Reducing Tardiness under Global Scheduling by Splitting Jobs

Abstract

Under current analysis, tardiness bounds applicable to global earliest-deadline-first scheduling and related policies depend on per-task worst-case execution times. By splitting job budgets to create subjobs with shorter periods and worst-case execution times, such bounds can be reduced to near zero for implicit-deadline sporadic task systems. However, doing so will result in more preemptions and could create problems for synchronization protocols. This paper analyzes this tradeoff between theory and practice by presenting an overhead-aware schedulability study pertaining to job splitting. In this study, real overhead data from a scheduler implementation in LITMUS RT was factored into schedulability analysis. This study shows that despite practical issues affecting job splitting, it can still yield substantial reductions in tardiness bounds.

Minimizing response times of automotive dataflows on multicore

“…1, functions in this way. Erickson et al [8] also showed that response time bounds can be further improved by simulating a G-FL schedule of a task system with more frequent releases of smaller jobs. This method is referred to as job splitting.…”

“…We randomly generated DAGs of varying characteristics and tested them for schedulability using the methods described in [8]. We now describe the experimental process we used.…”

“…We made several changes to the overhead-accounting techniques for C-FL as described in [8] to support DAGs under PGM RT . First, we account for non-preemptivity during consumer signaling, as this can delay higher priority work from being scheduled.…”

“…4, with T j i = T zi i , when scheduled with job splitting. The extent of splitting was determined from overhead data (which limits the extent of splitting) using a procedure from [8]. A proportional end-to-end latency metric was computed by dividing the computed the end-to-end latency by p i · (height(T i ) + 1).…”

“…A GEL scheduler uses these PPs as pseudodeadlines. Erickson et al also showed that job splitting may further minimize bounds on response times [8]. Here, the scheduling priority of a job decreases as the job consumes discrete chunks of execution budget, effectively splitting each "real" job into a sequence of sub-jobs.…”