14-06-2012, 05:37 PM
TRIPLE MODULAR REDUNDANCY WITH STANDBY (TMRSB)SUPPORTING DYNAMIC RESOURCE RECONFIGURATION
Types of Standby Systems
Hot Standby
Keeps primary and backup components running simultaneously with the backup tracking the primary system in real time
Cold Standby
A secondary component is only called upon when the primary component fails
Warm Standby
Periodically mirrors the primary component which means that there are times when both components do not contain the exact same data
Previous TMR-based Fault Tolerance Techniques for FPGAs
TMR + Bitstream scrubbing
The voting mechanism identifies the faulty configuration based and reconfigures (scrubs) the device with an alternative bitstream
Single failure assumption
Triple Module Redundancy Design Techniques for Virtex FPGAs
Different types of data structures such as Throughput Logic, State-Machine Logic and I/O Logic are illustrated in terms of a TMR technique
Conclusions
TMRSB Model:
New system model interpretation
Standby configurations do not require distinct physical resources
Provides an interesting resource cost vs. reliability tradeoff
Reliability of Standby-Based systems:
May not increase linearly in terms of the number of standby configurations
Reliability of the configurations which are both active and standby becomes an essential factor determining overall system reliability