CEF network design (including cross-border fibre issues)

The SEEFIRE project is co-funded by the European Commission under the FP6 IST contract no. 15817

www.seefire.org

SEEFIRECEF network design

(including cross-border fibre issues)

Stanislav ŠímaCESNET

SEEFIRE technical meeting – Sofia, July 15, 2005 2

Network construction and parts(what is really our task)

Collins COBUILD English Dictionary for Advanced Learners: „If you construct something such as an idea, piece of writing or system, you create it by putting different parts together.“

For network construction we need to elaborate specifications of:Network servicesNetwork building parts (elements)Implementation of services by connecting elements

Elaboration of above specifications is called network designNetwork design is semantically similar to computer design or electronic circuit design, but big difference makes complexity and cost of some building parts:

Building elements, implementation and operation of large scale networks are procured

Procurement is long process (especially if public funding is used): e.g. 7-24 month (but legal system in some countries is not so restrictive to research)

Returns and corrections of decisions could be very difficult or impossible in such procurement process (details depend on procurement regulations in given country)

Result of one-way design proces is not optimal (in general case)….


Legacy procedure of REN design(ISP for R&E community)

Decide about REN PoPs needed and then:Procure SDH or lambda servicesProcure „ISP-like“ equipmentProcure network operation and maintenance

The first goal is service for researchers (research in networking is secondary)Main Advantages:

Relatively simple design and operation (we use large and matured building elements and services, with guarantee)Transparency for donators and users (element selection is the best in the current commercial sense)

Main Weakness and Risk:The „best in current commercial sense“ is not the best generally (monopoly or dominant position of some vendors prevents it)Construction simplicity and network reliability received are very expensive, i.e. network is based on outdated technology, adaptation to user needs is far from perfect, both CAPEX and OPEX are highVendors prevent some improvements of network after installation (to safe their exclusive delivery position)


New possibilities in REN design

Main source of innovation is optical technology Fibres and fibre lighting devicesFree Space Optics (up to 2.5 Gb/s per line) with microwave backup

Change of P2P services by P2P fibres is first step onlyNew types of transmitters, receivers, amplifiers, gratings etc. are available, some of them even with MultiSource AgreementsNew network services are enabled, for example E2E lightpath on demand, fibre switching, facilities based networking …

New architecture for cost-effectivness:Overcome distances by light only (see CESNET2 and GEANT2 core)OEO conversion and switches should be rare (use OADM and OXC)Few routers (2-4) per NREN are sufficient (see Surfnet6 and Canet4)


CEF Network approachto REN design

Decide about REN PoPs needed and then:Procure Dark fibres including optical first mileProcure Open (multi-vendor) network equipmentProcure Network Integrator (if external support for design, deployment and operation is needed)

The goal is still service for researchers, but role of research in design is strongerCEF Network approach is field proved:

procurement of dark fibres instead services: proved by many NRENs, RONs, National LambdaRail, partially GEANT2procurement of open (multi-vendor) equipment: proved by SWITCH (including external design support), CzechLight (including CESNET-made Optical Amplifiers), partially CESNET2Many metropolitan CEF RENs has experience with network integration by own staff

Save own improvement freedom and independency on vendorsSave own ability to quick return and correct design (including re-tendering etc.)Convince vendors in word and deed, that their task is changed

you need dedicated fibre instead of pacifier you will decide on devices suitable for network development, etc.


Main strategies for SEE NRENs development

Short characteristics of main possibilities for SEE NRENs:„Safe way“: Step-wise repeat NREN development seen in other countries (it is slow, quite expensive and preserves gap)„Brute force way“: Use big investment to make „the second instance“ of some leading NREN (this approach is limited mainly by missing first mile fibres and by unsufficient funds)

„Innovation way“: Strong use of research results to solve NREN design problem. Search for cost-effective technology successively deployed in testbeds and tested in NRENs (or prepared for future NRENs)


Possible Government Support for R&E

Remove telecommunication regulations for research and education

Simplify tendering regulations for research and education

Establish fibre lines for research and education as infrastructure donated by state (such as roads)

Support municipalities deploying first mile fibres for research and education

It helps to national R&E and to national development. It helps to EU development too (by better connecting of SEE research capacity)


GEANT2 design characteristics

GEANT2 is hybrid network: Lambda services (Layer2 OEO services) is offered to participants besides IP service (Layer3 services)Small changes in network topology, local movings of GEANT PoPs requested by nRENsMixed procurement of services and dark fibresExtensive comparisons of dark fibres to services costs, despite that reasons for dark fibres are strategic rather than financial, and result of comparison strongly depends on question:

How many lambdas will users need in next years???GEANT2 is CEF network in the sense, that core is on leased dark fibres and lighting equipment is owned by DANTESeparated procurement of transmission system and switching equipment: one vendor selected for both (no CEF-like open result)


Our example presented on TNC 2004 Rhodes


Initial GEANT2 topology(dark fibre core footprint)


GEANT2 and SEE countries

GEANT2 will be one of world-leading continent-wide R&E network using dark fibres (another one will be National LambdaRail after merging with Abilene2 in USA)Further work is strongly needed, for example:

GN2 procurement of dark fibres failed for connection of some countries, despite that dark fibres where available (good offer was missing), for example to Ireland, Portugal, Poland, Estonia, Latvia, Lithuania: re-tendering in GN2 is possible

Transmission and switching equipment are from single-vendor: multi-vendor interoperability problems must be solved on demarcation line GEANT2-NRENs

Important and feasible tasks: acquire international dark fibres to GN2 countries Romania, Bulgaria and Greece

prepare international dark fibres to remaining SEEFIRE countries


International dark fibre acquiring and using

Cross border fibre (CBF) issueGCBF are GEANT2 cross border fibres, connecting GEANT2 PoPsRCBF are Regional corss border fibres, connecting neighbours NRENs

In principle, by means NRENs fibres and RCBF is possible to implement European-wide lambdas and GCBF looks redundantSituation is of course much more complicated:

GCBF could be sometimes less expensive (imagine quantity discounts)Lambdas implemented on GCBF should be more reliableGCBF could be used for lambdas inside country (using OADMs etc.)

In general, CBF procurement results achieved by NRENs and by DANTE are (and will be) differentGN2 support for RCBF prepared


Cross border fibre issues (lines are examples only)


CBF connection needed

For GN2 members toBucharestSofiyaAthena

For GN2 observers toBeogradeSkopje

For GN2 non-members toTiranaSarajevo

We should agree in SEEFIRE now about required steps, including suggestion to GN2


Acknowledgement

Lada Altmannova for topology mapsCzechLight team members for collaboration

CEF network design (including cross-border fibre issues)

Documents

network construction

network designnetwork

improvements of network

network reliability

network servicesnetwork

strongercef network

computer design

simple design