OPS: Optical Packet Switches

Hiroaki Harai (harai@nict.go.jp)

National Institute of
Information and Communications Technology
Sep 8, 2006
Optical Network Testbeds Workshop 3
Why do we Need OPS?
¾ Internet Traffic in Japan: approx. 500 Gbps
¾ Peta-bps backbone future: doubled per year Å 500 Tbps in 10 years
¾ Electronic packet switch
¾ Year 2004: Throughput 640Gbps (16x40 Gbps)
¾ Lightpath networks
¾ Need fully meshed connections/ feasible?
¾ Need complex traffic engineering
¾ Important technology for bandwidth-assured applications
¾ OPS networks
¾ Provide extremely high-throughput
¾ Much larger bandwidth for switching (> 40 Gbps)
¾ O/E/O: 40Gbps Æ 64 x 622 Mbps bus, SERDES
¾ May need MPLS-like control (labels can be merged)
¾ Important to ubiquitous society

Sep 8, 2006 H. Harai (NICT) 2

Optical Packet Switching
¾ Data-path is all-optical (No O/E/O)
¾ Switch, Buffer 1xNNLabel
1x Label NNx1
x1Buffer
Buffer
Switch
Switch
¾ Increase data bandwidth 1xNNLabel
1x Label NNx1
Switch
Switch x1Buffer
Buffer
¾ Label lookup (i.e. forwarding)
1xNNLabel
1x Label
¾ Electronic parallel processing? Switch
Switch
NNx1
x1Buffer
Buffer

¾ Optical processing 1xNNLabel

1x Label
Switch NNx1
x1Buffer
Buffer
Switch

Optical payload
payload payload
payload

Routing
Routing
Make
Make a routingtable
a routing tablefor
for
forwarding procedure
forwarding procedure
Electronic payload
payload payload
payload
Electrical
serial
Forwarding
Forwarding Scheduling
Scheduling Optical
Determine
Determineoutput
outputport
port Avoid
Avoidpacket
packetcollision
collision
from
from the routingtable
the routing table Priority
Prioritycontrol
control
payload
payload header
header payload
payload header
header
Switching
Switching Buffering
Buffering
Switch
Switchthe
thepacket
packet Store
Storethe
thepackets
packets
to
to the appropriateport
the appropriate port ininappropriate
appropriatetime
time

Sep 8, 2006 H. Harai (NICT) 3

What should be Solved for OPS?
¾ OPS
¾ Increasing number of ports of optical switch
¾ Electronic: 16x16, 40Gbps Æ 640Gbps
¾ Optical: 128x128, 160Gbps Æ 20.48 Tbps
¾ 25 Waves Æ 500 Tbps
¾ Increasing speed of label lookup and buffer management
¾ Wire-speed operation
¾ Increasing number of labels looked-up
Under
¾ Several thousands (New L2 possibility)
developing
¾ More (L3 switching)
in NICT
¾ Increasing buffer size
¾ At least tens of fiber-delay-lines
¾ Decreasing guard time between packets
¾ Several nanoseconds
¾ OPS Monitor/Analyzer
¾ Bit error / Optical packet error

Sep 8, 2006 H. Harai (NICT) 4

Remaining Topics of This Talk
¾ OPS Prototype
¾ Optical label lookup
¾ Optical buffer
¾ Electronic buffer management

Sep 8, 2006 H. Harai (NICT) 5

NICT’s 40Gbps-based OPS Experiment
N. Wada, H. Harai,
F. Kubota, OFC
2003 (no. FS7).

Sep 8, 2006 H. Harai (NICT) 6

Optical Code based Ultra Fast Label
Processing
Packet format

Header
Payload data
(label)

Replace to the optical code (label)

Optical label has different modulation format with payload data

Optical label is physically distinguished from payload data

Optical hardware based label processing is available

Fully passive, ultra high-speed optical label processing

Sep 8, 2006 H. Harai (NICT) 7
Time Domain Optical Code Processing
--Measured Waveform at 8-chip, 200Gchips/s
Ref.) K. Kitayama,
N. Wada, IEEE
Photonic Tech.
Lett., vol. 11,
pp. 1689—1691,
Dec. 1999.

Auto-correlation
10ps/div.

Cross-correlation
Sep 8, 2006
10ps/div.
H. Harai (NICT) 8
Optical Fiber-Delay-Line Buffer
Control signal
¾ Different lengths of FDLs 44 11

55 Buffer
22 Manager
¾ Need at least tens of FDLs 33 0
¾ H. Furukawa, H. Harai, N. Wada, T
11 2T
N. Takezawa, K. Nashimoto, T. 44 11
22
Miyazaki, ”A 31-FDL Buffer 55
Nx(B+1) 3T
Based on Trees of 1x8 PLZT 22 switch 44 33 4T
Optical Switches,” to be 33 55
presented at ECOC 2006, no. Optical packets
Tu4.6.5, Sep 2006. (B-1)T
Discard

Sep 8, 2006 H. Harai (NICT) 9

 Optical FDL Buffering at 160Gbps
Scheduler
LN-SW LN-SW LN-SW
LN-SW
Buffer 1 1 LN-SW
LN-SW Output
port

LN-SW LN-SW LN-SW

Buffer 2 LN-SW 1 1
1 LN-SW
LN-SW

Switch 1 out Buffer 1 out Output port

Intensity (a.u.)

Single SwitchTime
Packet collision!! Avoidance of collision (2µs/div)
Noise
Intensity (a.u.)

Double Switches

Switch 2 out Time (2µs/div) Buffer 2 out

Source: N. Wada (NICT)
High-Performance Buffer Management
for Optical Fiber-Delay-Line Buffer
¾ Establish practical-scale high-performance management for FDL buffer
(1) Develop buffer management by parallel and pipeline processing
◇ For number of ports, time complexity of each processor is O(1)
◇ Parallel expansion of sequential (i.e. round-robin) scheduling
◇ N-times higher throughput than sequential scheduling
(2) Confirm feasibility of support for 128x40Gbps packet switch by FPGA
◇ 8 times higher performance than ASIC based router (16x40Gbps)
◇ IP packet granularity (64byte or more; 10 Gpps), variable length
(3) Prototyping 8-port buffer management system
Parallel and pipeline buffer management (N =8)
l1
P41
∆1 cf) H. Harai and M. Murata, IEEE/ACM
l2
P12
P42
∆2 Transactions on Networking, Feb. 2006.
P43
l3 ∆3
P13 P23
P44
l4 P24 ∆4
P14
P45
l5 P15 P25 P35 ∆5
P46
l6 P16 P26 P36 ∆6
P47
l7 P17 P27 P37 ∆7
P48
l8 P18 P28 P38 ∆8
Pq
Sep 8, 2006 H. Harai (NICT) 8-port buffer management hardware 11
Performance Comparison

NICT OPS NICT’s Top Data Electronic

Prototype (As of Sep, 2006) Router
IN/OUT ports 2 2* 16
Bit rate 160 Gbps 160 Gbps 40 Gbps
Label processing 800 Mpps/port 10 Gpps (at 125 Mpps/port
40Gbps) **
Scheduling 4 Mpps 10 Gpps 2 Gpps **
Buffer 2/port 31/port 16000/port

* Can scale with nanosecond optical switches

** Estimated data: Assumption of wire rate processing of 40byte-packets

Sep 8, 2006 H. Harai (NICT) 12

Conclusion
¾ We need high-throughput backbone network for ubiquitous society
¾ OPS will provide extremely high-throughput
¾ Switching bandwidth is not limited
¾ Buffer size is increasing
¾ Electronic scheduling is fast
¾ NICT has developed OPS but,
¾ Need more advanced devices (e.g., ns-switch) and systems

¾Thank you for your attention

¾ Acknowledgment
¾ N. Wada, H. Furukawa of Photonic Network Group in NICT for
valuable discussion, collaboration, and some slides in OPS

Sep 8, 2006 H. Harai (NICT) 13