|
Most
existing real-time fault-tolerant scheduling algorithms for
heterogeneous distributed systems can achieve high reliability for
non-preemptive and aperiodic tasks. However, the existing scheduling
algorithms assume that status of each backup copy is either active or
passive status. To remedy this deficiency, we propose in this paper a
novel reliability model tailored for preemptive periodic tasks. Next, we
develop two real-time fault-tolerant algorithms (NRFTAHS and RDFTAHS)
for heterogeneous distributed systems. NRFTAHS manages to assign tasks
in a way to improve system schedulabilties, whereas RDFTAHS aims at
boosting system reliability without adding extra hardware. Unlike the
existing scheduling schemes, our algorithms consider backup copies in
both active and passive forms. Therefore, our approaches are more
flexible than the alternative scheduling schemes. Finally, we
quantitatively compare our schemes with two existing algorithms in terms
of performability measured as a function of schedulability and
reliability. Experiments results show that RDFTAHS significantly
outperforms NRFTAHS with respect to reliability with marginal
degradation in schedulability and, therefore, RDFTAHS substantially
improves overall performance over NRFTAH.
|