Network calculus: a theory of deterministic queuing systems for the internet
Network calculus: a theory of deterministic queuing systems for the internet
| Hi-index | 0.00 |
A shaper is a system that stores incoming bits in a buffer anddelivers them as early as possible, while forcing the output to beconstrained with a given arrival curve. A shaper is time invariantif the traffic constraint is defined by a fixed arrival curve; itis time varying if the condition on the output is given by a timevarying traffic contract. This occurs, for example, withrenegotiable variable bit rate (RVBR) services. We focus on theclass of time varying shapers called time varying leaky bucketshapers; such shapers are defined by a fixed numbers of leakybuckets, whose parameters (rate and bucket size) are changed atspecific transition moments. We assume that the bucket levels arekept unchanged at those transition moments (no reset assumption).Our main finding is an input-output characterisation for this classof time varying shapers. Then we apply it to the tradeoff inoptimising the RVBR service, assuming that a perfect prediction offuture traffic can be made. We provide two algorithms that solvethe problem of finding, at any renegotiation, the parameters for aRVBR service, respectively when the knowledge of the input trafficis limited to the next interval (local optimisation problem) andwhen we dispose of the complete input traffic description (globaloptimisation problem). We compare, by means of simulation, the tworesulting algorithms to study the validity of the local approach.We illustrate the impact of the no-reset assumption by analyzing onsome examples the losses that occur when the source chooses theopposite approach, namely, the reset approach. Furthermore wesimulate the RVBR service versus the renegotiable constant bit rate(RCBR) service and illustrate that the RVBR approach can providesubstantial benefits. Finally, we discuss the impact of the size ofthe renegotiation interval on the efficiency of the RVBR service.