82f7979ed69e6c09b16eed5fa586c620.ppt
- Количество слайдов: 50
Research in Communications Adam T. Drobot Telcordia Technologies, Inc. Piscataway, NJ 08854 In celebration of Ronald C. Davidson’s 40 Years of Plasma Physics Research and Graduate Education Princeton Plasma Physics Laboratory Princeton, NJ 12 June 2007
Agenda Introduction Background Technologies Research – Mobile Systems – Advanced Optical – Software Princeton, June 12, 2007 – 2
Introduction The Communication Industry – Size – $2. 7 Trillion worldwide, $0. 9 T U. S. – Complexity Wireline, Wireless, Cable, Fiber, Satellite, Powerline… Voice, Data, Video… – Pervasiveness – Global system with 1 B+ subscribers Convergence – Delivery of any service over any device or network, to any location, using common infrastructure Princeton, June 12, 2007 – 3
Introduction Networks – Core, Regional, Metro, Local Core Regional Local Metro Princeton, June 12, 2007 – 4
Introduction The Communication Industry Drivers – New Services – Efficiency of Operations – Deployment of Capital Direction – – Applications and Services IMS – IP Multimedia Subsystem Mediating Middleware and Core Capabilities Networks and Devices Princeton, June 12, 2007 – 5
Background Broadband Users per 100 pop. in 2000 Korea India Myanmar Canada China U. S Taiwan Lux. Czeck. Size shows population Princeton, June 12, 2007 – 6
Background Broadband Users per 100 pop. in 2003 Korea Japan Canada Taiwan India China Denmark Malta U. S Estonia Lux. Myanmar Niger Qatar Czeck. Size shows population Princeton, June 12, 2007 – 7
Background Broadband Users per 100 pop. in 2005 Korea Canada Taiwan Estonia Denmark Japan U. S Lux. Malta India China Myanmar Czeck. Qatar Niger Size shows population Princeton, June 12, 2007 – 8
Background Internet Users per 100 pop. in 2000 Iceland Taiwan Korea Malaysia India Czeck. China U. S Denmark Japan Lux. Qatar Niger Russia Size shows population Princeton, June 12, 2007 – 9
Background Internet Users per 100 pop. in 2003 Korea Iceland Japan U. S Taiwan Malaysia India China Czeck. Lux. Myanmar Niger Denmark Qatar Russia Size shows population Princeton, June 12, 2007 – 10
Background Internet Users per 100 pop. in 2005 Lux. Korea Taiwan U. S Japan Denmark Malaysia India China Qatar Myanmar Czeck. Russia Niger Size shows population Princeton, June 12, 2007 – 11
Technologies Basic Ingredients Computing & Processing Storage & Retrieval Networks & Interconnects – Wired, Optical, Cable, Wireless Software & Architectures Displays & User Devices Information & Content Princeton, June 12, 2007 – 12
CPU Speed (Intel) CPU Speed (GHz) Maximum Intel CPU Speed (IA-32) vs. Time Quad core Dual core Ref: http: //wi-fizzle. com/compsci/ and Intel Princeton, June 12, 2007 – 13
10 Transistors / Chip 10 10 9 8 7 6 5 4 3 2 1 10 DRAM cost / bit microprocessor 10 10 -1 -2 -3 -4 -5 -6 -7 -8 -9 Cost / bit (1995 $’s) Cost / function Chip Evolution http: //ocw. mit. edu/Ocw. Web/Electrical-Engineering-and-Computer-Science/6 -012 Fall 2003/Course. Home/index. htm Princeton, June 12, 2007 – 14
Memory: Price per Mb vs. Year (HD & RAM) http: //www. freewebs. com/adipor/Conclusion. pdf Princeton, June 12, 2007 – 15
Moore’s Law Pixels per Dollar Digital Cameras 10, 000 100 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 http: //en. wikipedia. org/wiki/Moore's_law Princeton, June 12, 2007 – 16
Computing Speed and Cost Speed Cost $M per MIP Princeton, June 12, 2007 – 17
Storage Capacity and Cost 10, 000 $100, 000. 00 1, 000 Capacity * 100, 000 $10, 000. 00 $1, 000. 00 10, 000 $100. 00 1, 000 $10. 00 Petabytes Shipped 100 Cost * $1. 00 10 $0. 10 1 Cost per Gigabyte $0. 01 1988 1992 1996 2000 2004 2008 Sources: IDC, "1999 Winchester Disk Drive Market Forecast and Review, “ Wall Street Journal, June 26, 2000 *Telcordia projected capacity 1988 -1994 & costs 2003 -2010. Princeton, June 12, 2007 – 18
Moore’s Law: MIT/IBM Raw From single transistors to CMP Itanium 2 Floating Point Integer Core Pentium 4 # transistors Itanium 2 Itanium Integer Core Renaming Decode Fetch Bus Trace Control Cache L 2 Data FPU Cache MMX/SSE 100, 000 P 4 P 3 L 2 Data Op Scheduling Cache 1, 000, 000 10, 000 P 2 Pentium 4004 486 1, 000 386 286 100, 000 8086 10, 000 8080 8008 4004 1970 Moore’s Law 1975 1980 1985 1990 1995 2000 1, 000 2005 Princeton, June 12, 2007 – 19
Content Delivery Quantity and Cost (USA 1960 2007) 1, 000 100, 000 10, 000 TV Radio 1, 000 Movies Trillion 100 10 words transmitted 1 per year Phone CATV Data Communication 0. 1 FAX 0. 01 0. 0001 Telegram 0. 001¢ 0. 1¢ 1¢ 10¢ $1 $100 Cost of transmitting 1000 words (1972 dollars) Princeton, June 12, 2007 – 20
Software Complexity Megabytes Size of Linux Kernel across Release Date Princeton, June 12, 2007 – 21
Software Integration Level Autonomous Software Web Services Integration Level Agents Numerical Libraries Unix Software Tools Class Libraries Single Functions 1950 1960 1970 1980 1990 2000 2010 2020 Year Princeton, June 12, 2007 – 22
Research • Mobility • Ubiquity 3 G and 4 G Networks Physical Sciences • Speed • Bandwidth • Immediacy • Relevance 5 G and 6 G Service Networks Princeton, June 12, 2007 – 23
Research 1. Wireless Princeton, June 12, 2007 – 24
MIMO Systems (Multiple Input Multiple Output Antenna Arrays) Application of signal processing expertise to achieve up to tenfold bandwidth increase for existing wireless spectrum. Princeton, June 12, 2007 – 25
MIMO Systems MIMO Basics – MIMO Capacity – Uninformed vs Informed Transmitter MIMO – Feedback and MIMO Advanced MIMO Receivers – Achieving Near-ML Performance: Space Time Bit Interleaved Coded Modulation & Iterative Detection – Soft Cancellation MIMO Measurements – Channel Phenomenology – High Spectral Efficiency Communications Cognitive Adaptive MIMO Testbed – Adaptive/Cognitive Behavior Princeton, June 12, 2007 – 26
MIMO Systems MIMO Channel Transmitter Nt transmit antennas For a narrowband transmitted signal, we can express the relationship between the signals on the transmit antennas and the received signals via the matrix H. Receiver Nr receive antennas Element hm, n represents the complex channel between transmit antenna n and receive antenna m. H is generally a function of frequency and time. Princeton, June 12, 2007 – 27
MIMO Systems Develop high spectral efficiency adaptive multi-user MIMO for practical communications. Advanced signal processing enables MIMO at low-to-moderate SNR in the presence of interference and mobility. MIMO will be will allow exchange of data at extremely high rates in bandwidth limited channels with sensitivity to power limitations and LPI/LPD constraints Telcordia’s MIMO measurement system – 8 element 450 MHz – 3 GHz dual polarized array is shown. Performance curves from overthe-air measurements showing zero BER operation at 25. 6 bps/Hz. Application of signal processing expertise to achieve up to tenfold bandwidth increase for existing wireless spectrum. Princeton, June 12, 2007 – 28
MIMO Systems Demonstration of a 25 information bps/Hz link using an 8 x 8 MIMO system in Spring 2005 This work was supported by the Army Research Laboratory C&N CTA Iterative Detection Results Using Telcordia’s ST-BICOM iterative detection receiver, which incorporates soft cancellation, multiple-layer iterative detection, we were able to demonstrate very high spectral efficiency at moderate SNRs. Princeton, June 12, 2007 – 29
Research 2. Optical Princeton, June 12, 2007 – 30
Example: OCDMA – for Future Access Networks Development and demonstration of prototype OCDMA hardware technologies, coding schemes and transmission systems operating up to 10 Gbit/s, with high spectral efficiency. Anticipated application of OCDMA technologies to high security environments, and to high-bandwidth PON’s. OCDMA-PON’s support the same maximum number of users with fewer lambdas than WDM, with the ability to gracefully add bandwidth and users to the network. Final phase of the program is a lab demonstration of a hybrid synchronous-downstream, asynchronous upstream OCDMA PON. • OCDMA: Optical Code Division Multiple Access • PON: Passive Optical Network • GPON: Gigabit Passive Optical Network Princeton, June 12, 2007 – 31
Example: OCDMA – For Beyond GPON A Compatible with existing fiber in the ground including the splitter B C Dx. A D Code converter Network operations & control Programmable OCDMA Code Conversion: Central Office to Home Princeton, June 12, 2007 – 32
Example: OCDMA – Physical Layer Security 1559. 0 nm 1560. 5 nm What makes it work – It is an overlay network operating in an existing WDM window in harmony with other traffic – It is robust to brute force attack: as good as electronic encryption – It offers inexpensive passive optical means suitable for 10 Gb/s and beyond data rates. How financial campuses are connected Star coupler Users within secure campus Dynamic scrambling coder Princeton, June 12, 2007 – 33
Example: OCDMA – Novel Network Topologies Physical layer security – Optical CDMA Architectural vision for securing DWDM optical networks with QKD Compatibility with WDM systems Useful for Multiple Access, PON’s, or enhanced security – ‘Quantum’ networking Quantum encryption Quantum key distribution Use quantum effects to provide ‘absolute’ or very high levels of security at the physical layer Focus on techniques compatible with real-world networks: Demonstrated coexistence of with multiple high power DWDM signals over metro-scale and greater fiber transmission links Quantum repeaters Quantum Testbed Optical Packet networks ‘Small buffer’ network performance Optical label switched technology Princeton, June 12, 2007 – 34
Example: OCDMA – Photonic Layer Security Compatibility with commercial WDM …this is an overlay technology “Security” for 100 Gb. E possible NOW …against exhaustive & archival attacks Optical integration Phase scrambling of inverse multiplexed tributaries creates large search space and hides the eye for resilience to exhaustive and archival attacks, respectively. OCDM …very compact technology allows wide application DWDM Optical Integration: Low cost, weight, size, and power consumption. Compatibility with DWDM allows deployment as a “secure” overlay network over existing networks and for multilevel security. Princeton, June 12, 2007 – 35
Example: Corning – Solid-Core and Hollow-Core Photonic Crystal and Photonic Band-Gap Fiber Hollow-Core Photonic Band-Gap Fiber (PBGF) – Losses as low as 13 d. B/km (small-core), 1. 5 d. B/km (large-core) – Bandwidth >400 nm – Extremely low nonlinearity Solid-Core Photonic Crystal Fiber (PCF) – – – High nonlinearity Losses, a < 1 d. B/km All-silica, dopant-free profile Zero-dispersion wavelength as low as 500 nm Unique dispersion, high birefringence Princeton, June 12, 2007 – 36
Example: Corning – Use of LEAF® fiber for a simple long-haul DWDM transmission system with no in-line dispersion compensation Motivation – System simplicity by eliminating in-line dispersion compensation modules reduces complexity and cost Result – Combined three advanced technologies to produce a simple and flexible LH transmission system requiring no optical dispersion compensation in-line or at Rx NZ-DSF (Corning® LEAF® fiber) Duobinary modulation format Receiver-based EDC Princeton, June 12, 2007 – 37
Example: Corning – Simple and flexible 1500 km transmission system and impact for optical networks • Simple change in system puts a fixed amount (-3360 ps/nm) of optical dispersion compensation in the transmitter. No in-line or post-compensation of dispersion required anywhere. • Reach of 1500 km demonstrated with ~1 d. B margin at longest wavelength. • Signal quality (Q) measurements at any intermediate node in 1500 km link show equal performance with no changes needed in Rx configuration and >4 d. B margin. • Demonstrates flexible system well suited for reconfigurable optical networks. • Cost-effective system architecture enabled by LEAF® fiber, duobinary, and MLSE-EDC Rx. Princeton, June 12, 2007 – 38
Example: Corning – System testing of ultra-low attenuation fiber Vascade® EX 1000 Motivation – Total span loss is the predominant limiting factor in unrepeatered single-span submarine systems. – System designers resort to expensive technologies such as high launch powers, advanced modulation formats, remote optically-pumped amplifiers (ROPAs), and strong FEC to achieve long distances > 300 km. – Lower fiber attenuation may be the most efficient and costeffective application space. Corning’s low-attenuation fiber allows simpler system design with lower cost and/or better performance Princeton, June 12, 2007 – 39
Example: Corning – Characterization results of Vascade® EX 1000 fiber for system testing Measured attenuation of fibers tested Average 1550 nm loss of 0. 169 d. B/km. Typical A 1550 even lower for current fiber. Average reach advantage of low-attenuation Vascade® EX 1000 vs. standard singlemode fiber was ~12%, measured across 4 different system configurations. Translates into extension of an unrepeatered span of up to ~40 km. Princeton, June 12, 2007 – 40
Example: Corning – Semiconductor Optical Amplifier Switches Switch entire data packets Fast switching: 0. 1 – 2 nanoseconds. Inherent gain (~20 d. B) High on-off ratio Low polarization sensitivity (<0. 6 d. B) Low noise figure (<6. 5 d. B) Broadband & WDM friendly (>80 nm) Monolithic integration feasible Future ultra-fast all-optical capability 8 x 40 Gb/s Capacity Discrete SOA Electrically Switched SOA at 1 GHz Optically Switched SOA at 80 GHz 10 -12 BER 20 d. B dynamic range Princeton, June 12, 2007 – 41
Research 3. Software Princeton, June 12, 2007 – 42
Software Challenge – How does one manage 130 Million Lines of Code with Acceptable Error Rates? Problem – We typically modify approximately 20 million lines each year. The industry average is 500 faults per million lines. This rate would make the communications systems in North America inoperable. The highest cost is not writing code but testing that it works – over half the effort! Increasingly, open source components are part of the software, leaving quality is in the hands of third parties. Princeton, June 12, 2007 – 43
Software Challenge – Complexity Megabytes Example: Growth of Size of Linux Kernel Release Date Princeton, June 12, 2007 – 44
Example: Telcordia – STRIDE (STructured Requirements & Interface Design Environment) Managing Development of Complex Software Requirements Modeling & Analysis Automatic generation with checks for: • Consistency • Completeness • Functionality Unified framework for expressing requirements as model components Design Documentation Reduced cost Better quality Code Skeletons Test Coverage Princeton, June 12, 2007 – 45
Example: Telcordia – STRIDE: Improved Quality & Productivity Led to dramatic quality improvement and cost reduction for a major B 2 B clearinghouse application Ø ~25% reduction in costs Ø ~90% reduction in requirements-related defects Led to an order-of-magnitude improvement in quality of Service Management Layer Solution for a major provider Ø Potential of 30% cost reduction for future release Ø Order of magnitude improvement in the quality of overall solution Reduced costs and improved quality of interfaces for a Network Activation OSS STRIDE based test automation Ø Dramatic improvement in coverage Ø Automatic generation of self-checking test cases from test and interface models Princeton, June 12, 2007 – 46
Example: Telcordia – STRIDE: Sample Productivity Improvements Princeton, June 12, 2007 – 47
Example: Telcordia – Impact of our Initial Quality Efforts. . . Faults Per Thousand Function Points Compliance ISO 9001 Certification (Phase I) ISO 9001 Certification (All Business Units) CMM Level 3 1994 1995 1996 1997 Princeton, June 12, 2007 – 48
Faults Per Thousand Function Points Example: Telcordia –. . . And their Long-Term Effects CMM Level 3 2002 Industry Average of 17. 5 Source: ISBSG Report 1/2004 (new product CMM Level 5 suite released) Princeton, June 12, 2007 – 49
End Princeton, June 12, 2007 – 50
82f7979ed69e6c09b16eed5fa586c620.ppt