|
Title
|
|
|
|
Evaluation of a strict priority scheduler, and cross-layer resource allocation
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| This thesis covers two parts: (I) the analysis of the end-to-end delay and delay variation of a strict priority scheduler for a particular combination of traffic inputs, and (II) cross-layer allocation of resources in shared systems. In part (I) we develop expressions for the end-to-end (E2E) delay and delay variation distributions, for the different classes of aggregate traffic that are served by a strict priority scheduler. This can help to dimension the network and ensure the QoS is not violated. We characterize the busy period, taking low priority traffic into account, of all priorities in order to calculate the additional delay each traffic class encounters. In particular, we characterize the busy period of any aggregate of constant bit-rate sources. We use those distributions to obtain the E2E delay bound, taking the through-traffic and cross-traffic (CT) into account. Methods used in literature are usually limited to two priority classes or do not account for the through-traffic. We evaluate our approach using simulation of a network, and find that our expressions are able to upper bound the E2E delay and delay variation for all the considered traffic priorities. In part (II) we look at cross-layer resource allocation, in which information is shared beyond the usual OSI layers, in order to increase performance. The research presented here applies a novel cross-layer scheduler, called the minimal delay violation (MDV) scheduler, that can be used when the achievable data rate of a user depends on the data rates the other users receive. We apply the MDV scheduler in a DSL context, where cross-talk between the copper cables of users reduces the maximum simultaneously achievable data rates, and long-term evolution (LTE) and 5G contexts where data rates can vary strongly from moment to moment. Through simulations, we show good performance of our scheduler, with respect to the delay and throughput. We additionally implement and discuss an algorithm to constrain the service rates in these settings. Finally, one other cross-layer allocation algorithm is developed that can be used when service rates are assigned dynamically, but there is a large delay between requesting and receiving the data rate, such as for example in satellite communication networks. |
| |
|
Language
|
|
|
|
English
| |
|
Publication
|
|
|
|
Antwerpen
:
Universiteit Antwerpen, Faculteit Wetenschappen
,
2022
| |
|
Volume/pages
|
|
|
|
xviii, 232 p.
| |
|
Note
|
|
|
|
:
Blondia, Chris [Supervisor]
:
Moonen, Marc [Supervisor]
| |
|
Full text (open access)
|
|
|
|
| |
|