Pacific Wave Evolution

John SilvesterUniversity of Southern California

APAN 33, Chiang Mai, ThailandFeb 14, 2012

University of Southern California and University of Washington develop local peering exchanges for their communities (remnants of the NSF regionals) due to proximity to major carrier hotels

Emergence of the next generation of REN’s (National and Regional) – purchased services

vBNS -> (Abilene) Internet2Calren (CENIC), PNWGP, …

Early Days

.com collapse (c 2000) made acquisition of dark fiber leases possible

CENIC and PNWGP acquire West Coast leases for dark fiber and light to provide multiple waves (CISCO equipment)

Concept of Pacific Wave emerges – distributed exchange on the West Coast of the US to expand reach of the Seattle and Los Angeles peering exchanges using low cost waves

Dark Fiber – the next Step

New Model: LA & Seattle Summer 2004

Greatly increased no-cost peering opportunities significantly reducing commodity costs for the participants

Concept grew to a larger set of participants over time and eventually led to the national commodity peering partnership with Internet2

Growth in importance of peering with R&E networks domestically and internationally

Science goes global

Pacific Wave Peering

Provided a uniform West facing (trans-Pacific) interface for International Peering◦ Cost reduction◦ Facilitating interconnects◦ Independence from landing points

(later added alternate fiber path to improve resiliency)

Number of connections and bandwidth increases (to 10G) led to addition of additional 10G PW backbone wave

International Trans-Pacific R&E Peering

Growth of regionals, trans-border connections and desire to direct connect to other partners results in a flattening of the network (moving away from the hierarchical structure)

Challenges the business models of the NREN’s

Network Flattening

Availability of dark fiber led to the development of new networks with the ability to provision lambdas, circuits, lightpaths, capacitated virtual circuits

Fast take up by “big science” eventually maturing (?) into NREN architectures and service models providing static and dynamic “lightpaths” (of various flavours)

Experimental “Lightpath” Services

Most exchanges are slow to provision the new services as they wait for agreement on what tools should be used for inter-domain lightpaths; and the fact that most International circuits do not support lightpaths

Exchanges mostly only support static lightpaths requiring manual interventions

But this is changing and SDN reinforces this shift 2012 may be the year of Dynamically reconfigurable

Inter-domain Lightpath deployment on a global scale. WIP (work in progress) at PW - see Thursday

presentation for more details

Exchanges slow to Adopt

Bandwidth upgrades◦ Internet2 => 100G◦ 40G in other NREN backbones and some

International links◦ Trans-Pacific 40G coming real soon now!

Provides the scaling needed for CVC’s to be economically feasible

Network research (GENI, SDN, …) needs (and can help provide) these capabilities

Other Drivers for Capacitated VC’s

CENIC/PNWGP and Internet2 partner to provision 100G on the US West Coast (win-win)

Allows for upgrades of the PW core to 100G (and even multiple 100G as necessary)

Plans afoot for dynamic lightpaths through PW

Rethinking of Gigapop as Exchange Point in the Internet2 network, with Internet2 becoming a provider of LP’s through the network

Recent Developments

One expectation of a GOLE is that it be non-blocking between participants.

This is hard (or expensive) for a distributed GOLE to provide

Presents an interesting challenge for PW ◦ Split into 3 GOLES◦ Continue to provide transparent (free) POP

interconnect (in a non-blocking way)◦ or <fill in the blank>

Distributed GOLE’s??