System Design 1 - Chapter 6

20 June 2025, Bastian Luettig

Consolidation Management

1. in the first block (above broadcast), there is the individual single computer-internal data
2. all single computers perform a broadcast of the individual information and now each single computer lane has the information from each other computer lane (be aware: a lane might not tell the truth)
3. each computer lane creates a voting / reference value $y_{\text{ref}}$, performs monitoring between the individual values and creates a monitoring failure indication
4. each computer lane $o$ creates fail state information
5. each computer lane $o$ creates not avail state information
6. each computer lane $o$ checks, if the yref-value was obtained from sufficiently valid sensors

Simmilar diagram for discrete sensors is available on slide 55 onward, chapter 6

In case of correct sensors, discrete sensors have $A_\in$ when and analog sensors have $A_\Delta$ after data acquisition by drivers

Resource Management

Based on the failure categories set by the consolidation management, the resource manager will passivate sensors

• Input Consensus is required to fulfill output consensus among redundant computers
• Sensor types are categorized into simple and complex (bus-based)
• Single sensors follow discrepancy models (analog: $A_\Delta$ and discrete: $A_\in$)
• Input Consolidation Management handles availability and failure detection at the individual computing lane
• Cross-Sensor Consolidation ensures correctness by comparing multiple single sensors
• If at least two sensors are valid, valid-voting is performed; otherwise, avail-voting. If no sensors are available, a default value is returned
• Faulty sensors are passivated if their values deviate too far from the reference value
• The PLAMA sensor ensures $A_\Delta$ for analog sensors and $A_=$ for discrete sensors