Fault Tolerant Scheduling of Mixed Criticality Real-Time Tasks under Error Bursts

Abhilash Thekkilakattil, Radu Dobrin and Sasikumar Punnekkat

Real-time tasks

Dependable Hard Real-time Systems

web images

Mixed-criticalityrequirements

Fault tolerance requirements

Timing requirements

Timing requirements

job1 job2

Worst Case Execution Time

Inter-arrival time (Period)

Release time

Real-time scheduling: find a schedule for any given taskset that guarantees all tasks’ completion before the deadline

Relative deadline

Real-time tasks

Dependable Hard Real-time Systems

web images

Mixed-criticalityrequirements

Fault tolerance requirements

Timing requirements

Fault Tolerance Requirements

Temporal redundancy

Fault tolerance related overhead

Fault: Hypothesized cause of an error that may lead to a failure

Spatial redundancy

replica 1

replica 2

voter

Our focus

The Error Burst Model

Increasing variety of error sources● Electromagnetic interference from ubiquitous devices e.g., cellphones, radars and radio sets

… having new characteristics● Occurs continuously over a period of time: error bursts

classical error model(errors as singleton events)

extended error model(errors as bursts of events)

Real-time tasks

Dependable Hard Real-time Systems

web images

Mixed-criticalityrequirements

Fault tolerance requirements

Timing requirements

Mixed Criticalities

Critical(Flight critical functionalities)

Non-critical(Inflight entertainment)

Normal operationAbnormal operation

discard

Taken from toonpool.com

System Model

1. Set of real-time tasks

2. Two levels of criticalities

3. Known error burst length

critical

non-critical

error burst

Problem Statement

How can we guarantee:

1. All critical tasks meet their deadlines under error bursts

2. All non-critical tasks meet their deadlines if they are not hit by error bursts

Overview of the Solution

Work generated in any time interval ≤ size of the interval.

Case 1 : If no critical job is hit by the error burst

Case 2 : When the error burst hits only a single critical job

Case 3 : When the error burst hits more than one critical jobs

Conclusions

• We propose a schedulability analysis for EDF scheduled mixed criticality real-time systems under error bursts

• Enables the use of “normal” EDF scheduler for mixed criticality scheduling

• Use of existing operating systems e.g., ERIKA

• No commercial operating system support for mixed criticality scheduling

• Future work:

• Exploit the use of frequency scaling to control transient overloads

Thank you !

Questions ?