Integrated Modular Avionics (IMA) architectural principles support Responsive Space goals
Common plug-and-play hardware modules and software time and space partitioning support rapid reconfiguration IMA is required to fly for 15 days following first fault condition; reliability of 10-9 chance of undetected failure Acquisition cost of AIMS2 IMA was 75% less than previous federated architecture with >20% shorter development schedule
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Modeling Tool(s)
Application
Run-Time
Application
Application
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Open architecture, common modules for RF Technology drives core processing performance
Integrated, LRM based system Open architecture, common modules for core processing
From a presentation by Ron Szkody on 29 May 1996 to the Integrated Sensor System (ISS) Open System Architecture (OSA) Joint Task Force (sponsored by United States Air Force Wright Laboratory /AAST-30
AIMS
VIA
Primus EpicTM
AIMS-2
2002
AIMS-787 Current
Space Exp
2004
1980s
4. System-level coordination tools Key IMA architectural principles may be included or excluded to meet program needs
Shared Resources
Unique Input/Output
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
...
Generic I/O Board a cPCI Backplane Generic I/O Board a Processor Board Network I/F Board
Processor Board
cPCI Backplane
Power Comm I/O Board Generic I/O Board a Generic I/O Board a Generic I/O Board a Processor Board Network I/F Board
Network Centric
Cabinet Centric
Power
Power Generic I/O Board Generic I/O Board a Generic I/O Board a Generic I/O Board a Processor Board Network I/F Board
...
cPCI Backplane cPCI Backplane GG I/O Board Generic I/O Board a
Cabinet Bus
Processor Board
Other Modules
Lock-Step Processors
Each side duplicates all processing, memory accesses, and I/O functions Compare logic determines when there is a fault on one side Each side enables external communication of the opposite side Upon fault detection, the processor pulls itself offline If a fault is recoverable (e.g. an SEU in one memory can be detected and the correct value loaded from its counterpart on the opposite side), the processor re-inserts itself into operation
DRAM
DRAM
System Memory
System Memory
Processor
Processor
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Combined with robust BIT/BITE, lockstep processors provide 100% fault coverage Does not require a cross-channel voting mechanism, providing cross channel trust Minimizes processing modules required to handle Byzantine failure conditions Provides a degree of self-healing
Lock-step processing enables minimal number of required channels to compensate for fault conditions
9 Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Partitioning Domains
1. 2. 3. 4. Memory Space Computation Time Input/Output Access Backplane Access
10
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
11
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
I/O
1394 1553 Ethernet Analog I/O Discrete I/O RS-422
OS API
OS API
OS API
OS API
OS API
Application Fault Response Time Management Software Loader Platform Load Verification Module Load Verification Cabinet Initialization
Hardware Abstraction Layer (HAL2) Non-Resident Boot Initialization Power-Up BITE (POST) Platform Fault Response
Resident Boot
Boot Initialization
Partitioning allows software running on 1 processor to be treated as if applications were physically separated Table-Driven operations removes system responsibility from the software developers and transfers it to the system integrator
12
I/O and inter-application communication is table-driven and managed by the system software Many system modifications only impact tables values no impact to applications software
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Benefits of Partitioning
Allows software applications to be developed and tested independent of other developers Software changes do not require recertification of unmodified applications
Allows each applications verification level to be based on its criticality. Supports coexistence of applications of differing criticality levels on the same hardware Supports coexistence of applications with differing tendencies to change Significantly reduces integration, test, and V&V activities
13
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Applications Developers
Resource Manager
System Engineers
Physical Implementation
tem s Sy ch Ar
Memory Mapper
ctu e t i
re
Application Scheduler
1 per Processor Bus Bus Access Bus Access Schedule Access Application Schedule Schedule Schedule
System and partitioning tools are required in addition to traditional development tools
14 Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.
Partition A
Memory Code
System Tables System Tables
Middleware
Code
Virtual Backplane
Middleware
I/O Partition
Memory Drivers
Test Script
Inter-Application Comm, I/O
I/O
15
Run-Time
17
Copyright by Honeywell International Inc. Published by AIAA 4th Responsive Space Conference 2006 with permission.