[181627] in North American Network Operators' Group
Re: How long will it take to completely get rid of IPv4 or will it
daemon@ATHENA.MIT.EDU (manning)
Mon Jun 29 21:01:55 2015
X-Original-To: nanog@nanog.org
From: manning <bmanning@karoshi.com>
In-Reply-To: <558F2AA3.1060608@satchell.net>
Date: Mon, 29 Jun 2015 17:58:38 -0700
To: Stephen Satchell <list@satchell.net>
Cc: nanog@nanog.org
Errors-To: nanog-bounces@nanog.org
actually, 1500 byte frames require a very different buffering technique, =
since you have so many in flight at a given time.
if your old enough, this equates to the 53byte ATM cells when the data =
rates were in the Megabit range.
manning
bmanning@karoshi.com
PO Box 12317
Marina del Rey, CA 90295
310.322.8102
On 27June2015Saturday, at 15:58, Stephen Satchell <list@satchell.net> =
wrote:
> On 06/27/2015 11:48 AM, manning wrote:
>> This is kind of like asking when we will stop using ethernet framing
>> (ethernet was designed for a 3Mbps transmission rate) yet we are
>> deploying 100Gbps networks. Still stuck on that 1500byte limitation.
>> When can we get rid of that?
>=20
> Speed has nothing to do with frame size. The 1500 byte limitation is =
more a function of the CRC algorithm. (Oh, the initial frame size was =
selected for 3-mbit Ethernet so that collision mitigation was =
reasonable.)
>=20
> Think about jumbo frames (9000 bytes) and their robust error =
detection. Research is being done in even larger frames, because the =
rule is that as your transmission rate increases, you should increase =
the frame size and use a FRC algorithm that detects all one-bit errors =
and most two-bit errors, at least.
