In this paper, we derive a utilization bound on
schedulability of apriodic tasks with arbitrary arrival times,
execution times, and deadlines. Earlier utilization
bounds considered only periodic and sporadic tasks.
To the author's knowledge, this is the first time
a utilization bound is derived for the aperiodic task
model. We prove that the optimal bound is 5/8. Our
result is an extension of the well-known Liu and
Layland's bound of ln 2 (derived for periodic tasks).
The new bound is shown to be tight. Our findings are
especially useful for an emerging category of soft
real-time applications, such as online trading and
e-commerce, where task (request) arrival times are
arbitrary, task service times are unknown, and service
has to be performed within a given deadline. Our result
provides theoretical grounds for guaranteeing deadlines
of individual aperiodic requests by observing only the
aggregate utilization conditions.
