Reconfigurable Computing for Telecommunications

1
Reconfigurable Computing for Telecommunications

• William D. Bishop
• wdbishop_at_computer.org
• Wayne M. Loucks
• wmloucks_at_pads.uwaterloo.ca

2
Presentation Outline

• Introduction to reconfigurable computing
• Definition of a reconfigurable computer
• Characteristics
• Advantages and disadvantages
• Niche applications
• Active research at UW
• Goals
• Reconfigurable coprocessor platform
• Prototype application
• Potential applications in telecommunications
• Voice Services
• Data and Multimedia Services
• Wireless Services
• Conclusions

3
Introduction to Reconfigurable Computing

• Definition of a reconfigurable computer
• Characteristics of reconfigurable computers
• Flexible control logic
• Flexible datapaths
• In-system reconfigurability
• Unlimited reconfiguration cycles

A reconfigurable computer is a computing machine
that incorporates programmable logic devices to
create a hardware architecture that may be
modified at runtime.
4
Classification of Reprogrammable Computers
5
Computing Hierarchies
6
Summary of Reconfigurable Computing

• Advantages
• Flexible architecture
• Improved performance
• Easily upgraded
• Additional control logic capacity
• Niche applications
• Image processing Athanas, 1995
• Cryptography Vuillemin, 1996
• Hardware emulation Dubois, 1995

• Disadvantages
• Hardware complexity
• Software complexity
• Additional development cost

7
Active Research at UW

• Goals
• Investigate the potential value of reconfigurable
  coprocessors for non-niche software applications
• Identify mainstream computing tasks that can
  benefit from reconfigurable computing
• Develop a framework for a reconfigurable
  coprocessor system
• Design and implement a library of reconfigurable
  hardware objects to accelerate mainstream
  computing tasks
• Analyze the quantitative and qualitative benefits
  of a reconfigurable coprocessor

8
Reconfigurable Coprocessor Platform
Logic Analyzer Connections
SIMMs
ARC-PCI Board
Altera 10K50 User CPLDs
Altera 10K50 Controller CPLD
9
Prototype Application

• Acquired the source code for CSIM, a commercial
  software package licensed by Mesquite Software,
  Inc.
• CSIM is a process-oriented, discrete-event
  simulation library
• Applications of CSIM include the following
• Modelling queuing systems
• Simulating ASICs (Application Specific Integrated
  Circuits)
• Any system that may be modelled using a finite
  set of communicating processes may be simulated
  using CSIM
• Designed and implemented a reconfigurable
  component for pseudo-random number generation
  within CSIM
• Several other reconfigurable components are
  planned for CSIM

10
Enhanced CSIM System Overview
11
Reconfigurable Coprocessor System Components
12
Potential Applications for Reconfigurable
Computing in Telecommunications

• Voice Services
• Suitable for voice processing at end-points
• Signal processing may be added (voice disguise,
  filtering, QoS monitoring, etc.)
• Private branch exchange hardware may be upgraded
  remotely
• Data and Multimedia Services
• Suitable for data processing at end-points
• Network interface cards may have simplified
  control logic
• New functionality may be added to network
  components
• Intelligent switch designs may require fewer
  ASICs
• Wireless Services
• Suitable for base stations
• Base station hardware may be upgraded remotely
• New functionality and communication protocols may
  be incorporated
• One chip may implement different protocols at
  different times

Provided that the users will not perceive a drop
in the quality of service, reconfigurable
computing is a viable option for
telecommunications equipment
13
Conclusions

• Possible to completely hide the use of a
  reconfigurable component from the end-user
• Reconfigurable computing is not suitable for
  simple tasks due to reconfiguration latency ( 10
  ms) and communication overhead (implementation
  dependent)
• Only barriers to the use of reconfigurable
  computing for telecommunications equipment are
  reconfiguration latency and high power consumption

14
Acknowledgements

• This research would not have been possible
  without the support of the following
  organizations
• Natural Sciences and Engineering Research Council
  of Canada
• Altera Corporation
• Mesquite Software, Inc.