Seventy essential terms of real-time and embedded systems engineering — from admission control to WCET — explained in plain language. Type to filter instantly, jump by letter, or narrow the list by category. Each entry links to an in-depth article where one exists.
Real-Time Systems Glossary
70 terms
A
- Abstract Interpretation Timing Analysis
- A formal static analysis technique, introduced by Patrick and Radhia Cousot in 1977, that computes safe over-approximations of all possible program states without executing the code. In WCET analysis it is used for value analysis (possible register and memory contents), cache analysis (classifying accesses as always-hit or always-miss using must and may analyses) and pipeline analysis, producing per-block timing bounds that feed a path analysis such as IPET.
Read more: Worst-Case Execution Time (WCET) Analysis: Static, Measurement-Based and Hybrid Methods - Adaptive Mixed Criticality (AMC) Scheduling
- A fixed-priority scheduling scheme and analysis for mixed-criticality systems, proposed by Baruah, Burns and Davis in 2011. The system starts in low-criticality mode. If any high-criticality task executes beyond its low-criticality WCET budget, the system switches to high-criticality mode and stops releasing low-criticality tasks. Response-time tests such as AMC-rtb and AMC-max verify that high-criticality tasks still meet their deadlines across the mode change.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform - Admission Control Scheduling
- A run-time or design-time check that decides whether a new task, reservation or contract can be accepted without jeopardizing the timing guarantees already given to others. It typically applies a schedulability test, such as a utilization bound or response-time analysis, to the existing workload plus the new request. If the test fails, the request is rejected or renegotiated. Linux SCHED_DEADLINE and contract-based frameworks both rely on admission control.
Read more: Contract-Based Scheduling Explained: Making Embedded Software Composable - Aperiodic Server Scheduling
- A scheduling mechanism that serves aperiodic or sporadic requests within a fixed-priority system using a reserved budget C every period T. A polling server checks for pending work at each period start and forfeits its budget if none is waiting. A deferrable server preserves its budget for the whole period, improving responsiveness. A sporadic server replenishes consumed budget one period after it was used, so it can be analyzed like an ordinary periodic task.
Read more: Contract-Based Scheduling Explained: Making Embedded Software Composable - Architecture Analysis & Design Language (AADL) Design
- An SAE standard (AS5506) modeling language for describing the software and hardware architecture of embedded, real-time systems. AADL models represent threads, processes, processors, memories, buses and data flows, annotated with properties such as periods, execution times, deadlines and dispatch protocols. Tools such as OSATE can analyze these models for schedulability, end-to-end flow latency and resource budgets before implementation, and can support code or configuration generation.
Read more: Model-Based Design and Timing: From Simulink and AADL Models to Verified Real-Time Code
B
- Best-Case Execution Time (BCET) Timing Analysis
- The shortest time a piece of code can take to execute on a given hardware platform, over all possible inputs and initial hardware states. BCET is the lower counterpart of WCET. Together they bound execution-time variability, which matters for computing output jitter, analyzing timing anomalies, checking minimum separation constraints, and validating control loops whose performance depends on consistent sampling and actuation delays.
Read more: Worst-Case Execution Time (WCET) Analysis: Static, Measurement-Based and Hybrid Methods - Blocking Scheduling
- The time a task is prevented from running by a lower-priority task, usually because the lower-priority task holds a shared resource such as a mutex, or executes in a non-preemptive section. Blocking appears as the term B_i in response-time analysis. Without a suitable protocol it can be unbounded; priority inheritance, the priority ceiling protocol and the stack resource policy all limit its maximum duration.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol - Busy Period Timing Analysis
- An interval during which the processor is continuously occupied executing work of a given priority level or higher, starting and ending with no such work pending. The level-i busy period that starts at a critical instant contains the worst-case response time of task i. Analyzing the full busy period is necessary when deadlines can exceed periods, because several jobs of the same task may then be pending at once.
Read more: Response-Time Analysis: A Step-by-Step Worked Example
C
- Constant Bandwidth Server (CBS) Scheduling
- A resource reservation algorithm for EDF scheduling, proposed by Abeni and Buttazzo in 1998. Each server has a budget Q and period P, giving bandwidth Q/P. Jobs are scheduled using the server's deadline; when the budget is exhausted, it is recharged and the deadline is postponed by P. This guarantees temporal isolation: a misbehaving task cannot consume more than its reserved share. Linux SCHED_DEADLINE is based on CBS.
Read more: SCHED_DEADLINE and the Constant Bandwidth Server: Resource Reservations in Linux - Context Switch OS & Kernel
- The operation by which a kernel stops running one thread and starts another, saving the outgoing thread's registers, stack pointer and program counter and restoring those of the incoming thread. Its cost includes the switch itself, scheduler decision time, and indirect effects such as cache and TLB pollution. In real-time analysis, context-switch overhead is usually charged to each job, for example by adding twice the switch time to its execution time.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - Contract-Based Scheduling Scheduling
- A scheduling approach in which each application component declares its resource requirements as a contract, typically a minimum budget, period and deadline plus optional ranges for using extra capacity. The framework admits contracts only if all can be guaranteed together, enforces them through resource reservations, and redistributes spare capacity according to importance and weight. It enables independent development and composable integration, and was central to the FRESCOR research project.
Read more: Contract-Based Scheduling Explained: Making Embedded Software Composable - Controller Area Network (CAN) Networking
- A serial bus developed by Bosch and standardized as ISO 11898, widely used in vehicles and industrial systems. CAN uses bitwise arbitration on message identifiers, so the frame with the lowest identifier, and therefore highest priority, wins access to the bus without collision. This makes the bus behave like non-preemptive fixed-priority scheduling, allowing worst-case message response times to be computed. Classical CAN runs up to 1 Mbit/s; CAN FD allows faster data phases.
Read more: Real-Time Networking: CAN, Time-Sensitive Networking (TSN) and Wireless with Guarantees - Cooperative Scheduling Scheduling
- A scheduling model in which a running task keeps the processor until it voluntarily yields, blocks or finishes; the scheduler never preempts it. Cooperative scheduling is simple, has low overhead, and reduces the need for locking, because tasks cannot be interrupted at arbitrary points. Its drawback is that a long-running task delays everything else, so worst-case response times depend on the longest non-yielding section of every task.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - Critical Instant Timing Analysis
- The release time that produces the largest response time for a task. For independent periodic tasks under fixed-priority preemptive scheduling, Liu and Layland showed that a task's critical instant occurs when it is released simultaneously with all higher-priority tasks. Analyzing this single scenario is therefore enough to determine the worst case, which is the basis of response-time analysis. Blocking, jitter and offsets modify the scenario.
Read more: Response-Time Analysis: A Step-by-Step Worked Example
D
- Deadline Scheduling
- The time by which a job must complete. A relative deadline D is measured from the job's release; an absolute deadline is the release time plus D. Deadlines may be implicit (D equals the period), constrained (D at most the period) or arbitrary. They are classified by the consequence of a miss: missing a hard deadline is a failure, a late firm result is useless but rare misses are tolerated, and a late soft result loses some value.
Read more: What Is a Real-Time Embedded System? Hard, Firm and Soft Deadlines Explained - Deadline Monotonic Scheduling (DM) Scheduling
- A fixed-priority assignment in which tasks with shorter relative deadlines receive higher priorities. Leung and Whitehead showed in 1982 that deadline monotonic is optimal among fixed-priority assignments for independent tasks whose deadlines are less than or equal to their periods. When every deadline equals its period, it reduces to rate monotonic scheduling. Schedulability is normally verified with response-time analysis.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Dhall Effect Scheduling
- A phenomenon identified by Dhall and Liu in 1978 showing that global EDF or rate monotonic scheduling on m processors can miss deadlines for task sets with total utilization only slightly above 1. A classic example combines m very light tasks with short periods and one heavy task with a long period: the light tasks occupy every processor first and the heavy task misses its deadline. It motivated partitioned and hybrid priority schemes.
Read more: Multicore Real-Time Scheduling: Partitioned, Global and Semi-Partitioned Approaches - DO-178C Safety & Certification
- "Software Considerations in Airborne Systems and Equipment Certification," published by RTCA in 2011 and by EUROCAE as ED-12C. It is the primary guidance used for certifying airborne software. Software is assigned a Design Assurance Level from A (catastrophic failure effect) to E, which determines the objectives to satisfy. Verification includes demonstrating timing behavior, such as worst-case execution time and stack usage. Supplements cover tools, model-based development, object-oriented technology and formal methods.
Read more: Timing in Flight Control: How UAV Autopilots Meet Their Deadlines - Dynamic Voltage and Frequency Scaling (DVFS) Architecture
- A power-management technique that lowers a processor's supply voltage and clock frequency when full performance is not needed. Because dynamic power grows roughly with the square of voltage times frequency, running slower at lower voltage can reduce energy per operation. In real-time systems, execution times stretch at lower frequency, so speed choices must keep the task set schedulable. Frequency transitions also have latency that analysis must account for.
Read more: Energy-Aware Real-Time Scheduling: DVFS, Sleep States and Deadlines
E
- Earliest Deadline First (EDF) Scheduling
- A dynamic-priority scheduling algorithm that always runs the ready job with the nearest absolute deadline. On a single preemptive processor, EDF is optimal: if any algorithm can meet all deadlines, EDF can. For independent periodic tasks with implicit deadlines, a set is schedulable under EDF exactly when total utilization is at most 1. EDF needs deadline tracking at run time and can behave unpredictably under overload.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison
F
- Feasibility Timing Analysis
- The property that a task set can be scheduled so that all deadlines are met by at least one scheduling algorithm. Feasibility is a property of the workload and platform, while schedulability refers to a specific algorithm. For example, any set of independent periodic tasks with implicit deadlines and total utilization at most 1 is feasible on one processor, because EDF can schedule it.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - FIRST Scheduling Framework (FSF) Scheduling
- An application programming interface and framework for contract-based flexible scheduling, developed in the EU-funded FIRST research project that preceded FRESCOR. FSF lets applications negotiate contracts specifying a minimum budget, maximum period and deadline, plus ranges for using spare capacity, and implements them on top of fixed-priority or EDF schedulers using resource reservations. The FRESCOR project later built on this work, extending contracts to additional resources.
Read more: Contract-Based Scheduling Explained: Making Embedded Software Composable - Fixed-Priority Preemptive Scheduling (FPPS) Scheduling
- A scheduling policy in which each task has a static priority and the highest-priority ready task always runs, preempting lower-priority tasks immediately when it becomes ready. It is the default policy of most RTOSes and of POSIX SCHED_FIFO. Priorities are usually assigned with rate monotonic or deadline monotonic rules, and schedulability is checked using utilization bounds or exact response-time analysis.
Read more: Response-Time Analysis: A Step-by-Step Worked Example
G
- Generalized Precision Time Protocol (gPTP) Networking
- A profile of the IEEE 1588 Precision Time Protocol, defined in IEEE 802.1AS, that synchronizes clocks across a Time-Sensitive Networking domain. A grandmaster clock is elected, and every link measures its propagation delay using peer-to-peer delay measurements, allowing sub-microsecond synchronization in typical deployments. A common time base is a prerequisite for TSN features such as the time-aware shaper defined in IEEE 802.1Qbv.
Read more: Real-Time Networking: CAN, Time-Sensitive Networking (TSN) and Wireless with Guarantees - Global Scheduling Scheduling
- A multiprocessor scheduling approach in which all tasks share a single ready queue and any job can run on any processor, migrating between cores as needed. Global EDF and global fixed-priority scheduling are common examples. Global scheduling balances load automatically and can make better use of idle cores, but migration overheads, cache effects and phenomena such as the Dhall effect make worst-case analysis more difficult than in partitioned scheduling.
Read more: Multicore Real-Time Scheduling: Partitioned, Global and Semi-Partitioned Approaches
H
- Hyperbolic Bound Timing Analysis
- A sufficient schedulability test for rate monotonic scheduling, published by Bini, Buttazzo and Buttazzo, stating that n independent periodic tasks with implicit deadlines are schedulable if the product of (Ui + 1) over all tasks is at most 2. It is less pessimistic than the Liu and Layland bound and equally cheap to compute, accepting every task set the classic bound accepts and some it rejects.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Hyperperiod Timing Analysis
- The least common multiple of the periods of all tasks in a periodic task set. After one hyperperiod, a synchronous periodic schedule repeats exactly, so simulating or checking it over that interval is sufficient for many analyses and for building static cyclic schedules. Hyperperiods can grow very large when periods are not harmonic, which is one reason designers often choose harmonic periods.
Read more: Response-Time Analysis: A Step-by-Step Worked Example - Hypervisor Architecture
- Software that creates and manages virtual machines, allowing several operating systems to share one hardware platform. Type 1 hypervisors run directly on hardware; type 2 hypervisors run on a host operating system. In embedded systems, partitioning hypervisors are used to isolate an RTOS from a general-purpose OS or to consolidate functions of different criticality, providing spatial and temporal isolation, though shared caches and memory buses can still cause interference.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform
I
- IEC 61508 Safety & Certification
- The International Electrotechnical Commission's basic standard for the functional safety of electrical, electronic and programmable electronic safety-related systems. It defines a lifecycle from hazard analysis to decommissioning and four Safety Integrity Levels, SIL 1 to SIL 4, that set requirements for risk reduction, development techniques and verification. Many sector standards, including ISO 26262 for road vehicles and IEC 62304 for medical device software, derive from it.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform - IEEE 802.1Qbv (TAS) Networking
- The Time-Sensitive Networking amendment that defines the time-aware shaper (TAS), now incorporated into IEEE 802.1Q. Each egress port has a gate control list that opens and closes the transmission gates of its traffic-class queues according to a repeating schedule synchronized across the network. Reserving exclusive windows for scheduled traffic gives bounded latency and very low jitter. Guard bands keep lower-priority frames from overrunning a protected window.
Read more: Real-Time Networking: CAN, Time-Sensitive Networking (TSN) and Wireless with Guarantees - Implicit Path Enumeration Technique (IPET) Timing Analysis
- A path analysis method for static WCET estimation, introduced by Li and Malik. The program's control-flow graph is described with integer variables representing how often each basic block or edge executes, constrained by flow conservation, loop bounds and other flow facts. Maximizing the sum of execution counts multiplied by per-block time bounds, typically with integer linear programming, yields the WCET bound without enumerating paths explicitly.
Read more: Worst-Case Execution Time (WCET) Analysis: Static, Measurement-Based and Hybrid Methods - Interrupt Latency OS & Kernel
- The delay between a hardware device asserting an interrupt and the processor beginning to execute the corresponding interrupt service routine. It includes hardware recognition time, completion or abandonment of the current instruction, the longest period during which interrupts are disabled, and context saving. Worst-case interrupt latency, rather than average latency, matters for real-time guarantees, and it is a key figure when comparing RTOSes and real-time Linux configurations.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - Interrupt Service Routine (ISR) OS & Kernel
- A function invoked by the processor in response to a hardware or software interrupt. ISRs run outside normal task scheduling, typically at a priority above all tasks, so their execution time directly delays application tasks and must be included in schedulability analysis. Good practice is to keep ISRs short and defer longer processing to a task or deferred handler; PREEMPT_RT Linux runs most handlers as schedulable kernel threads.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - ISO 26262 Safety & Certification
- The international functional safety standard for electrical and electronic systems in road vehicles, adapted from IEC 61508. First published in 2011, with a second edition in 2018. Hazards are classified into Automotive Safety Integrity Levels, ASIL A to ASIL D, which set requirements for development, verification and analysis. Part 6 covers software and includes requirements relevant to timing, such as freedom from interference between software elements.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform
J
- Jitter Timing Analysis
- Variation in the timing of a recurring event relative to its ideal time. Release jitter is the variation between when a job should be released and when it actually becomes ready; output or completion jitter is the variation in when results are produced. Jitter matters in control systems, where it degrades stability and performance, and in networks, and it must be included explicitly in response-time analysis.
Read more: What Is a Real-Time Embedded System? Hard, Firm and Soft Deadlines Explained
L
- Least Laxity First (LLF) Scheduling
- A dynamic-priority scheduling algorithm that runs the job with the smallest laxity, defined as the time to its absolute deadline minus its remaining execution time. Like EDF, it is optimal on a single preemptive processor. In practice it is rarely used, because laxities change continuously as jobs execute, and jobs with equal laxity can cause frequent context switches, which adds overhead and requires knowing remaining execution times.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Liu and Layland Bound Timing Analysis
- The utilization bound published by C. L. Liu and James Layland in 1973: n independent, preemptible periodic tasks with deadlines equal to periods are schedulable under rate monotonic scheduling if U = Σ Ci/Ti ≤ n(2^(1/n) − 1). The bound decreases toward ln 2, about 0.693, as n grows. It is sufficient but not necessary; task sets above it may still be schedulable, which exact response-time analysis can confirm.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Logical Execution Time (LET) Design
- A programming abstraction, introduced with the Giotto language, in which a task reads its inputs at the start of a fixed logical interval and publishes its outputs exactly at the end, regardless of when it actually executes within the interval. LET makes input and output timing deterministic and independent of scheduling or platform, simplifying composition and multicore migration at the cost of added latency. It has been adopted in automotive software architectures.
Read more: Model-Based Design and Timing: From Simulink and AADL Models to Verified Real-Time Code
M
- Measurement-Based Probabilistic Timing Analysis (MBPTA) Timing Analysis
- A timing analysis approach that collects execution-time measurements and applies extreme value theory to estimate a probabilistic WCET: a curve giving the probability that an execution exceeds any given time, such as one in 10^15 runs. Its statistical validity relies on independence and identical distribution of measurements, which is commonly achieved with time-randomized hardware, such as random cache placement and replacement, or software randomization.
Read more: Worst-Case Execution Time (WCET) Analysis: Static, Measurement-Based and Hybrid Methods - Mixed Criticality Architecture
- A system property in which software components of different criticality levels, for example flight control and cabin infotainment, share the same hardware platform. In the model introduced by Vestal in 2007, tasks have multiple WCET estimates, with more conservative values at higher criticality levels. The goal is to guarantee high-criticality tasks under pessimistic assumptions while using resources efficiently for lower-criticality tasks in normal operation.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform - Model-Based Design Design
- An engineering approach in which system behavior and architecture are captured in executable or analyzable models, such as Simulink, SCADE or AADL, that drive simulation, verification and automatic code generation. For real-time systems, models can carry timing information, such as sample rates, periods and execution budgets, allowing timing problems to be found before hardware exists. Preserving model semantics in generated multitask code requires careful scheduling and communication design.
Read more: Model-Based Design and Timing: From Simulink and AADL Models to Verified Real-Time Code - Mutex OS & Kernel
- A mutual-exclusion lock that ensures only one thread at a time enters a critical section protecting shared data. Unlike a general semaphore, a mutex has an owner, and only the owner may release it, which allows the kernel to implement priority inheritance or priority ceiling protocols to bound priority inversion. Mutexes may block the waiting thread, making them unsuitable for use inside most interrupt service routines.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol
P
- Partitioned Scheduling Scheduling
- A multiprocessor scheduling approach in which each task is statically assigned to one processor and each processor runs its own uniprocessor scheduler, such as fixed-priority or EDF, with no migration. Well-understood uniprocessor analysis applies directly to each core. The main challenge is allocating tasks to cores, which resembles bin packing, so some task sets with total utilization well below the processor count cannot be partitioned.
Read more: Multicore Real-Time Scheduling: Partitioned, Global and Semi-Partitioned Approaches - Period Scheduling
- For a periodic task, the fixed interval T between successive job releases, such as 10 ms for a 100 Hz control loop. For a sporadic task, T denotes the minimum inter-arrival time between releases. Periods, together with execution times, determine processor utilization, and deadlines are often set equal to periods. Choosing harmonic periods, where each divides the next, simplifies schedules and improves schedulability under rate monotonic.
Read more: What Is a Real-Time Embedded System? Hard, Firm and Soft Deadlines Explained - PREEMPT_RT OS & Kernel
- A set of Linux kernel changes that make the kernel almost fully preemptible to reduce worst-case latency. It converts most spinlocks into sleeping locks with priority inheritance, runs interrupt handlers in kernel threads, and minimizes non-preemptible sections. After many years as an external patch set, PREEMPT_RT was merged into the mainline kernel in version 6.12 in 2024. It enables millisecond and often sub-100-microsecond worst-case latencies on suitable hardware.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - Priority Ceiling Protocol (PCP) Scheduling
- A resource access protocol for fixed-priority systems, published by Sha, Rajkumar and Lehoczky in 1990. Each resource has a ceiling equal to the highest priority of any task that uses it. A task may lock a resource only if its priority exceeds the ceilings of all resources currently locked by other tasks. PCP prevents deadlock and chained blocking, so a task is blocked by at most one lower-priority critical section.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol - Priority Inheritance Protocol (PIP) Scheduling
- A resource access protocol in which a task holding a lock temporarily inherits the priority of the highest-priority task it blocks, until it releases the lock. This prevents medium-priority tasks from preempting the lock holder and extending priority inversion indefinitely. PIP is widely supported, including in POSIX and Linux futexes, but it does not prevent deadlock and allows chained blocking across several resources.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol - Priority Inversion Scheduling
- A situation in which a high-priority task waits for a lower-priority task, typically because the latter holds a shared resource. Bounded inversion is inherent to resource sharing; unbounded inversion occurs when medium-priority tasks preempt the lock holder and prolong the wait. The 1997 Mars Pathfinder resets are a well-known example. Priority inheritance and priority ceiling protocols bound the inversion so it can be included in analysis.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol
Q
- Quality of Service (QoS) Architecture
- A measure of how well a system delivers a service beyond simple correctness, such as frame rate, resolution, latency, throughput or control accuracy. In flexible real-time systems, applications may operate at several QoS levels, each needing different resources. Contract-based frameworks allow a component to declare a minimum acceptable QoS and use spare resources to improve it, degrading gracefully when demand increases.
Read more: Real-Time Embedded Vision: Timing Guarantees for Video Analytics at the Edge
R
- Race-to-Idle Architecture
- An energy-management strategy that executes work at the highest available speed so the processor can finish quickly and enter a deep low-power sleep state for longer. It contrasts with DVFS-based slowdown. Race-to-idle can save more energy when static or leakage power is significant, sleep states are very efficient, or peripherals consume power while active. The best choice depends on the platform's power characteristics and sleep transition costs.
Read more: Energy-Aware Real-Time Scheduling: DVFS, Sleep States and Deadlines - Rate Monotonic Scheduling (RMS) Scheduling
- A fixed-priority assignment in which tasks with shorter periods receive higher priorities. Liu and Layland proved in 1973 that rate monotonic is optimal among fixed-priority assignments for independent, preemptible periodic tasks with deadlines equal to periods. Schedulability can be checked with the Liu and Layland bound, the hyperbolic bound, or exact response-time analysis. Its simplicity and predictable overload behavior make it very common in practice.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Real-Time Operating System (RTOS) OS & Kernel
- An operating system designed to provide predictable, bounded timing for application tasks. An RTOS typically offers priority-based preemptive scheduling, low and bounded interrupt and scheduling latencies, timers, and synchronization primitives with priority inversion control. Examples include Zephyr, FreeRTOS, NuttX and RTEMS, while PREEMPT_RT turns Linux into a real-time capable system. Determinism, rather than raw speed, is the defining property.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared - Resource Reservation Scheduling
- A scheduling technique that guarantees a task or group a specified share of a resource, such as processor budget Q every period P, and prevents it from exceeding that share. Reservations provide temporal isolation, so timing faults or overloads in one component do not affect others. They are implemented by server algorithms such as the constant bandwidth server and underpin contract-based scheduling and Linux SCHED_DEADLINE.
Read more: SCHED_DEADLINE and the Constant Bandwidth Server: Resource Reservations in Linux - Response Time Timing Analysis
- The interval between a job's release and its completion. It includes the job's own execution, interference from higher-priority work, blocking by lower-priority tasks, and system overheads. The worst-case response time of a task is the largest response time over all its jobs; a task meets its deadlines if its worst-case response time does not exceed its relative deadline.
Read more: Response-Time Analysis: A Step-by-Step Worked Example - Response-Time Analysis (RTA) Timing Analysis
- An exact schedulability test for fixed-priority preemptive scheduling, developed by Joseph and Pandya and extended by Audsley and colleagues. The worst-case response time of task i is the smallest solution of R = Ci + Bi + Σ ⌈R/Tj⌉ Cj over higher-priority tasks j, found by fixed-point iteration starting from R = Ci. The task is schedulable if R does not exceed its deadline. Extensions handle jitter, overheads and arbitrary deadlines.
Read more: Response-Time Analysis: A Step-by-Step Worked Example
S
- SCHED_DEADLINE OS & Kernel
- A Linux scheduling class, available since kernel 3.14, that implements EDF combined with the constant bandwidth server. Each task specifies runtime, deadline and period parameters through the sched_setattr() system call. The kernel performs admission control, rejecting tasks whose total bandwidth would exceed a configured limit, and throttles tasks that exceed their runtime, providing temporal isolation. SCHED_DEADLINE tasks run ahead of SCHED_FIFO and SCHED_RR tasks.
Read more: SCHED_DEADLINE and the Constant Bandwidth Server: Resource Reservations in Linux - Schedulability Timing Analysis
- The property that a task set meets all its deadlines when scheduled by a specific algorithm on a given platform. Schedulability tests may be sufficient, such as utilization bounds, which guarantee success when passed but may reject schedulable sets, or exact, such as response-time analysis, which accept precisely the schedulable sets under the model's assumptions. Schedulability is distinct from feasibility, which concerns whether any algorithm can succeed.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison - Semaphore OS & Kernel
- A synchronization primitive, introduced by Edsger Dijkstra, consisting of a counter with atomic wait and signal operations. A counting semaphore manages a pool of identical resources; a binary semaphore takes only the values 0 and 1 and is often used to signal events from an interrupt to a task. Because semaphores generally have no owner, many RTOSes cannot apply priority inheritance to them, so mutexes are preferred for protecting shared data.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol - Spare Capacity Scheduling
- Processor or other resource capacity left after guaranteeing the minimum requirements of all admitted tasks or contracts, together with capacity reclaimed when jobs finish earlier than their worst case. Flexible scheduling frameworks redistribute spare capacity to components that can use it to improve quality of service, often in proportion to declared importance and weight, while preserving every component's minimum guarantee.
Read more: Contract-Based Scheduling Explained: Making Embedded Software Composable - Spinlock OS & Kernel
- A lock in which a waiting thread repeatedly polls the lock variable, busy-waiting until it becomes free, instead of being suspended. Spinlocks are useful on multicore systems for very short critical sections, where suspending and resuming would cost more than waiting. For real-time analysis, spin time must be bounded, which requires fair or FIFO-ordered locks and short, non-preemptive critical sections, as in multiprocessor protocols such as MSRP.
Read more: Multicore Real-Time Scheduling: Partitioned, Global and Semi-Partitioned Approaches - Stack Resource Policy (SRP) Scheduling
- A resource access protocol proposed by Theodore Baker in 1991 that works with both fixed-priority and EDF scheduling. Each task has a static preemption level, and each resource has a ceiling. A job may start executing only when its preemption level exceeds the current system ceiling. As a result, a job blocks at most once, before it starts, deadlock is impossible, and tasks can share a single run-time stack.
Read more: Priority Inversion, Priority Inheritance and the Priority Ceiling Protocol - System Tick OS & Kernel
- A periodic timer interrupt, commonly at 100 Hz to 1 kHz, that a kernel uses to keep time, expire timeouts and trigger time-slice scheduling. The tick resolution limits timing granularity and causes regular wake-ups. A tickless kernel instead programs the hardware timer for the next pending event, avoiding unnecessary interrupts and enabling longer sleep. Zephyr, FreeRTOS tickless idle and Linux NO_HZ modes offer tickless operation.
Read more: Energy-Aware Real-Time Scheduling: DVFS, Sleep States and Deadlines
T
- Time Partitioning Architecture
- A technique that divides processor time into fixed, pre-scheduled windows assigned to separate partitions, guaranteeing each partition its time regardless of what others do. ARINC 653, the avionics application software interface standard used in Integrated Modular Avionics, defines a repeating major time frame split into partition windows, along with spatial partitioning of memory. Time partitioning supports incremental certification and the safe integration of software of different criticality.
Read more: Mixed-Criticality Systems: Running Safety-Critical and Best-Effort Software on One Platform - Time-Sensitive Networking (TSN) Networking
- A set of IEEE 802.1 standards that add deterministic, low-latency communication to standard Ethernet. Key elements include clock synchronization (IEEE 802.1AS), traffic shaping and scheduling such as the time-aware shaper (802.1Qbv) and credit-based shaper, frame preemption, redundancy for reliability (802.1CB) and centralized configuration. TSN allows time-critical control traffic and best-effort traffic to share one network in industrial, automotive and professional media systems.
Read more: Real-Time Networking: CAN, Time-Sensitive Networking (TSN) and Wireless with Guarantees - Time-Slotted Channel Hopping (TSCH) Networking
- A medium access mode of IEEE 802.15.4, introduced in the 802.15.4e amendment, for low-power wireless networks requiring reliability and predictable timing. Time is divided into slots grouped into repeating slotframes, and each scheduled transmission is assigned a slot and channel offset. The physical channel changes from slot to slot, reducing the effect of interference and multipath fading. The IETF 6TiSCH work runs IPv6 over TSCH.
Read more: Real-Time Networking: CAN, Time-Sensitive Networking (TSN) and Wireless with Guarantees - Timer OS & Kernel
- A hardware or software facility that measures time and triggers actions at specified instants or intervals. Hardware timers count clock cycles and raise interrupts; kernel software timers build on them to provide timeouts, delays and periodic activation of tasks. Timer resolution, accuracy and the latency between expiry and task activation directly affect release jitter, so they are important parameters in real-time system design.
Read more: Choosing an RTOS in 2026: Zephyr, FreeRTOS, NuttX, RTEMS and PREEMPT_RT Linux Compared
U
- Utilization Scheduling
- The fraction of processor time a task requires, computed as its execution time divided by its period, Ui = Ci/Ti. Total utilization U is the sum over all tasks. A uniprocessor task set with U above 1 cannot be scheduled. Utilization bounds, such as U ≤ 1 for EDF with implicit deadlines or the Liu and Layland bound for rate monotonic, provide quick schedulability tests.
Read more: Rate Monotonic vs Earliest Deadline First: A Practical Comparison
W
- Watchdog Timer Safety & Certification
- A hardware or software timer that resets or places a system in a safe state if it is not periodically serviced, or "kicked," within a set time. It detects software hangs, deadlocks and runaway code. Effective use requires servicing the watchdog only after checking that critical tasks are actually making progress. Windowed watchdogs also flag servicing that happens too early, which can indicate faulty control flow.
Read more: Timing in Flight Control: How UAV Autopilots Meet Their Deadlines - Worst-Case Execution Time (WCET) Timing Analysis
- The longest time a piece of code can take to execute on a specific hardware platform, over all possible inputs and initial hardware states, without interruptions. WCET is a key input to schedulability analysis. It is estimated with static analysis, measurements or hybrid methods, and a safe bound must not underestimate the true value. Caches, pipelines and multicore interference make tight, safe bounds difficult to obtain.
Read more: Worst-Case Execution Time (WCET) Analysis: Static, Measurement-Based and Hybrid Methods
No matching terms. Try a shorter query.