...Skeeve -- Skeeve Stevens, CEO/Technical Director eintellego Pty Ltd - The Networking Specialists skeeve at eintellego dot net / www.eintellego.net Phone: 1300 753 383, Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 / skype://skeeve www.linkedin.com/in/skeeve ; facebook.com/eintellego -- NOC, NOC, who's there? -----Original Message----- From: arin-ppml-bounces at arin dot net [mailto:arin-ppml-bounces at arin dot net] On Behalf Of michael.dillon at bt dot com Sent: Saturday, 6 February 2010 1:18 AM To: arin-ppml at arin dot net Subject: Re: [arin-ppml] FW: [sig-policy] prop-083: Alternative criteria for subsequentIPv6 allocations > The crux is that even if aggregation of announcements as a > policy is dropped, there is no policy that satisfies the need > of LIR's to be able to build non-connected IPv6 networks in > the APNIC region. > > With some basic justification, an LIR should be able to gain > subsequent /32 allocations for the purpose of remaining 'as > routable as possible'. I think you should consider some editing of your policy proposal because I can't understand it. But, above, you seem to be suggesting that a network operator who is operating multiple disconnected IPv6 networks should be able to get multiple discrete /32 allocations. This seems reasonable to me. However before making a final determination I would like to see some input from the operators of large Internet data centres. It seems to me that a data centre is rather like a multi-tenant building, and since a tenant in a multi-tenant building deserves a /48 assignment, it seems reasonable that a data centre operator should give everyone a /48. That means if you rent a cage, you get a /48, if you rent a rack you get a /48, and if you colocate a 1U server with a single ethernet connection, you get a /48. Let's face it, lots of these servers will have multiple cores and could be running dozens or hundreds of virtual machines with a complex internal network created by softrouters and softswitches. See <http://openvswitch.org/> and <http://conferences.sigcomm.org/sigcomm/2009/workshops/wren/slides/tripa thi-crossbow-wren09.pdf> for examples of the kind of things that people are doing. Out of that analysis we might learn that a block size less than /32 would be more appropriate for these disconnected data centres. Perhaps a /36 per centre would work, and they could be allocated out of a block designated for /36 blocks. --Michael Dillon _______________________________________________ PPML You are receiving this message because you are subscribed to the ARIN Public Policy Mailing List (ARIN-PPML@arin.net). Unsubscribe or manage your mailing list subscription at: http://lists.arin.net/mailman/listinfo/arin-ppml Please contact info at arin dot net if you experience any issues.