Extending Contemporary Network Modeling Towards the Photonic Layer ˇıma Jan Kundr´at Stanislav S´
[email protected] CESNET, z.s.p.o.
April 20, 2015
Meet CESNET
I
Czech National Research and Education Network (NREN)
I
Grid Computing
I
Data Storage
I
Remote Collaboration
Děčín
NIX.CZ
Cheb
I
Future Internet
I
Cloud Computing
Beroun Dolní Břežany
Plzeň
...
Kuks
GÉANT Internet (6 Gb/s)
Krnov
Pardubice Litomyšl
Česká Třebová
Vodňany Budkov
Opava
Ostrava Tábor
Olomouc Jihlava
Temelín Kašperské Hory
PIONIER
Hradec Králové
Humpolec
Písek
AMS-IX
100 Gb/s 10 Gb/s 1–2,5 Gb/s <1 Gb/s
Dvůr Králové
Jenštejn PoděbradyLázně Bohdaneč Kostelec n. Č. L. Kutná Hora Ondřejov
Příbram Blatná
I
Turnov
Řež
Praha Mariánské Lázně
Jablonec n. N.
Ústí n. L.
Most
n×100 Gb/s n×10 Gb/s uzel (PoP) uživatel (user)
Liberec
Brno
Poněšice
České Budějovice
Telč Jindřichův Hradec Třeboň Nové Hrady
Vyškov
Zlín Uherské Hradiště Kyjov
Znojmo
ACONET
Lednice
Břeclav
SANET
Karviná
Photonic Networks
Photonic Networks I
Transmission without conversion between light and electricity I I I
I
Handy when “just” increasing speeds to Tbit+ is not enough I I I
I
Optical amplifiers All-optical switching No DSPs, no buffering, etc., along the path Low Latency Negligible jitter Determinism
Opens up new possibilities for advanced services I I I
Time and Frequency transfer Quantum crypto Remote sensing
Comparing Atomic Clocks of National Time Laboratories
Properties of Photonic Networks
I
The network provides a capacity to carry a light signal from point A to point B I I
I
Wavelength as a spectral range With a defined “quality”
Real-world thing subject to real physics I I I
Just some wavelength, not everything Attenuation as a function of frequency Different speeds of different wavelengths I
I I
I
. . . in different types of fibre
Non-linear effects ...
End result is “interesting” for a non-physicist
Network Modeling
I
We are modeling something else than an IP network I
I
Edge 6= Tuple(capacity , cost)
We also have to ask different questions I I I I I I I I
Usability for a given application Feasibility of carrying concurrent signals Signal degradation Regulatory compliance Energy consumption Reliability, resiliency, BER, HW life, fault tolerance Reconfigurability, “future-proofing” ...
Why Should We Model Photonic Networks? I
Foster adoption of Photonic Services and start offering them to the users
I
Help engineers design their networks Save money
I
I I I I
I
Provide interoperability between networks I
I
Bidirectional single-fibre transmission 100 G vs. n× 10 G Fibre Bragg Grating vs. Dispersion Compensating Fibre Reduce vendor lock-in through specification of desired properties Intercganging machine-readable description of the infrastructure(s)
Raise awareness of the problems
A Brief Overview of Modeling in Optical Networks
Network Description Language (NDL) I
Jeroen van der Ham et al, 2005
I
Based on RDF, “Semantic Web”, describing onthologies via triplets
Rembrandt3 I
Extended by CESNET’s M. H˚ ula with description of optical properties I
I
passBandTransmit,. . .
Not deployed anymore, obsoleted by NML
Network Markup Language (NML) I
Unification of competing modeling languages I
cNIS, NDL, perfSONAR,. . .
I
Developed within the Open Grid Forum (OGF) umbrella
I
Multi-layer, multi-domain
I
Encoding-agnostic Subject of active research
I
I I I
I
´ GEANT’s GN3+, GN4, US-based GENI Network Service Interface (NSI) Applications for Automated Provisioning
Optical/Photonic layer is out of scope of the current standard I
An extension is required
Conclusion & Future Work
I
Photonic Networks extend the set of applications which can make use of a network
I
Models have to describe different properties compared to traditional packet-/flow-based networks
I
Even the most popular models do not accommodate the description of photonic layers yet Ample opportunities for future work and bringing these to the users
I
I I I
Extending NML/NSI with support for Photonic Layer Verifying the model against a real network CAD tooling for engineers
Thank you! Questions? This work was supported by the Czech institutional funding of research by project Large Infrastructure CESNET LM2010005 and by the GN3 project under the EU FP7 programme. The authors would like to thank Jan Radil and Josef Vojtˇech for their valuable feedback.