follow up on gigabit either

k claffy kc at caida.org
Thu Dec 24 01:57:11 UTC 1998



hi

ok, summary as warranted:


==================
who responded:
==================

	folks from following spaces answered (some to list so you saw)
	(forgot to ask for permission to use names 
	so won't, but you who answered know who you 
	are since you've been karma++'ed today)

		ic.gc.ca
		canarie.ca
		gxn.net (gx networks)
		cisco
		bluestar.net
		linx.net
		surfnet.nl
		shrubbery.net
	 
==================
upshot:
==================
mostly all atm for now, though several folks pondering/aiming
for gigE for connectivity to NAP's like CA*Net3.
(usable (high density) gigether switches only recently available)

2.5 notable exceptions:  LINX and AMS-IX, and CaNET3.soon

==================
per instance details:
==================

LINX does: http://www.linx.net/
	   http://www.linx.net/tickets/1998_migrate.html

 	. 2 Packet Engines PowerRail 5200s
	  2 Extreme Summit 48s, linked w gigE

	. 2 physical locs in Docklands about 5km apart
	  dark fibre between, gigE happily over it.  

	. 2 redundant conns from two telcos to avoid backhoe risk.

	. interswitch traffic is >100Mbps on some links. 
	  no access >100Mbps to ISPs.
	  no members even on at gigE yet, 
	  biggest LINX members only push ~40M/sec i/o each,	
	  --> no problem so far


Amsterdam IX (AMS-IX) (since sept98) http://www.ams-ix.net/

	. ISPs connect on 10BaseT and 100BaseTX

	. may add Fast EtherChannel and GigE.  

CANARIE

	. working w Nortel, Ameritech, UoIllinois and iCAIR to 

	. build an "optical Internet exchange" in Chicago.  

	. initally for HPC (research) network (MREN, CA*net 3, SuperNet)

	. peering connections at 2.5 Gbps (OC-48) using 
	  any desired protocol: e.g., GigE, ATM, SONET, etc
 
	Canarie has fine paper on this stuff, fwiw (tnx billStA)
	
		http://www.canet3.net/Optical/architectural.htm


Steve also made exclnt point that the ATM vs 'promiscuous' medium
	have other inherent distinction of potential interest 
	to ISPs with policy/privacy/security concerns


hope this worth 
the aal5 headers it cost...


tnx again for data
k




More information about the NANOG mailing list