Efficient object sharing in quantum-based real-time systems

Abstract

We consider the problem of implementing shared objects in uniprocessor and multiprocessor real-time systems in which tasks are executed using a scheduling quantum. In most quantum-based systems, the size of the quantum is quite large in comparison to the length of an object call. As a result, most object calls can be expected to execute without preemption. A good object-sharing scheme should optimize for this expected case, while achieving low overhead when preemptions do occur. In this paper, we present several new shared-object algorithms for uniprocessors and multiprocessors that were designed based upon this principle. We also present scheduling analysis results that can be used in conjunction with these algorithms.

On Spin Locks in AUTOSAR: Blocking Analysis of FIFO, Unordered, and Priority-Ordered Spin Locks

“…To avoid long non-preemptable sections, prior work has explored the design and implementation of preemptable spin locks [3,12,19], where busy-waiting jobs can be preempted at any time and only the actual critical sections are executed non-preemptably. The primary benefit of preemptable spinning is that the maximum non-preemptable section length is independent of the number of processors (which, given rising core counts, is highly desirable), albeit at the cost of increased worst-case blocking, which arises because preempted jobs must "reissue" their lock requests after a preemption (e.g., in FIFO-ordered spin locks, preempted jobs must re-queue again at the end of the queue [3,12]). While preemptable spinning may be unavoidable in demanding latencysensitive applications, no analysis of the associated increase in blocking has been proposed to date, which renders such locks unsafe in the context of predictable hard real-time systems.…”

A Fully Preemptive Multiprocessor Semaphore Protocol for Latency-Sensitive Real-Time Applications

“…Under clustered scheduling [4,17], processors are grouped into m c non-overlapping sets (or clusters) of c processors each, where C k denotes the k th cluster. We require uniform cluster sizes and for simplicity assume that m is an integer multiple of c. 2 Each task is statically assigned to a cluster; we let C(T i ) denote the cluster to which task T i has been assigned. Partitioned (c = 1) and global (c = m) scheduling are special cases of clustered scheduling.…”

“…Unfortunately, non-preemptive spinlocks can cause unacceptable delays for latency-sensitive workloads because tasks may remain non-preemptive for prolonged times. While it is possible to reduce the negative effects by allowing spinning tasks to remain preemptive (e.g., see [2,20,35]), critical sections are still executed non-preemptively in such spinlocks and thus have a major impact on latency, not unlike the priority-boosting of critical sections under the MPCP.…”

Schedulability Analysis on Generalized Quantum-Based Fixed Priority Scheduling

“…In [9], Anderson et al defined quantum-based scheduling as follows: "Under quantum-based scheduling, processor time is allocated to tasks in discrete time units called quanta. When a processor is allocated to some task, that task is guaranteed to execute without preemption for q time units, where q is the length of the quantum, or until it terminates, whichever comes first."…”

“…When a processor is allocated to some task, that task is guaranteed to execute without preemption for q time units, where q is the length of the quantum, or until it terminates, whichever comes first." Quantum-based scheduling was applied to dynamic priority scheduling for efficient object sharing [9], flow protection in communication [10], and reduction of the scheduling overhead [11].…”