Modeling, Analysis, and Hard Real-Time Scheduling of Adaptive Streaming Applications

Abstract

In real-time systems, the application's behavior has to be predictable at compile-time to guarantee timing constraints. However, modern streaming applications which exhibit adaptive behavior due to mode switching at run-time, may degrade system predictability due to unknown behavior of the application during mode transitions. Therefore, proper temporal analysis during mode transitions is imperative to preserve system predictability. To this end, in this paper, we initially introduce Mode Aware Data Flow (MADF) which is our new predictable Model of Computation (MoC) to efficiently capture the behavior of adaptive streaming applications. Then, as an important part of the operational semantics of MADF, we propose the Maximum-Overlap Offset (MOO) which is our novel protocol for mode transitions. The main advantage of this transition protocol is that, in contrast to self-timed transition protocols, it avoids timing interference between modes upon mode transitions. As a result, any mode transition can be analyzed independently from the mode transitions that occurred in the past. Based on this transition protocol, we propose a hard real-time analysis as well to guarantee timing constraints by avoiding processor overloading during mode transitions. Therefore, using this protocol, we can derive a lower bound and an upper bound on the earliest starting time of the tasks in the new mode during mode transitions in such a way that hard real-time constraints are respected.

On the implementation and execution of adaptive streaming applications modeled as MADF

“…To address the first challenge and facilitate the analysis of adaptive streaming applications, several parallel models of computation (MoCs), referred as adaptive MoCs in this paper, such as modeaware dataflow (MADF) [1], mode-controlled dataflow (MCDF) [2], and finite-state machine scenario-aware dataflow (FSM-SADF) [3] have been proposed. These adaptive MoCs are able to capture the behavior of an adaptive streaming application as a collection of a finite number of different dataflow graphs, called scenarios or modes, that are controlled by parameters which values need to be updated at run-time to activate different graphs.…”

“…Each scenario or mode is modeled using a static MoC such as synchronous dataflow (SDF) graph [4] or cyclo-static dataflow (CSDF) graph [5] and is individually analyzable in terms of performance and resource usage at design-time. As a result, design-time analyzability of the adaptive streaming applications, modeled with the aforementioned adaptive MoCs, can be provided to some extent, e.g., hard real-time (HRT) analysis [1], worst-case performance analysis [3], etc.…”

“…This protocol, however, introduces timing interference of one mode execution with another one that can significantly affect and fluctuate the latency of an adaptive streaming application across a long sequence of mode transitions. To avoid such undesirable behavior caused by the ST transition protocol, a simple, yet effective transition protocol, called maximum-overlap offset (MOO), is proposed for the MADF model [1]. The MOO protocol can resolve the timing interference between modes upon mode transitions by properly offsetting the starting time of the new mode.…”

“…The MOO protocol can resolve the timing interference between modes upon mode transitions by properly offsetting the starting time of the new mode. In [1], it has been shown that such offset for the MOO protocol can be calculated when using the strictly periodic scheduling (SPS) [6] in each mode to schedule the application tasks.…”

“…This extension enables the applicability of many different schedules to the MADF model, thereby generalizing the MADF model and making MADF schedule-agnostic as long as K-periodic schedules are considered; • We propose a generic parallel implementation and execution approach for adaptive streaming applications modeled with MADF that conforms to the analysis model and its operational semantics [1]. We demonstrate our approach on LITMUS RT [8] which is one of the existing RT extensions of the Linux kernel; • Finally, to demonstrate the practical applicability of our parallel implementation and execution approach and its conformity to the analysis model, we present a case study on a real-life adaptive streaming application.…”