1. Time-Triggered Architecture
A summary
Tim Arrowsmith
2/6/2006

2. TTA- Introduction
Infrastructure/guidelines for partitioning large applications into nearly autonomous subsystems.
Also control the complexity of the evolving system.
Decomposes a large embedded application into clusters and nodes
Provides a FT global time base of known precision at each node
Taking advantage of the global time to simplify communications and ensure timeliness of real-time applications
Tim Arrowsmith

3. TTA – Architecture Model
Broken into 6 sections:
Model of Time
Time and State
RT Entities and RT Images
State Information vs. Event Information
Structure of the TTA
Interconnection Topology
Tim Arrowsmith

4. Model of Time Real time progresses as an infinite set of instants
A happening that occurs at an instant is called an Event
Ordering example:
Node j increments clock
Event e occurs
Node k increments clock
Tim Arrowsmith

5. Model of Time – cont. TTA introduces a sparse time base
Time is partitioned into alternating durations of activity and silence
External representation of time modelled according to the GPS time representation
Time-stamp is an eight-byte integer
Tim Arrowsmith

6. Time and State Sparse-time provides a system-wide notion of time
“Interval of silence” on the sparse time base forms a system-wide consistent dividing line between the past and future and the interval when the state of the distributed system is defined
Tim Arrowsmith

7. RT Entities and RT Images
Dynamics of a real-time application are modeled by a set of relevant state variables, the RT-Entities that change their state as time progresses
State Variable
TT-model
A RT Image is a temporally accurate picture of a RT entity at instant t
Tim Arrowsmith

8. State Information vs. Event Information
State Attribute – and property of a RT entity tha tis observed by a node of the distributed RT at a particular instant.
State Information – corresponding information
State Observation – records the state of a state variable at particular instant
Event – sudden change of state of an RT entity that occurs at and instant
Event information – information that describes an event, difference between the state before and the state after the event
Tim Arrowsmith

9. Structure of the TTA Basic building block of the TTA is a node
Tim Arrowsmith

10. Interconnection Topology
TTA – bus configuration
At every physical node there are three subsystems: the node and two guardians
Tim Arrowsmith

11. Interconnection Topology
TTA – star configuration
In cluster of n node n+2 packages are needed (as opposed to 3n with bus)
Tim Arrowsmith

12. Design Principles Discusses principles that guided TTA design
Divided into 6 sections:
Consistent Distributed Computing Base
Unification of Interfaces
Composability
Scalability
Transparent Implementation of FT
Openness
Tim Arrowsmith

13. Consistent Distributed Computing Base
TTA exploits the short error detection latency of a TT protocol to perfome immediate error detection and distributed agreement membership
Tim Arrowsmith

14. Unification of Interfaces
The time-triggered transport protocol carries autonomously – driven by TT schedule – messages from the sender’s CNI to the receiver’s CNI
Tim Arrowsmith

15. Unification of Interfaces – cont.
An interface that prevents propagation of control errors by design is called a temporal firewall
There are three types of interfaces of a node:
Real-time service (RS)
Diagnostic and Maintenance (DM)
Configuration Planning (CP)
Tim Arrowsmith

16. Composability
Must distinguish between architeture design and node design
Stability-of-prior service principle ensure that the validated service of a node is not refuted by the integration of a node into a system
Tim Arrowsmith

17. Composability – cont.
Constructive integration principle requires that if n nodes are already integrated then the integration of the n+1 node must not disturb the correct operation of the n already integrated nodes
Tim Arrowsmith

18. Composability – cont.
Replica Determinate if all members of this set have the same externally visible state, and produce the same output messages at points in time that are at most an interval of d time units apart
‘d’ is the time it takes to replace a missing message from redundant replicas
Tim Arrowsmith

19. Scalability
TTA is designed for very large distributed real-time applications
Horizontal layering (abstraction)
Vertical layering (partitioning)
Tim Arrowsmith

20. Transparent Implementation of FT
In TTA the FT mechanisms are implemented in a dedicated FT layer
The FT CNI is identical in structure and timing to the basic non-FT CNI
Tim Arrowsmith

21. Openness
“ Provided that the CORBA security clearance is passed, it is thus possible to investigate remotely (via the Internet) the internals of every TTA node while the system is delivering its real-time service.”
Tim Arrowsmith

22. Communication Divided into 4 sections: The TTP/C Protocol
The TTP/A Protocol
Event Message Channels
Performance Limits
Tim Arrowsmith

23. TTP/C Protocol Fault-tolerant time-triggered protocol that provides:
Autonomous FT message transport with know delay and bounded jitter between CNI (via TDMA)
FT clock synchronization, without relying on a central time server
Membership service to inform every node about the “health-state” of every other node
Clique avoidance
Tim Arrowsmith

24. TTP/A Protocol Time-triggered fieldbus protocol of TTA.
Connect low-cost smart transducers to a node of the TTA.
Interface file system (IFS) holds real-time data, calibration data, diagnostic data, and configuration data.
Information between the IFS of the smart transducer and the CNI of the TTA node is exchanged by TTP/A.
TTP/A supports a “plug-and-play” mode.
Tim Arrowsmith

25. Event Message Channels
Event message channels constructed on top of basic TT communications
Bytes designated a priori
Two message queues provided at CNIs:
Sender queue at sender’s CNI
Receiver queue at receiver’s CNI
Filter service and garbage collection service
Tim Arrowsmith

26. Performance Limits Must maintain a 5µs inter-frame gap
Testing currently being perfomed on 1GBit/s systems using COTS
Tim Arrowsmith

27. Fault Tolerance
Fault Hypothesis – it is assumed that a chip is a single fault-containment region.
Fault-Tolerant Units – CNI implements replica determinism, it is up to host software to ensure replica determinism within the complete node. Also supports self-checking pairs.
Never-Give-UP Strategy –
highly application specific.
Redundant Transducers –
uses an agreement protocol.
Tim Arrowsmith

28. TTA Design Methodology
Architecture Design – application decomposed into clusters and nodes.
Node Design – application software for host computers developed. Testing from the bottom-up.
Validation – designed to reduce the validation effort.
Design Tools – supported by a comprehensive set of integrated design tools of TTTech AG
Tim Arrowsmith

29. Conclusion
Guiding principle: take maximum advantage of the availability of global time.
TTA currently occupies a niche position.
The designers hope to broaden as mainstream application designers start to utilize time instead of attempting to dismiss it.
Tim Arrowsmith