Transcript Internet-2 Fall Meeting Optical Panel

Internet-2 Fall Meeting Optical Panel
Tuesday September 20th 2005 [email protected]
Electrical vs Optical Switching in
R&E Networks
 Capabilities
 Economics
 Reliability
Capabilities
Functionality
OEO
OOO
Transparency
(bit rate and protocol)
Yes
- wide range of signals
Yes
Low Latency
Yes
Yes
Single wavelength granularity
(I.e. no wavelength stranding)
Yes
Yes
Mesh Support (multi-degree)
Yes
Yes
- Blocking issues
Wavelength Translation
Yes
Yes – extra components
- Available (at a cost)
Grid Conversion
(e.g. CWDM to DWDM)
Yes
No
Protocol Performance
Monitoring
Easy + optical
power/OSNR
Optical power/OSNR
only
Wavelength Protection & Hitless
Maintenance
Easy
Ring – Easy
Mesh – More difficult
Economics
 The most cost effective solution is dependent on

The architecture of the node (East-West / Mesh)

How much traffic is required to add/drop at the node

How much traffic is Pass-through at the node
 OEO and OOO available in same node for most cost
effective solution over network lifespan.
Economics
Example 1: All traffic is ‘available’ for add/drop at the node but
can be pass-through if necessary – assumes 10GigE traffic
2-Degree - 100% add/drop node
Normalized cost (au)
1.2
1
0.8
0.6
0.4
0.2
0
0
5
10
15
20
25
30
# add/drop w avelengths
OEO
OOO
OEO is more cost effective at all capacities
35
Economics
Example 2: Fixed capacity of 32 wavelengths – %age of traffic
available at the node is variable – Assumes 10GigE traffic
2-Degree Node- 32 wavelength Capacity
1.2
Cost (normalized)
1
0.8
0.6
0.4
0.2
0
0
20
40
60
80
100
120
% add/drop traffic
OOO
OEO
OOO more cost effective with increasing pass-through traffic
Economics
Example 3: QUILT Network Model (October 2004) : Metro 2-1 node.
Mix of 2.5G(50%) and 10G(50%) traffic. 25% pass-through traffic (10G)
Quilt Node Metro 2-1 Example
1.2
Cost (normalized)
1
0.8
0.6
0.4
0.2
0
5
10
15
20
25
30
35
Network Capacity (wavelengths)
OEO
OOO
OOO / OEO equally cost effective at Day-1 :OEO
becomes more attractive as more capacity is added
Reliability
Reliability Factor (32  availability, 8 add/drop)
2
Reliability index (au)
1.8
1.6
OPTICAL
1.4
1.2
1
ELECTRONIC
0.8
0.6
0.4
0.2
0
4
5
6
7
8
9
10
11
Year (20xx)
Electronic switching is a well proven, field deployed technology.
With increase in market volumes (2008) Optical ROADM reliability
will be on a par with electrical reliability
Questions
 Are there performance issues related to OOO vs OEO ?

With OOO, the express (pass-thru) wavelengths will traverse a longer
optical path, though more optical amplifiers and filters than the
equivalent OEO solution so . .
 The required OSNR for the OOO case will be higher than OEO due to the
build up of non-linear impairments and a reduction in the dispersion window.
 OOO will experience more pass-band narrowing through the cascaded optical
ROADM – therefore more signal distortion effects must be managed within
the link engineering.
 OOO does however provide dynamic gain flattening and power equalization.

Optical amplifiers with advanced transient suppression and monitoring
features are required for OOO.
 The number of OEO nodes traversed will eventually be limited by jitter
accumulation along the path.

If design is within link budget limitations, both will perform equally well
Questions
 What are the advantages/disadvantages of OOO vs OEO related
to monitoring ?

An Electronic switch fabric has access to layer 0, 1 and higher
Performance Monitoring which provides for remote network
troubleshooting and for fast, effective protection switching.

OOO has access to optical power/OSNR measurement only.

Electronic switching fabrics allow loop-backs to be set remotely which
cuts the cost and time required for network troubleshooting – there is no
equivalent in an OOO system.
Questions
 What are the advantages/disadvantages of OOO vs OEO related
to switching especially in providing dynamic capabilities ?




OEO is by definition non-blocking and so multi-degree ring
interconnection and mesh networking are simple compared to OOO from
an engineering and control perspective. Where the number of passthrough wavelengths is small, OEO provides a more cost effective
solution.
In the cases where significant pass-through traffic is required (in east
west or multi-degree networks), OOO provides a more cost effective
solution than OEO.
For cases where the traffic is a mix of 2.5G and 10G, add/drop and
pass-through at a node, the cost differential is less well defined (as was
discussed re the Quilt network model !)
In large rings and multi-degree cases OOO requires complex path
dispersion management which is fiber type dependent.
Questions
 Which is better suited beyond 10G and what are the
advantages/disadvantages of each ?

From a cost perspective, OOO is better suited to 40G traffic when passthrough wavelengths are required at a node.

For 40G pass-through, OEO would cost a minimum of 4 times that of an
equivalent 10G port – switching electrically at native 40G rates is not on
the development radar

For OOO and depending on the 40Gbps+ technology used, there will be
a limit on the number of cascaded OOO hops in a chain (ring or mesh)
before pass-band narrowing starts to cause significant ISI.

40Gbps+ will require more strict control of dispersion in OOO systems
and will have a reduced OSNR tolerance compared to 10G systems.