# Ima Partitioning > Use when sizing an ARINC 653 partition schedule, writing the partition configuration table, or reviewing inter-partition communication ports against frame, period, and latency budgets for certification. Compute and validate ARINC 653 integrated modular avionics (IMA) partition configurations: check partition schedule feasibility by summing the partition durations within the major frame and verifying that each partition receives its period slot, build the partition configuration table from frame and window data, bound sampling port and queuing port message latency for inter-partition communication, and scope the health monitoring responsibilities for fault detection and recovery. Trigger: ARINC 653, IMA, partition scheduling, major frame, MAF, sampling port, queuing port, inter-partition communication, health monitoring. - Skill: `ashfordeou/ima-partitioning` (Agent Skill, multi-file: 3 files) - Install (CLI): `npx skillmds@latest add ashfordeou/ima-partitioning` - Raw SKILL.md: https://api.skillmd.com/api/skills/ashfordeou/ima-partitioning/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - License: Apache-2.0 - Author: ashfordeOU (https://skillmd.com/u/ashfordeou) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/ashfordeou/ima-partitioning --- # ARINC 653 IMA Partitioning (avionics/ima/ima-partitioning) Use when the task is integrated modular avionics (IMA) partitioning for civil avionics: ARINC 653 partition schedule feasibility, the partition configuration table, inter-partition communication through sampling and queuing ports, and health monitoring responsibilities. The module is data-driven: you supply the major frame (MAF) duration and the partition list with window durations and periods, and the functions compute the frame load, the utilization, the slack, and the feasibility verdict, then bound the sampling and queuing port latencies. ## Domain quick reference - ARINC 653 defines partitioning for IMA: applications run isolated from one another in partitions, each with its own memory space and its own schedule windows on the processor. - A partition configuration table assigns each partition a window duration d and a period p inside a repeating major frame (MAF) of duration M; the MAF must be a common multiple of every period so the cyclic schedule repeats evenly (period divides the MAF). - Frame load: each partition with period p executes M / p windows per frame, so the allocated time per frame is the sum of d x (M / p). Worked: MAF 40 ms with partitions of 10 ms at period 20 ms and 15 ms at period 40 ms allocates 20 + 15 = 35 ms, utilization 0.875, slack 5 ms, feasible. - Feasibility: the frame load must not exceed the MAF (equivalently the utilization sum of d / p must be at most 1), and each window duration must fit inside its own period slot. Worked: MAF 40 ms with windows 10, 20, and 15 ms at period 40 ms allocates 45 ms, an over-subscribed frame that is infeasible. - Sampling ports carry the latest message (freshness semantics): a new message overwrites the previous one, and the receiver reads whatever is current. Worst-case latency is one sending period plus the wire transmission time of the message. Worked: 100-byte message at 100 Mbps takes 0.008 ms on the wire; a 10 ms sampling period gives 10.008 ms worst-case latency. - Queuing ports carry messages in FIFO order with a bounded queue depth. Worst-case latency is the queue depth times the sending period plus the transmission time of the final message. Worked: depth 4 at 10 ms with the same message gives 40.008 ms. - Health monitoring responsibilities: each partition monitors its own execution, the module monitors partition states and the schedule, and faults are reported to the health monitor, which runs the configured recovery action for the fault level. - The exact ARINC 653 configuration rules, window arithmetic, port attributes, and health monitor behavior are revision-specific standard data; confirm them against the current revision before freezing a partition configuration table. ## Workflow 1. Collect the MAF duration and, for each partition, its name, window duration, and period. 2. Run schedule_feasibility(maf_ms, partitions): the function validates the inputs, checks every period against the MAF and every duration against its period slot, sums the frame load, and returns the verdict with the utilization and the slack. 3. An infeasible verdict lists the violations: re-tune durations, periods, or the MAF until the frame load fits and every slot is legal. 4. Wrap the accepted schedule in a PartitionSchedule object to keep the frame, load, utilization, slack, and violations together for the configuration record. 5. For each inter-partition communication path, choose the port type: sampling_port_latency_ms for freshness traffic and queuing_port_latency_ms for FIFO traffic, passing the sending period, the message size, the link bit rate, and (for queuing) the queue depth. A message larger than the port capacity raises ValueError. 6. Record the partition configuration table (MAF, per-partition window and period, frame load, slack) and the port latency bounds for the certification package and the integrator. 7. Confirm the health monitoring responsibilities for the module: which faults are detected in each partition, where they are reported, and which recovery action runs for each fault level. ## Pitfalls - Routing this leaf: ARINC 653 partition scheduling and IPC ports are this leaf; a 429 word decode routes to arinc429-protocol, a virtual link bandwidth question routes to arinc664-afdx, and a MIL-STD-1553 command/response bus routes to mil-std-1553. A partition schedule is not a data bus question. - Confusing the MAF with a frame size: the major frame is a time interval (milliseconds) that repeats; it is not a byte count and not a network frame. - A period that does not divide the MAF: the cyclic schedule cannot repeat evenly, so the configuration is invalid even when the load looks small. - A window duration larger than its own period: the partition cannot finish one execution before its next slot starts. - Over-subscription: the frame load must fit the MAF; a set that allocates more than the frame is re-tuned, not force-fitted. - Sampling versus queuing semantics: sampling ports overwrite (latest wins), queuing ports preserve order; applying freshness semantics to a queuing path, or FIFO ordering to a sampling path, mis-states the latency bound. - Treating the configuration as fixed: window arithmetic, port attributes, and health monitor behavior are revision-specific standard data; confirm every value against the current revision before freezing the table. ## Behavior contract (gate 3) The partition schedule feasibility, sampling port latency, queuing port latency, and transmission time helpers are exercised by the gate 3 contract test: scripts/test_ima_partitioning.py against scripts/ima_partitioning_logic.py (stdlib unittest, offline). Run: python3 scripts/test_ima_partitioning.py ## Compliance - Standards referenced, not reproduced: the ARINC 653 text is proprietary (ARINC/SAE ITC) and is not yet in standards-map.yaml; this leaf keys to the closest existing map entry, do-178c (IMA software and robust partitioning are DO-178C topics), listed reference-only. Summary only per standards-map.yaml. - The module implements the schedule arithmetic, port latency, and health monitoring scoping from common engineering practice; no standard table is embedded in the code or this page. - compliance: STANDARDS-REF, gated: false.