Beyond 10G

	Hong Liu
	February 14, 2005

Outline

	Today’s landscape and Future Direction
		10G and 40G
		Ethernet vs. Sonet
	Juniper’s Core Platform and OC-768c Solution
		Core Routing Platform Roadmap
		OC-768c and engineering challenges
	Beyond 40G

Today’s Landscape

	True 10G roll out and deployment:
		10GE customer and IntraPoP connections
		OC-192c/STM-64 InterPoP circuits
		DWDM transport
	40G (OC-768c) at very early stage
		WDM transport gear available, with some deployed
		One vendor with very expensive solution in trial
	40GE/100GE? standards yet to get off the ground

Future Directions

	Long term – convergence is key
		IP over MPLS, clear winner
		Ethernet with transport-friendly features?
		SONET with Ethernet cost/commodity advantages?
	Near/Medium term – each has its place
		Ethernet lacks OAM for transport/larger network
		SONET expensive for intra-cluster
		Virtual Concatenation, Link Capacity Adjustment Scheme and Generic Framing Procedure allows Ethernet over SONET

Applications

	Bigger bandwidth increments mean fewer devices to manage
	IntraPoP connections
		For short distances 40G is feasible now
		Ethernet’s lack of OAM not a barrier
	InterPoP connections
		40G single wavelength, over long distances requires fiber laid in the last few years
		DWDM with many 10G wavelengths to extend the usable life of installed fiber
		SONET is only near-term choice, due to OAM requirements of transport providers

Core Routing Platform Roadmap

	T-series platforms were built to accommodate multiple generations of PFEs
		Switch fabric capacity upgrades
		Packet forwarding engine technology upgrades
	Next-generation, 40G PFE in development now
		Leverages design points from current architecture
		Straightforward extension of existing 20G chipset

Juniper’s OC-768c

	Designed to work with installed T640s
	OC-768c = Next step for ultra-high BW pipes
		1999 - 2003 = deployment of 10G circuits
		2003 - 2005 = aggregate existing 10G circuits (link bundling)
		2005 / 2006 and beyond = deployment of OC-768c
	Initial offering will be SR (2km), link limited primarily by fiber chromatic dispersion
	Could be extended to IR (40km), link limited by chromatic and polarization mode dispersion.

Beyond 40 Gigabit Speeds

	Engineering lessons:
		Multiplexing is most cost-effective solution, at least initially
	Engineering challenges:
		Forwarding with intelligence at these speeds
		Memory requirements – bandwidth and capacity
		Power and cooling issues – temperature stability of optics
		Chromatic and PMD (polarization mode dispersion) compensation are needed for links beyond 40G
	Fiber capacity/readiness for more than 40G  using single lamda
	As you approach technological limits, cost rises asymptotically

Thank you!

Hong Liu
hongliu@juniper.net