Slide 1

Overview: The Clouds Gather

	IPv4 address space exhaustion
	Continued growth in the DFZ
	Hardware limitations
		DFZ FIB approaching capacity of many popular routers
		RIB size now often exceeds router capacity
		Routing table recalculations exceeding time between updates
	IPv6 to the rescue?
		Should have been able to fix many of the concerns
		Very limited implementation
		No evident transition plan
	Where do we go from here?

It's the end of the world...as we know it

	But NOT the end of the world!
	We are not running out of address space...just a way to use it effectively
	Likely to lead to a market in address space
	Will massively grow the number of routes in the DFZ
	Will lead to increased breakage of peer to peer model
		not to be confused with file sharing P2P networking which is only a sub-category
		More likely to affect the R&E community than the typical user

IPv4 Address Exhaustion

	In May of last year ARIN announced the approaching exhaustion of IPv4 address space
		Less than 18% of the space remains unallocated
		Called for transition to IPv6
		Called for policy changes for handling the exhaustion
	IPv4 address space remains fairly sparse, though it continues to grow less so
		Several /8 prefixes are largely unused
		Many /16 prefixes are either unused, abandoned, or hidden on private/classified networks
		Many old /24 networks are no longer in use
	These networks will re-appear when they develop value

Continued Growth of the DFZ

	Default Free Zone (DFZ) growth has shown no tendency to slow
		DFZ exceeds 240,000 prefixes (or will when you get home)‏
		Will pass 244K routes in a month (More on this later)‏
	Myth that running out of allocatable space will slow the growth of the DFZ
		Scarcity will generate increased demand for ever smaller pieces of address space
		Demand will produce an exchange for address space
		Abandoned space will reappear as corporate asset

Hardware Limitations are approaching (1/2)‏

	Many popular routers in the DFZ are approaching Forwarding Information Base (FIB) capacity
		Cisco 6500/7600 routers (excluding Sup750-3BXL) can support 244K FIB entries
			That is just over a month form now! (What's in your TCAM?)‏
	RIB growth is pressing route processor capacity
		Highly configuration dependent
		May already exceed capacity
		May have reasonable headroom
		Some routers allow DRAM expansion to accommodate growth

Hardware Limitations are approaching (2/2)‏

	Route churn is approaching the point where routes will never completely converge
		This way lies madness (rather literally)‏
		Routing loops
		Black holed traffic
		Complete loss of state
	Routers typically don't use the fastest processors
		Upgrades will be possible and fairly painless
			Except to your budget!
		Optimization of protocol stacks may buy capacity
		May not be enough as the RIBs grow and churn increases

IPv6 to the Rescue?

	IPv6 will probably come, but when?
		Standards are many years old but there is almost no traffic
		Implemented on most hosts (and often enabled!)‏
		Implemented on most R&E and a few commercial nets
		Almost no services are available!
		Brokenness of IPv6 stacks discourages services implementation
			Reports on NANOG of 10% traffic drops when IPv6 is enabled for a service
			Web is too valuable to risk such losses

No viable transition plan

	You won't see services without IPv6 customers
	You won't see IPv6 customers without services
	No way exists for IPv6 customers to reach IPv4 services
		NAT-PT was the proposed solution
		NAT-PT has been deprecated
			RFC now listed as ÒhistoricalÓ
	Only solution is universal dual-stack capability
		But...

IPv6 to the Rescue (Part Deux)?

	Universal dual stacks will blow up the FIB
		IPv6 entries require 2-4 times the space in the FIB
		Equivalent to 720K prefixes in the FIB
		Will soon exceed the capacity of even very large routers
	Routing dual protocols will vastly increase CPU requirements to converge the RIBs
		IPv6 stacks are often not as carefully optimized as IPv4
		Twice as many routes to converge
		Increased complexity of multiple RIBs to converge?

Where do we go from here? (1/2)‏

	Retirement?
		Probably not that bad
	Watch the budget!
		New routers may be needed
			At least major upgrades required
			If you have Sup2 systems, things may get dicey
				When TCAM is full, new routes are passed to the SUP for forwarding
				Router dies an ugly death

Where do we go from here? (2/2)‏

	We need a viable transition to IPv6 now
		See http://www.civil-tongue.net/clusterf/
			Contribute ideas
			Prepare to feel Randy's wrath :-)‏
	Look for ways to limit FIB growth (e.g. LISP)‏
	Look around for unused address space
		Maybe you can sell it to get the budget for Sup750-3BXL upgrades
	Don't panic! (Note the large, friendly letters)‏
		The answer is 42

Recommended Reading

	Talks from NANOG41
		http://www.nanog.org/mtg-0710/bush.html
		http://www.nanog.org/mtg-0710/bicknell.html
		http://www.nanog.org/mtg-0710/farinacci.html
		http://www.nanog.org/mtg-0710/meyers.html
		Both slides and RealMedia recordings available
	RAM mailing list
		http://www.ieft1.org/mailman/listinfo/ram/

Summary

	Yes, we have a problem here
		The Galactic Construction Corps is not about to start an intersteller bypass (The world is not about to end)‏
	If nothing is done the FIB and RIBs will continue to grow
		This will at least require some re-design and some new hardware
	IPv6 is not just around the corner
		It is coming...but not this week
	IPv6 will not solve all of our problems

Thank You to (in no particular order)‏

	Kathy Aronson
	Randy Bush
	REM
	Dave Meyers
	Dino Farinacci
	Douglas Adams
	Vince Fuller
	Capital One Card
	Lots of others who slipped my mind
	You, who have to deal with these problems and listened to me babble on about it