Transcript Optical Waves Who needs them and why?

Optical Waves
Who needs them and why?
Some thoughts on how wave-based lightpath
services can be effectively employed in emerging
applications
Jerry Sobieski
Mid-Atlantic Crossroads
Presented at the Internet2 Fall Member Meeting
December 6th, 2006
Chicago, IL
“Optical Waves” vs “LightPaths”

Waves are subtly different from Light Paths
• LPs are often provisioned thru switched or aggregated network
infrastructure in ways that “privatize” traffic but often produce
unexpected performance issues (e.g. buffer overruns)
• LPs are often not optical at all (e.g. UTP Ethernet)
• LPs often require only a portion of the capacity of an optical link
(e.g. Fast Ethernet)

“Waves” imply a point-to-point optical signal that exists inside
a very narrow spectral band and that typically exists in
conjunction with many other similar waves on a single fiber.
• While the photonic concept would lead one to assume that
anything can be modulated onto a wave, there are many practical
and engineering constraints that prevent this – thus the
significant development that equipment vendors put into DWDM
transport systems.
Optical Waves as LightPaths



LightPaths := Circuits broadly [re-]defined
LPs provide a mechanism by which the user or application can
explicitly and apriori specify certain network service requirements
and then access those services– i.e. they enable predictable and
repeatable network services.
LPs (circuits) provide a mechanism by which the network can
effectively manage network resources allocated (and dedicated to)
many different users and applications.
• This also allows the network to set resources aside for use at some
future time… -> reservations and bookahead scheduling.


If your application has explicit network service requirements that
must be met for the application to function properly or well, then
your application should consider lightpath services to instantiate
those service requirements.
LightPaths are complementary to conventional IP network services
• IP networks employ circuits of some ilk underneath the IP layer
• The vast majority of LightPaths use IP within those paths.
• The LightPath is used to guarantee (or at least bound) the IP
performance
Who needs ‘em?
Isolation of application components


Waves/circuits (hybrid networks) can be used to segregate
traffic bewteen/among certain facilities in order to provide
that traffic with specific guarrantees
Example: A regional GRID cooperative may wish to
provision a full mesh of high capacity network links
between the participating computational clusters and data
storage facilities.
• The core components may be separated by geographically
significant distances or might otherwise share network
resources with unrelated and interfering traffic.
• Communications capabilities between clusters should be
predictable and repeatable in order to allow effective mapping
of the computational tasks/workflow across the available
resources
• Access to a large distributed data repository by different
computational layouts should be independent of other incident
network traffic.
Real Time and nearRT Applications


Some applications have a real-time or near real-time
component:
Example: Distributed Interactive Simulations; Gaming
• Agents participating in the simulation (potentially thousands)
require realtime notification of events that affect them; (Often
all events must be broadcast in order for automated simulation
agents to identify those that are relevant)


Architecting the distribution tree and providing adequate network
performance at each branch is critical
Example: Very Long Baseline Interferometry – Capturing
simultaneously sampled sensor data in realtime to
intermediate storage from many globally distributed radio
telescope facilities.
• And then subsequently streaming that captured data to
computational facilities for analysis
Traffic Engineering

In order to effectively utilize all available backbone
transport capacity, LSPs can be established and forwarding
adjacencies created to forward particular IP traffic flows
along alternate (underutilized) paths
- TE provides a degree of load balancing and more efficient use
of network resources
-
Large flows (near or exceeding backbone capacity) require
provisioned path(s) in order to provide the necessary
overall capacity end-to-end.
- Otherwise saturation of conventional IP forwarding paths will
occur and network users as a whole will experience desiccated
performance
Where do “Optics” play?

Emerging technologies are enabling the dynamic allocation of wave-based
connections to add raw capacity when and where it is needed
•
•


Dynamic on-demand photonics is doable in limited environments, but still
very complex and technically challenging
Well engineered DWDM systems can provide wave based transport for
almost any anticipated link technology
•
•
•

ITU compliant XFPs can be inexpensively deployed into end user facilities
(Ethernet switches, NICs, etc.) for activation when needed
Tunable lasers, tunable filters, multi-degree ROADMs allow hands-free, high
service velocity provisioning
Ethernet, SONET/SDH, IB, FiberChannel,
Even 2R regeneration of non-standard digital streams such as SMPTE-292 (raw
HDTV)
Forward Error Correction found on transponder cards can add significant distant or
overcome significant impairments
Features such as FEC can provide “soft failover” capabilities - indicate
impending failure before a hard (unrecoverable) failure occurs.
•
These types of features, when integrated with higher layer routing and VCAT +
LCAS capabilities can allow the network to groom at-risk circuits from bad paths to
better paths hitlessly and before unrecoverable bit errors are expressed to the
user.
Summary

Waves - as connection oriented
LightPaths- can provide customized
collaboration and application
environments
• Predictable
• Repeatable
• Schedulable
• Deternministic Network Performance

(when provisioned well)