The Optical Internet WDM/TDM fiber links Backbone ISP POP, CO Backbone

Slides:



Advertisements
Similar presentations
1 EE384Y: Packet Switch Architectures Part II Load-balanced Switch (Borrowed from Isaac Keslassys Defense Talk) Nick McKeown Professor of Electrical Engineering.
Advertisements

1 Maintaining Packet Order in Two-Stage Switches Isaac Keslassy, Nick McKeown Stanford University.
Chapter 9 Introduction to MAN and WAN
Lecture 4. Topics covered in last lecture Multistage Switching (Clos Network) Architecture of Clos Network Routing in Clos Network Blocking Rearranging.
Configuring a Load-Balanced Switch in Hardware Srikanth Arekapudi, Shang-Tse (Da) Chuang, Isaac Keslassy, Nick McKeown Stanford University.
Flattened Butterfly Topology for On-Chip Networks John Kim, James Balfour, and William J. Dally Presented by Jun Pang.
Wavelength-Routing Switch Fabric Patrick Chiang, Hossein Kakvand, Milind Kopikare, Uma Krishnamoorthy, Paulina Kuo, Pablo Molinero-Fernández Stanford University.
EE 230: Optical Fiber Communication Lecture 16 From the movie Warriors of the Net Active WDM Components and Networks.
Optical communications & networking - an Overview
Router Architecture : Building high-performance routers Ian Pratt
Scaling Internet Routers Using Optics UW, October 16 th, 2003 Nick McKeown Joint work with research groups of: David Miller, Mark Horowitz, Olav Solgaard.
Isaac Keslassy, Shang-Tse (Da) Chuang, Nick McKeown Stanford University The Load-Balanced Router.
A Scalable Switch for Service Guarantees Bill Lin (University of California, San Diego) Isaac Keslassy (Technion, Israel)
10 - Network Layer. Network layer r transport segment from sending to receiving host r on sending side encapsulates segments into datagrams r on rcving.
The Concurrent Matching Switch Architecture Bill Lin (University of California, San Diego) Isaac Keslassy (Technion, Israel)
Scaling Internet Routers Using Optics Producing a 100TB/s Router Ashley Green and Brad Rosen February 16, 2004.
1 Architectural Results in the Optical Router Project Da Chuang, Isaac Keslassy, Nick McKeown High Performance Networking Group
1 OR Project Group II: Packet Buffer Proposal Da Chuang, Isaac Keslassy, Sundar Iyer, Greg Watson, Nick McKeown, Mark Horowitz
Using Load-Balancing To Build High-Performance Routers Isaac Keslassy, Shang-Tse (Da) Chuang, Nick McKeown Stanford University.
1 ENTS689L: Packet Processing and Switching Buffer-less Switch Fabric Architectures Buffer-less Switch Fabric Architectures Vahid Tabatabaee Fall 2006.
048866: Packet Switch Architectures Dr. Isaac Keslassy Electrical Engineering, Technion The.
A Load-Balanced Switch with an Arbitrary Number of Linecards Isaac Keslassy, Shang-Tse (Da) Chuang, Nick McKeown Stanford University.
Scaling Internet Routers Using Optics Isaac Keslassy, Shang-Tse Da Chuang, Kyoungsik Yu, David Miller, Mark Horowitz, Olav Solgaard, Nick McKeown Department.
A WDM Passive Optical Network Architecture for Multicasting Services Student : Tse-Hsien Lin Teacher : Ho-Ting Wu Date :
Chapter 10 Introduction to Wide Area Networks Data Communications and Computer Networks: A Business User’s Approach.
1 EE384Y: Packet Switch Architectures Part II Load-balanced Switches Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University.
Optimal Load-Balancing Isaac Keslassy (Technion, Israel), Cheng-Shang Chang (National Tsing Hua University, Taiwan), Nick McKeown (Stanford University,
1 Chapter 10 Introduction to Metropolitan Area Networks and Wide Area Networks Data Communications and Computer Networks: A Business User’s Approach.
1 Chapter 10 Introduction to Metropolitan Area Networks and Wide Area Networks Data Communications and Computer Networks: A Business User’s Approach.
Localized Asynchronous Packet Scheduling for Buffered Crossbar Switches Deng Pan and Yuanyuan Yang State University of New York Stony Brook.
Nick McKeown CS244 Lecture 7 Valiant Load Balancing.
Optics in Internet Routers Mark Horowitz, Nick McKeown, Olav Solgaard, David Miller Stanford University
Fiber-Optic Network Architectures. OSI & Layer Model This Course.
Data Communications & Computer Networks, Second Edition1 Chapter 10 Introduction to Metropolitan Area Networks and Wide Area Networks.
1 Optical Burst Switching (OBS). 2 Optical Internet IP runs over an all-optical WDM layer –OXCs interconnected by fiber links –IP routers attached to.
Advance Computer Networking L-8 Routers Acknowledgments: Lecture slides are from the graduate level Computer Networks course thought by Srinivasan Seshan.
Univ. of TehranAdv. topics in Computer Network1 Advanced topics in Computer Networks University of Tehran Dept. of EE and Computer Engineering By: Dr.
Designing Packet Buffers for Internet Routers Friday, October 23, 2015 Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford.
Routers. These high-end, carrier-grade 7600 models process up to 30 million packets per second (pps).
Applied research laboratory 1 Scaling Internet Routers Using Optics Isaac Keslassy, et al. Proceedings of SIGCOMM Slides:
Packet Forwarding. A router has several input/output lines. From an input line, it receives a packet. It will check the header of the packet to determine.
1 Performance Guarantees for Internet Routers ISL Affiliates Meeting April 4 th 2002 Nick McKeown Professor of Electrical Engineering and Computer Science,
Interconnect Networks Basics. Generic parallel/distributed system architecture On-chip interconnects (manycore processor) Off-chip interconnects (clusters.
1 A quick tutorial on IP Router design Optics and Routing Seminar October 10 th, 2000 Nick McKeown
1 How scalable is the capacity of (electronic) IP routers? Nick McKeown Professor of Electrical Engineering and Computer Science, Stanford University
A Load Balanced Switch with an Arbitrary Number of Linecards I.Keslassy, S.T.Chuang, N.McKeown ( CSL, Stanford University ) Some slides adapted from authors.
Burst Transmission, Burst Switching and Dynamic Circuit Switching Prof. Leonid Kazovsky, PNRL Stanford presented by 리준걸 INC Lab. Seoul Nat’l.
Network layer (addendum) Slides adapted from material by Nick McKeown and Kevin Lai.
1 Building big router from lots of little routers Nick McKeown Assistant Professor of Electrical Engineering and Computer Science, Stanford University.
Chapter 3 Part 3 Switching and Bridging
EE384Y: Packet Switch Architectures Scaling Crossbar Switches
Network Resources.
Connecting Network Components
Weren’t routers supposed
Packet Forwarding.
Introduction to Metropolitan Area Networks and Wide Area Networks
Pablo Abad, Pablo Prieto, Valentin Puente, Jose-Angel Gregorio
Addressing: Router Design
Chapter 3 Part 3 Switching and Bridging
Static and Dynamic Networks
BASIC OVERVIEW OF AN ALL OPTICAL INTERNET
CS 740: Advance Computer Networks Hand-out on Router Design
Advance Computer Networking
Computer Networking A computer network, often simply referred to as a network, is a collection of computers and devices connected by communications channels.
EE382C Lecture 6 Adaptive Routing 4/14/11 What is tornado traffic?
Chapter 3 Part 3 Switching and Bridging
CS 6290 Many-core & Interconnect
Optical communications & networking - an Overview
Techniques and problems for
An Engineering Approach to Computer Networking
Presentation transcript:

The Optical Internet WDM/TDM fiber links Backbone ISP POP, CO Backbone AP MAN/LAN Access Point Backbone MANs, LANs & Access AP AP DS3 AP AP AP OC-3c WDM OC-12 SONET OC-12 Gigabit Ethernet FFTH, FTTC, PON or others.

Project Goals Determine where and how optics can be used best in the Optical Internet; Work across the traditional boundaries between networks, systems, sub-systems, device and materials research; Facilitate and foster interactions between network, system, device & material scientists.

Optical Internet: The Next Generation Optical Internet Systems Leonid Kazovsky Kapil Shrikhande Ian White Matt Rogge Joseph Kim (ST) Akira Nakamura (SONY) Hiroshi Okagawa (Fujitsu) Yoshiaki Ikoma (Hitachi) William J Dally Amit K Gupta Arjun Singh Brian Towles Nick McKeown Shang-tse Chuang Isaac Keslassy Sundar Iyer Greg Watson [Kyoungsik Yu]

Research Overview (A) Optical Internet Systems: Metropolitan Area Networks design (Prof. Leonid. G. Kazovsky) Scalable switching fabrics (Prof. William J. Dally) 100Tb/s router design (Prof. Nick McKeown) (B) Optical Internet Devices: All-optical MEMs switches and devices (Prof. Olav Solgaard) 2D Arrays of optical modulators (Prof. David Miller) Low-cost Long wavelength VCSELs (Prof. James S. Harris)

The HORNET architecture . . . Electronic switching, scheduling, queuing etc. sub-systems Tunable Tx FIXED l RX MAC LAN side equipment

The HORNET Architecture Single (tunable) transceiver per node, no central switch, full-mesh connectivity Node throughput potentially = bit-rate of transceiver (minus overhead) Big cost savings Data optimized packet transport A “new” architecture, many open issues

Project Status: Highlights Experimentally demonstrated the following: Tunable transmitter (2.5Gb/s) Control channel MAC Burst mode receiver (2.5Gb/s) Survivable architecture Theoretical study: optical amplification, fairness protocols Latest experimental results: Circuits over HORNET Complete testbed demonstration with MAC, reservation protocols, and continuous mode bit-error-rate tests at 1.25Gb/s 13 conference papers (3 invited), 7 journal papers and several trade journal articles

Node implementation Data and control processing sub-system CW, CCW Control channel O/E + SERDES (MAN-side) 155/622 Mb/s data-rates CW, CCW Data SERDES & I/O (MAN-side) 1.25/2.5Gb/s data-rates Data and control processing sub-system Fast tunable transmitter sub-system DAC FPGA FPGA PLD LAN-side interfaces (SONET, Ethernet) DAC DAC

Control channel MAC Time Slotted Network: Reservations: Nodes align (data) packet transmission to slot boundaries. Control channel signals time slots. Dispersion shifted fiber aligns slot on control and data channel. Reservations: O/E/O conversion at every node. Control channel indicates (Reservation, Occupancy) for all data wavelengths (2 bits/l). Nodes read control channel to determine openings on data channels.

Converged Network: Reservation Protocol 2000 slots live in a 100km ring (a frame), for 2.5Gb/s channel bit-rate and 64Byte slots. Reserve ‘X’ slots/frame: fixed bit-rate circuit. Handshake between source and destination, followed by source reserving required slots Circuit reservation in <= 3 RTTs, teardown in <= 2 RTTs Best-effort data is transmitted in unreserved slots

Research Overview (A) Optical Internet Systems: Metropolitan Area Networks design (Prof. Leonid. G. Kazovsky) Scalable switching fabrics (Prof. William J. Dally) 100Tb/s router design (Prof. Nick McKeown) (B) Optical Internet Devices: All-optical MEMs switches and devices (Prof. Olav Solgaard) Quasi-transparent optical switching devices (Prof. David Miller) Long wavelength VCSELs for LANs/MANs (Prof. James S. Harris)

Scheduling Optical Switches Emulation Architecture A switch with non-zero switching time can emulate a zero overhead switch plus delay. Existing “Exact” Algorithm # Configs Speedup Delay (cell times) 16,310 1 3,226,000 MIN 128 44 25,600 DOUBLE 256 2 51,200 Example for 128x128 switch with 200 cell time switching overhead New Algorithms Scheduling Algorithms New algorithms trade a small amount of bandwidth for greatly reduced batch size and switch delay

Worst-case Oblivious Routing Throughput-sensitive applications Emerging applications of interconnection networks are throughput sensitive: packet routing and I/O interconnect Designers want to accurately characterize the worst-case throughput of these networks Finding the worst-case Oblivious routing worst-case found by “building” a bad traffic pattern using a bipartite graph Intractable problem turned into a polynomial time algorithm

Design Space for Tori Improved load-balance, lower power and cost Reduced latency Region of feasible algorithms shown in gray. Two new algorithms developed that have optimal load-balance without sacrificing latency

RLB and GOAL Oblivious Global Load Balance Local Load Balance within Q Every source(S)–destination(D) pair has a short and long direction in each dimension. Obliviously pick a direction in each dimension favoring short over long. Defines a quadrant Q for routing to D. y k-y S D x k-x Local Load Balance within Q For RLB, route first to an intermediate node in Q, then to D. Randomize order of dimensions traversed. For GOAL, at each hop, pick the next dimension to be traversed adaptively (based on queue length).

Performance comparison Oblivious Adaptive Throughput for 1000 random traffic permutations Among oblivious algorithms, RLB has best average throughput. It cannot match VAL’s optimal worst case performace (but it gives much lower latency than VAL). GOAL is the only algorithm that can match VAL in the worst-case and performs best in the average case.

Latency and Cost 3 topologies considered: Torus – has higher cost and latency for most configurations Star – has slightly lower (3%) cost than Clos for large sizes, primarily due to distributed switching Folded Clos – has lowest cost and latency over wide range of network sizes, due to smallest diameter

Other Design Parameters Distributed vs. centralised switching Distributed switching Torus and Star (Cayley) networks Can reduce expensive optical links, only 1 required per node. Centralised switching Clos and hierarchical networks 2 long links required – to-and-from central switch cabinet Easier packaging and network manageability Channel sliced folded Clos network built with 256x256 2-bit routers

Research Overview (A) Optical Internet Systems: Metropolitan Area Networks design (Prof. Leonid. G. Kazovsky) Scalable switching fabrics (Prof. William J. Dally) 100Tb/s router design (Prof. Nick McKeown) (B) Optical Internet Devices: All-optical MEMs switches and devices (Prof. Olav Solgaard) Quasi-transparent optical switching devices (Prof. David Miller) Long wavelength VCSELs for LANs/MANs (Prof. James S. Harris)

Motivating Design: 100Tb/s Optical Router Switch Electronic Linecard #1 Electronic Linecard #625 160- 320Gb/s 160- 320Gb/s 40Gb/s Line termination IP packet processing Packet buffering Line termination IP packet processing Packet buffering 40Gb/s 160Gb/s Arbitration 40Gb/s Request Grant 40Gb/s (100Tb/s = 625 * 160Gb/s)

Research Problems Linecard Architecture Switch Fabric Memory bottleneck: Address lookup and packet buffering Architecture Arbitration: Computation complexity Switch Fabric Optics: Fabric scalability and speed Optics: Optical modulators Electronics: Low power optical links Electronics: Optical switch control Electronics: Clock recovery for intra-system links

The Arbitration Problem A packet switch fabric is reconfigured for every packet transfer. At 160Gb/s, a new IP packet can arrive every 2ns. The configuration is picked to maximize throughput and not waste capacity. Known algorithms are too slow.

Load-Balanced Switch External Inputs Internal Inputs External Outputs 1 1 2 1 N 1 N 1 N 2 CS Chang had a clever idea in order to make this non-uniform traffic uniform. The idea is that we could load-balance the arrival traffic among all the linecards. [explain on animation: on the right is the “switching” cyclic shift as before. Let’s focus on the left, the load-balancing cyclic shift: when a first cell arrives, it is sent to the middle-stage linecards according to the cyclic shift pattern. This pattern changes, and this is true again for the next cell] Thus, intuitively, the traffic is distributed uniformly among all linecards, and therefore the arrival traffic at all linecards should have about the same shape. This is what led CS Chang to prove that this simple two-stage switch architecture has 100% throughput for a broad range of traffic types. Load-balancing cyclic shift Switching cyclic shift 100% throughput for broad range of traffic types (C.S. Chang et al., 2001)

Implement as a Passive Optical Mesh R/N Passive mesh 1 2 3 1 2 3 R R Cyclic Shift Cyclic Shift 2R/N Passive mesh 1 2 3 Mention that there is a range of choices: 1) Complete Cyclic Shift 2) Complete Passive Mesh 3) Something in between: part cyclic shift and part passive mesh. Give the full form of WGR. No more arbitrations, no more reconfigurations!

AWGR (Arrayed Waveguide Grating Router) A Passive Optical Component 1 , 2 … N l 1 Linecard 1 Linecard 1 1 l 1 Linecard 2 Linecard 2 2 NxN WGR l 1 Linecard N Linecard N N Wavelength i on input port j goes to output port (i+j-1) mod N Can shuffle information from different inputs

WGR Based Solution Problems: We want N > 600 1 , 2 … NxN WGR Fixed Laser/Modulator Linecard 1 Linecard 2 Linecard N Detector 3 N-1 N^2 channels A linecard will use multiple lambdas which will evenly spread Impt slide: new way of considering switching Problems: We want N > 600 What if some linecards are missing?

From Linecard Mesh to Group Mesh

From Linecard Mesh to Group Mesh

Hybrid Optical Electrical Switch Fabric Static MEMS Electronic Switches Fixed Lasers Optical Receivers Electronic Switches 1 1 GxG MEMS Linecard 1 Linecard 1 2 2 LxM Crossbar MxL Crossbar Linecard 2 3 3 Linecard 2 1 Linecard L M M Linecard L GxG MEMS Group 1 Group 1 1 1 Linecard 1 2 Linecard 1 2 2 LxM Crossbar MxL Crossbar Linecard 2 3 3 Linecard 2 GxG MEMS Linecard L M M Linecard L 3 Group 2 Group 2 1 1 Linecard 1 Linecard 1 2 2 LxM Crossbar MxL Crossbar Linecard 2 3 3 Linecard 2 GxG MEMS Linecard L M M Linecard L M Group G Group G