SERENATE WP 3 Equipment Study WP 3 Equipment

SERENATE WP 3 Equipment Study

WP 3 (Equipment) Mission • A study of into the availability and characteristics of equipment for next-generation networks • More specifically, to look at developments of routing, switching and transmission equipment over the next 2 -5 years • Efforts concentrated on addressing: • • higher capacities (i. e. 40+Gbps) optical technology for switching and transmission developments in network management and the control plane impact on network architectures

Work Plan • Bi-lateral meetings with 11 equipment vendors and 2 university research labs during November and December 2002 • Equipment vendors: • Alcatel, Calient, Ciena, Cisco, Corvis, Juniper, Lucent, Nortel, Photonex, Tellium, Wavium* • e. g. cross-section of players from the well established to the newly started-up • University research labs: • University of Essex (Prof Mike O’Mahoney) • University of Ghent (Prof Piet Demeester) • Attempted to contact a number of other vendors who either did not respond or declined to take part

Questionnaire • A confidential questionnaire was developed to: • set the context of the bi-lateral meetings for the vendors (questionnaire was sent to them in advance) • provide some guidance for discussion during the meetings • Questionnaire addressed the following broad topics: • • 40+Gbps capabilities (drivers & technical difficulties) Device scalability New control plane paradigms switching and transmission developments

The Team • DANTE (leader) • TERENA • NREN Consultants from: • CESNET • PSNC • HEANet

Routing developments • Scalable to terabits, in multi-chassis platforms • require experts for installation? • 40 Gbps backplane support and slot capability exists today • 40 Gbps interface capability “planned”, but not yet available • SONET/SDH framing • coloured interfaces ? • Maybe but proprietary solutions

Router functionality • • • Differentiated Classes of Service multicast ipv 6 MPLS-based VPNs G-MPLS • following standards, expected improved interoperability • interdomain functionality still questionable

Switching developments • Optical Cross Connects (OXC) • Essentially digital cross connects with optical interfaces • Also called O-E-O switches • Photonic Cross Connects (PXC) • Devices that work entirely in the optical domain • Also called O-O-O switches

OXCs • Scale to hundreds of Gbps, using advanced ASICs • bandwidth grooming performed with proprietary techniques (not interoperable!) • GMPLS developments: implementations still have proprietary features, although some interoperability demonstrated • Colour DWDM interfaces: some proprietary examples • Will only work with same vendor’s transmission equipment

PXCs • All the rage a few years ago • Now all but a few vendors have either moth-balled their products or gone out of business • Can save on O-E-O conversions hence: • footprint • power consumption • cost • Bit rate, protocol & wavelength independence • Scale up to tens of Tbps switching capacity • Earliest envisaged use (of smaller products) as “remotely manageable optical patch panel”

PXC difficulties • re-routing of wavelengths leads to optical channels in different route length: amplification and dispersion control difficult • Qo. S hard to control • Need external TDM devices for BW grooming • interoperability

Transmission equipment • Capabilities of current state-of-the-art DWDM transmission equipment far exceeds BW needs for the next few years • Little vendor interoperability amongst transmission components nor is this likely to happen • nature of systems is proprietary and analogue • only “standards” are ITU grid wavelength specs • may be possible to mix & match for low capability systems (CWDM, lower bit rates e. g. 2. 5 Gbps) • Every DWDM link is bespoke: • No “off the shelf” deployments

Reach • Very complex equation. Depends on: • • • fibre type (G. 652, G 655…. ) capacity of each wavelength number of wavelengths amplification technology used transmission technology used FEC

Reach with Nothing In Line (NIL) • Using pre and post amplification • up to 280 km at 2. 5 Gbps (Cesnet experience) using RAMAN • using cheaper equipment (1 GE, EDFA amplifier) result was 189 km • 350 km demonstrated

LH and ULH systems • LH (to 1, 500 km) and ULH (to 4, 000 km) require amplification at each span (40 -100 km) • larger spans if less wavelengths (200 km) at 10 Gbps • RAMAN • FEC • 40 Gbps can reach 1, 000+km with 80 km spans • RAMAN • dispersion compensation at receiver • PMD mitigators (depending on fibre)

Some conclusions • 40 Gbps: first in LH? Some say metro-area…. . • depends where economics work in its favour • common view is that main driver will be router interface cards • still more than 4 x cost of 10 Gbps… • 80 Gbps, 160 Gbps technically possible, but in labs. (600 Gbps has been demonstrated)

Network architectures • New set of requirements for research networks: • traditional users at large • relatively limited number of users with requirements for limited coverage but very high capacity • accessibility of “cheap” wavelengths in some parts of Europe • developments of transmission technology • in some cases NRENS can “do better without carriers”

Network Options • Traditional IP (layer-3) only • mixed layer-2 + layer-3, with OXCs (or PXCs) • Owned fibre network • a mix of all

Network Management and control • Different network architecture means NRENS will manage new network elements • Use traditional telco-style management systems as well as SNMP-based ones • Integration of two needs work! • G-MPLS has potential for integrated control, but interoperability and implementations conformant to standards not there yet