The impact of 100G and 1000G wavelengths on Internet architectures

[David Farber <dave () farber net>]

From: "Bill St.Arnaud" <bill.st.arnaud () canarie ca>
Date: April 17, 2008 10:23:41 AM PDT
Subject: [CAnet - news] The impact of 100G and 1000G wavelengths on  
Internet architectures

For more information on this item please visit my blog at
http://green-broadband.blogspot.com/ or http://billstarnaud.blogspot.com
-------------------------------------------


[Here is an excellent paper by European researchers working on the  
European
equivalent of US's GENI called Frederica. The main goal is to support  
the
virtualization of physical as well as logical routers and to create an
innovative service that provides the Network Operations Center and end  
users
with the ability to customize the configuration of its own dedicated IP
physical and/or logical network. The Software Architecture is built on  
top
of the Infrastructure as a Service (IaaS) Framework, a generalized  
approach
to the outcome of years of research under the UCLP Research programs  
funded
by CANARIE.  The significance is further elaborated on in my blog on the
Internet Evolution site --BSA]

<http://tnc2008.terena.org/programme/presentations/show.php?pres_id=98>

This document provides the principles behind the MANTICORE project,  
whose
main goal is to create an innovative service for the Network Operations
Centre and end users that allows them to customize the configuration of
their own dedicated IP physical and/or logical network  through the  
use of a
Web Services based Resource Management System. The motivation and the
concept of this logical IP Network Service will be detailed and the  
software
architecture of the first MANTICORE implementation will be explained  
(this
implementation is based on the IaaS Framework, Infrastructure as a  
Service
Framework). The conceptualization of the IP routing (internal and  
external)
is important when providing logical IP networks; thus, the basic ideas  
will
be explained in this paper. Future implementations will be able to  
extend
the concept to layer 0, 1, 2 (by integrating MANTICORE with the Argia  
and
Ether products) and/or layers above layer 3.


The impact of 100G and 1000G wavelengths on Internet architectures

<http://www.internetevolution.com/author.asp?section_id=506&doc_id=151266&; 
>

Researchers are now starting to explore the next big threshold of  
optical
networking -- 1000 Gigabit wavelengths. As with 100G wavelengths,
researchers are hoping to develop systems that fit within spectral  
envelopes
of existing 10G systems. This will make upgrades easy to implement as  
only
the transponders at the end of a network will have to be upgraded.
Nevertheless, 100G and 1000G wavelength systems are going to put  
enormous
stresses on the design of existing Internet networks.

1000G wavelength technology will most likely be premised on coherent  
optical
detection, which allows receivers to detect phase, polarization, and
amplitude combined with some form of complex modulation formats, such as
DPSK (differential phase-shift keying), DQPSK (differential quadrature
phase-shift keying) or OAM (Orbital Angular Momentum). These  
technologies
have long been in use in wireless systems and have dramatically  
increased
the available bandwidth on these systems.

However, routing manufacturers, in particular, will face significant
challenges in designing and deploying routing engines that will work at
these speeds. It is very unlikely that individual router interfaces  
will be
able to support 100G and 1000G interfaces. Instead, such optical  
wavelengths
will effectively be an inverse mux of many 10s, if not hundreds of 1G  
or 10G
"lightpaths." Each of these lightpaths will likely terminate on a  
virtual
router, with the possibility of many parallel routed networks being
deployed.

[..]In fact, with this type of architecture it would even be  
conceivable to
split up the forwarding table and assign different routing address  
blocks to
different sets of lightpaths. Not only would this optimize the network
architecture, it would also significantly reduce the size of the  
forwarding
table lookups on each virtual (or real) router.

[..]A good analogy is with airplane networks: A hub and spoke network is
much more efficient use of airline capacity, as passengers can be  
aggregated
at major hubs before being forwarded to their final destination. But
travelers generally prefer point-to-point flights as they can avoid the
chaos and inevitable congestion at the hub.

Point-to-point flights are much more expensive for the airline as they
cannot aggregate as much traffic on a single flight, and there is a much
greater risk of planes flying nearly empty. If airplane capacity could  
be
increased the same way and relatively at the same cost as we are doing  
with
100G and 1000G wavelengths, then hub and spoke networks would disappear
almost entirely.

Although Mike O'Dell has often been pilloried as one of the architects  
of
the Internet bubble, many years ago he grasped the significance of  
this type
of architecture and the impact of Internet traffic growth and resultant
network capacity. At the time he made his famous predictions of  
networking
doubling, O'Dell was thinking of ATM circuits. But the real impact  
will most
likely be felt with these new 100G and 1000G optical circuits. As we
increasingly move to point-to-point architectures network capacity will
increase at a much faster rate than the offered traffic load.


-------------------------------------------
Archives: http://www.listbox.com/member/archive/247/=now
RSS Feed: http://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com