• Specifically, when requesting admission into the system, each sporadic task presents to the scheduler its traffic parameters and the collection of these parameters are called the flow specification of the task in communications literature.
  • They define the constraints on the inter-arrival times and execution times of jobs in the task.
  • The performance guarantee provided by the system to each task is conditional which means that the system delivers the guaranteed performance conditioned on the fact that the task meets the constraints define by its traffic parameters.
  • The flip side is that if the task misbehaves (that is, its actual parameters deviate from its traffic parameters), the performance experienced by it may deteriorate
  • . Different traffic models use different traffic parameters to specify the behavior of a sporadic tasks.

Leaky Bucket Model:

We first introduce the notion of a (Λ, E) leaky bucket filter, then define the leaky bucket model. This kind of fillter is specified by its input rate Λ and size E: The filter can hold at most E tokens at any time and it is filled with tokens at a constant input rate of Λ tokens per unit time. A token is lost if it arrives at the filter if the filter already contains E tokens. It can be assumed of sporadic task that meets the (Λ, E) leaky bucket constraint as if its jobs were generated by the filter in the following manner. The filter may release a job with execution time e when it has at least e tokens. After the release of this job, the e tokens are removed from the filter. A job cannot be released when the filter has no token. Therefore, not any job in a sporadic task that satisfies the (Λ, E) leaky bucket constraint, has execution time larger than E. The total execution time of all the jobs which are released within any time interval of length E/Λ is less than 2E. A periodic task with period equal to or larger than E/ Λ and execution time equal to or less than E satisfies the (Λ, E) leaky bucket constraint. A sporadic task S = (1, p, p’, I) that fits the FeVe model satisfi• es the constraint if p’ = 1/ Λ and E = (1− p/p’)I/p’.

                                 Leaky Bucket traffic model