do297
Use when scoping an IMA platform, planning module acceptance, or laying out development assurance steps for an integrated modular avionics system. Plan the integrated modular avionics (IMA) platform a
SKILL.mdskills/avionics/ima/do297/SKILL.md
GitHubDO-297 IMA Architecture and Acceptance (avionics/ima/do297)
Use when the task is integrated modular avionics (IMA) architecture and acceptance planning for civil avionics in the spirit of DO-297: identify the platform modules, partitions, and shared resources, allocate applications to partitions with CPU, memory, and I/O budgets, check for resource contention, map integrity and availability requirements, and lay out module acceptance and incremental certification evidence. The module is data-driven: you supply the module resources and the application list with integrity and availability requirements, and the functions allocate applications to partitions, compute partition and module budgets, flag over-subscription, and generate deterministic acceptance criteria.
This leaf covers the IMA integration and acceptance process. It is different from avionics/ima/ima-partitioning, which sizes the ARINC 653 partition schedule, major frame, and inter-partition communication ports.
Domain quick reference
- DO-297 (proprietary RTCA guidance; name and paraphrase only here) describes the IMA development and integration approach: a platform of modules hosts multiple applications in isolated partitions, acceptance evidence is gathered at module and application level, and incremental certification lets an accepted module or application be reused with reduced re-verification.
- Platform elements: modules (line replaceable units that host partitions), applications (functions hosted inside partitions), and shared resources (CPU time, memory, I/O ports, buses, power).
- Allocation rule: each application runs in exactly one partition, one or more partitions share a module, and every partition budget must fit inside the module resources in every dimension.
- Contention: when the summed demand of the hosted applications exceeds a module budget dimension (CPU, memory, or I/O), the platform is over-subscribed and the offending application must be flagged.
- Integrity levels: derived from failure-condition severity (catastrophic maps to level A, hazardous to B, major to C, minor to D, no effect to E), consistent with the DO-178C software level approach (referenced only; see standards-map.yaml).
- Availability classes: class 1 functions must remain available on demand, class 2 permits loss of function with a warning, class 3 permits loss of function without a warning.
- Module acceptance: evidence that the module hosts its allocated applications within resource budgets and preserves failure containment between partitions. Incremental certification records the accepted module configuration so later applications reuse it with limited re-acceptance.
IMA architecture model
- Module: name, cpu_units, memory_bytes, io_ports.
- Application: name, integrity (A-E), availability (1-3), cpu_units, memory_bytes, io_ports.
- Partition: hosts one or more applications; the partition budget is the sum of the demands of the applications it hosts.
- Module totals: sum of the demands of every partition on the module.
- Acceptance criteria: deterministic statements covering platform definition, module acceptance testing, resource usage verification, failure containment, availability demonstration, and incremental certification credit.
Workflow
- Identify the platform: list the modules with their CPU, memory, and I/O budgets.
- Identify the applications with their integrity and availability requirements; map failure-condition severity to integrity level.
- Allocate applications to partitions (one partition per application by default, or group applications into shared partitions).
- Compute the partition budgets and the module totals; check every dimension against the module budget.
- Run the contention check: any dimension over budget flags the over-budget application for re-allocation or module upgrade.
- Generate the module acceptance criteria for the incremental certification record.
- Lay out the development assurance steps per integrity level.
Worked example
A module has 100 CPU units, 1,000,000 bytes of memory, and 16 I/O ports. Three applications are hosted:
- FMS: integrity A, availability class 1, 40 CPU, 400,000 bytes, 6 ports.
- ADIRU: integrity B, availability class 1, 30 CPU, 300,000 bytes, 4 ports.
- Display: integrity C, availability class 2, 20 CPU, 200,000 bytes, 3 ports.
Module totals: 40 + 30 + 20 = 90 CPU, 900,000 bytes, 13 ports. Every dimension fits with slack 10 CPU, 100,000 bytes, 3 ports, so the allocation is accepted with no contention. If FMS grows to 60 CPU units the total becomes 110 CPU against a budget of 100, the contention check flags FMS as the over-budget application, and the module must either drop the growth or move an application to another module.
Pitfalls
- Checking applications against the module budget one at a time: the contention check sums the demand of every hosted application per dimension - FMS at 40 CPU fits alone, but with ADIRU and Display the module total is 90, and FMS growing to 60 pushes the total to 110 against the 100 CPU budget, flagging FMS as the over-budget application.
- Verifying only one resource dimension: the module budget must hold in CPU, memory, and I/O ports at once - an allocation that fits 90 of 100 CPU units can still bust the 16-port I/O budget, so every dimension needs its own total-versus-budget check.
- Double-counting partition demand: the module total is the sum of the partition budgets, and each partition budget is the sum of the applications it hosts - an application grouped into a shared partition must not also be counted at the module level on its own.
- Confusing integrity with availability: integrity levels A-E come from failure-condition severity (catastrophic to A, hazardous to B, major to C), while availability classes 1-3 say whether loss of function is permitted and whether it warns - a class 1 display is not automatically level A, and a level D function can still demand class 1 availability.
- Treating module acceptance as a one-time certificate: incremental certification records the accepted module configuration so later applications reuse it with limited re-acceptance - a changed module (an application moving in or a budget growing) changes the accepted configuration and re-opens the acceptance evidence.
- Routing software-lifecycle questions here: DO-297 is the platform integration and acceptance frame (name and paraphrase only - the text is proprietary RTCA guidance); per-application software assurance belongs to the DO-178C leaves and partition scheduling to avionics/ima/ima-partitioning.
Behavior contract (gate 3)
Run the deterministic contract test (stdlib unittest, offline):
python3 scripts/test_do297.py