaking program now in progress. :-)
While this is certainly true, HDLC is a point-to-point protocol and not a network protocol, like frame relay, SMDS or ATM. And HDLC itself isn't quite enough, IMHO, you really need PPP.
We run PPP on all non-frame-or-smds and less-than-DS3 links so that the customer (or ourselves, if it came to it) could switch to non- Cisco gear instantly. But we usually leave most cisco-cisco high-speed links at HDLC. My impression is that HDLC was the same efficiency - or moderately more so - than PPP. For what do feel that one *really* needs PPP? Or, to put it another way, maybe we're talking about different things. I'm talking about the HDLC *point-to-point* *network* protocol, as implemented by Cisco, not HDLC the low-low-level point-to-point protocol.
And the efficiency lost to ATM is not 40% as often claimed on this list, but rather it is 12% less efficient than PPP for TCP. 10% is the cell header overhead and 2% is due to modulo 48 padding, given actual traffic at FIX West as measured by kc at NLANR.
Quite believable.
--Kent
We return to our regularly scheduled ATM tweaking program now in progress. :-)
:) Avi
participants (2)
-
Avi Freedman -
Kent W. England