CS 268: Optical Networks Ion Stoica April 21, 2004 (Based in part on slides from Ed Bortolini (Network Photonics), Ling Huang (UC Berkeley), Shivkumar Kalyanaraman (RPI), Larry McAdams (Cisco))

Big Picture Metro Long Haul Metro Access Access DWDM SONET Data Center DWDM SONET BPS2000, Baystack 450 & Passport 8600 (edge ethernet distribution & aggregation) OPTera Packet Edge on OPTera Metro 3000 series & OPTera OC-48 (ethernet over metro optical/Sonet) OPTera Metro 5000 series (ethernet over DWDM) Preside (provisioning & management software) Juniper & Shasta (collateral IP services & routing platforms) Contivity (VPN/secure, encrypted tunnel services) SONET SONET Metro Long Haul Metro Access Access

Overview Optical Transmission Dense Wavelength Division Multiplexing (DWDM) Synchronous Optical Network (SONET) Generic Framing Procedure (GFP)

Optical Transmission Transmitted data waveform Waveform after 1000 km PEAKING CROSS-OVER POINTLOWERED ZERO LEVEL BROADENED

Fiber Attenuation Telecommunications industry uses two windows: 1310 & 1550 nm 1550 window is preferred for long-haul applications Less attenuation Wider window Optical amplifiers 1550 window 1310 window l

Dispersion Dispersion causes the pulse to spread as it travels along the fiber Chromatic dispersion Light propagation in material varies with the wavelength Degradation scales as (data-rate)2 (Figure from http://lw.pennnet.com/)

Dispersion Modal dispersion Only for fiber that carry multiple light rays (modes) Different modes travel at different speeds Multimodal fiber used only for short distances

Overview Optical Transmission Dense Wavelength Division Multiplexing (DWDM) Synchronous Optical Network (SONET) Generic Framing Procedure (GFP)

DWDM λ1 λ2 λ3 λ4 λ5 λ16 16 uncorrelated vawelengths 2.488 Gbps (1) 1310/1510 nm 2.488 Gbps (16) λ1 λ2 λ3 λ4 λ5 λ16 1530-1565 nm ramge 1310/1510 nm 16*2.488 Gbps = 40 Gbps 16 uncorrelated vawelengths 16 stabilized, correlated vawelengts

DWDM System Design 40-80 km Terminal Terminal Regenerator - 3R (Reamplify, Reshape and Retime) 120 km Terminal Terminal Optical Amplifiers (OA) Terminal Terminal Terminal Terminal Terminal Terminal OA amplifies all ls

DWDM System Design 1 1 Optical Combiner 2 DWDM Filter 2 3 3 4 4 5 5 1550 1 1551 1 2 1552 2 3 1553 Optical Combiner DWDM Filter 3 4 1554 4 5 1555 5 Amplify 6 1556 6 7 1557 7 1310 nm Rx External Modulator Laser 15xx nm 1310 nm Reamplify Reshape Retime Rx Tx 15xx nm

All-Optical Switching Natively switch s while they are still multiplexed Eliminate redundant optical-electronic-optical conversions DWDM Fibers in Demux Fibers out Mux All-optical OXC

1-D MEMS MEMS: Micro-electromechanical systems 1-Dimensional array of micro-mirrors 1 mirror per wavelength Digital control; no motors

Input & Output fiber array Wavelength Dispersive Element Optical Switch 1-input 2-outoput illustration with four wavelengths 1-D MEMS with dispersive optics Dispersive element separates the ’s from inputs MEMS independently switches each  Dispersive element recombines the switched ’s into outputs Input Fiber Output Fiber 1 Output Fiber 2 Input & Output fiber array Wavelength Dispersive Element 1-D MEMS Micro-mirror Array Digital Mirror Control Electronics 1011

Optical Switch

Optical Add-Drop Multiplexer Add-drop one l Each l is associated with a fixed add/drop port Used to implement ring topologies DEMUX MUX

Overview Optical Transmission Dense Wavelength Division Multiplexing (DWDM) Synchronous Optical Network (SONET) Generic Framing Procedure (GFP)

SONET Encode bit streams into optical signals propagated over optical fiber Uses Time Division Multiplexing (TDM) for carrying many signals of different capacities A bit-way implementation providing end-to-end transport of bit streams All clocks in the network are locked to a common master clock Multiplexing done by byte interleaving

Synchronous Transport Signal (STS) First two bytes of each frame contain a special bit pattern that allows to determine where the frame starts Receiver looks for the special bit pattern every 810 bytes Size of frame = 9x90 = 810 bytes Data (payload) overhead 9 rows Synchronous Payload Envelope (SPE) 90 columns SONET STS-1 Frame

Encoding Overhead bytes are encoded using Non-Return to Zero high signal  1; low signal  0 To avoid long sequences of 0’s or 1’s the payload is XOR-ed with a special 127-bit patter with many transitions from 1 to 0 Duration of a frame is 125 µsec (51.84 Mbps for STS-1)

SONET Overhead Processing Three layers of overhead Path overhead (POH): end-to-end transport Line overhead (LOH): mux-to-mux transport Section overhead (SOH): adjacent network element Section Section Section Section Intermediate Multiplexer (ADM or DCS) DEMUX MUX Regenerator Regenerator Line Line Path

STS-1 Frame Format Two-dimensional: 9*80 = 810 bytes Or “Columns” 9 Rows Small Rectangle =1 Byte Two-dimensional: 9*80 = 810 bytes Time Frame: 125 µsec Rate: 810*8 bit/125 µsec = 51.84 Mbps For STS-3 only the number of columns changes (3*80 = 270)

STS-1 Headers Section Overhead (SOH) Path Overhead (POH): 90 Bytes Or “Columns” 9 Rows Line Overhead (LOH) Path Overhead (POH): Floating; can begin anywhere

Section Overhead (SOH) First 3 lines in the header Main functions Framing (A1, A2) Monitor performance Local orderwire (E1): select repeater/terminal within communication complex Proprietary OAM (Operation, Administration, and Maintenance) (F1) A1 =0xF6 A2 =0x28 J0/Z0 STS-ID B1 BIP-8 E1 Orderwire F1 User D1 Data Com D2 Data Com D3 Data Com SOH 3 B 87 B

Line Overhead (LOH) Last 3 lines in the header Main functions Pointer H2 Pointer H3 Pointer Act Last 3 lines in the header Main functions Locating payload (SPE) in the frame (H1, H2) Muxing and concatenating signals Performance monitoring Automatic protection switch (K1, K2) Switchover in case of failure Line maintenance B2 BIP-8 K1 APS K2 APS D4 Data Com D5 Data Com D6 Data Com D7 Data Com D8 Data Com D9 Data Com D10 Data Com D11 Data Com D12 Data Com S1 Sync M0 REI E1 Orderwire LOH 3 B 87 B

Path Overhead (POH) 1st column in SPE Main functions J1 Trace 1st column in SPE Main functions Info about SPE content Performance monitoring Path status Path trace SPE can start anywhere in the current frame and span the next one Avoids buffer management complexity & artificial delays B3 BIP-8 C2 Sig Label G1 Path Stat F2 User H4 Indicator Z3 Growth 1st STS-1 Z4 Growth P O H SPE 9 rows Z5 Tandem 2nd STS-1

N Individual STS-1 Frames STS-N Frame Format 90xN Bytes Or “Columns” N Individual STS-1 Frames Examples STS-1 51.84 Mbps STS-3 155.520 Mbps STS-12 622.080 Mbps STS-48 2.48832 Gbps STS-192 9.95323 Gbps Composite Frames: Byte Interleaved STS-1’s Clock Rate = Nx51.84 Mbps 9 colns overhead Multiple frame streams, w/ independent payload pointers Note: header columns also interleaved

STS-N: Generic Frame Format

STS-Nc Frame Format Concatenated mode: 90xN Bytes Or “Columns” Transport Overhead: SOH+LOH Concatenated mode: Same header structure and data rates as STS-N Not all header bytes used First H1, H2 Point To POH Single Payload In Rest Of SPE Current IP over SONET technologies use concatenated mode: OC-3c (155 Mbps) to OC-192c (10 Gbps) rates a.k.a “super-rate” payloads

Practical SONET Architecture ADM – Add-Drop Multiplexer DCS – Digital Crossconnect

Protection Technique Classification Restoration techniques can protect the network against: Link failures Fiber-cables cuts and line devices failures (amplifers) Equipment failures OXCs, ADMs, electro-optical interface. Protection can be implemented In the optical channel sublayer (path protection) In the optical multiplex sublayer (line protection) Different protection techniques are used for Ring networks Mesh networks

1+1 Protection Traffic is sent over two parallel paths, and the destination selects a better one In case of failure, the destination switch onto the other path Pros: simple for implementation and fast restoration Cons: waste of bandwidth

1:1 Protection During normal operation, no traffic or low priority traffic is sent across the backup path In case failure both the source and destination switch onto the protection path Pros: better network utilization Cons: required signaling overhead, slower restoration

1:1 Ring Protection Each channel on one ring is protected by one channel on the other ring When faults  loop around ADM ADM ADM ADM ADM ADM ADM ADM

Protection in Ring Network (Unidirectional Path Switched Ring) (Bidirectional Line Switched Ring) 1:1 Line Protection Used for interoffice rings 1+1 Path Protection Used in access rings for traffic aggregation into central office 1:1 Span and Line Protection Used in metropolitan or long- haul rings

Protection in Mesh Networks Network planning and survivability design Disjoint path idea: service working route and its backup route are topologically diverse. Lightpaths of a logical topology can withstand physical link failures. Working Path Backup Path

Path Protection / Line Protection Path Switching: restoration is handled by the source and the destination. Line Switching: restoration is handled by the nodes adjacent to the failure. Line Protection. Normal Operation Line Switching: restoration is handled by the nodes adjacent to the failure. Span Protection: if additional fiber is available.

Shared Protection Normal Operation 1:N Protection Backup fibers are used for protection of multiple links Assume independent failure and handle single failure. The capacity reserved for protection is greatly reduced. In Case of Failure

Overview Optical Transmission Dense Wavelength Division Multiplexing (DWDM) Synchronous Optical Network (SONET) Generic Framing Procedure (GFP)

Generic Framing Procedure (GFP) GFP provides a generic mechanism to adapt traffic from higher-layer client signals over an octet synchronous transport network (e.g., SONET) The GFP is an approach driven to standards by legacy equipment suppliers (e.g., Lucent, Nortel, NEC) in T1X1 and ITU. It intends to introduce a common data-link layer, on top of either SONET or OTN L1, so as to more efficiently accommodate many different L2 data signals. In particular, it allows statistical multiplexing of several bursty sources into a GFP frame. Initial standards for point-to-point GFP allow such multiplexing in the originating node on the ring. Later specification (ring GFP) would also allow other nodes to multiplex data traffic into the same frame. The GFP framing uses HEC and frame length indication for synchronization and delineation, similar to ATM and as opposed to a flag field as used in HDLC frames (e.g., in POS). Several vendors, and the ITU, envision the use of GFP and virtually concatenated SONET channels to circumvent SONET data deficiencies (granularity and burstiness) without eliminating SONET as a physical layer and for circuit bandwidth management. Packet-aware features may also be combined with GFP/VC to decouple VC bandwidth from the tributary interface (e.g., GbE) bandwidth, to provide flow control, and packet layer performance monitoring.

Generic Framing Procedure (GFP) Frame-mapped Need to know the client protocol Associate a length to each higher level frame Efficient: eliminate the need for byte stuffing or for block encoding (e.g., 8B/10B) Transparent No need to know the client protocol Less efficient; can transmit signal even when the client is idle The GFP is an approach driven to standards by legacy equipment suppliers (e.g., Lucent, Nortel, NEC) in T1X1 and ITU. It intends to introduce a common data-link layer, on top of either SONET or OTN L1, so as to more efficiently accommodate many different L2 data signals. In particular, it allows statistical multiplexing of several bursty sources into a GFP frame. Initial standards for point-to-point GFP allow such multiplexing in the originating node on the ring. Later specification (ring GFP) would also allow other nodes to multiplex data traffic into the same frame. The GFP framing uses HEC and frame length indication for synchronization and delineation, similar to ATM and as opposed to a flag field as used in HDLC frames (e.g., in POS). Several vendors, and the ITU, envision the use of GFP and virtually concatenated SONET channels to circumvent SONET data deficiencies (granularity and burstiness) without eliminating SONET as a physical layer and for circuit bandwidth management. Packet-aware features may also be combined with GFP/VC to decouple VC bandwidth from the tributary interface (e.g., GbE) bandwidth, to provide flow control, and packet layer performance monitoring.