Packet scheduling for wireless QOS
Queen's University Belfast, 2008
Online
Hochschulschrift
Zugriff:
Since their arrival in the late 90's wireless packet schedulers have sought to seamlessly re-apply the concepts of fair queuing (FQ) algorithms (traditionally employed in wire-mediated communications) to provide fairness, throughput and delay gutUlUltees to individual sessions in the wireless environment. What has militated against a smooth transition in this regard is the presence of unpredictable bursty location-dependent channel errors due to fading on the wireless channel. The approach generally employed by wireless schedulers for minimizing the effects of channel errors is to use a compensation policy. Compensation has traditionally been applied in two -ways: (a) penalizing sessions that have utilized extra bandwidth gained from the loss ofanother session and (b) reserving a fraction ofthe system bandwidth. This dissertation presents a novel scheduling approach that employs multiple sources of compensation so that an emphasis is placed on allowing sessions to recover as quickly as possible from channel errors, while at the same time preserving the guarantees ofother sessions in the system. Two crucial factors are identified that influence how wireless packet scheduling should be performed namely the channel bandwidth and the traffic variation. In order to better understand the dynamics ofthese two driving forces, this dissertation quantifies and analyzes the behavior of each ofthese parameters in typical real-world worst-case scenarios. Through investigating the nature ofthese two aspects ofthe network, a number of conclusions are drawn on how certain wireless scheduling procedures should be implemented. In particular the recording of lost service and the application of compensation are significantly affected by the nature of these two system attributes. It is shown that wireless scheduling algorithms must incorporate far more system knowledge than what was traditionally required for the wired environment. This research proposes a novel traffic profile and priority based wireless fair queuing (TPP-WFQ) scheduling algorithm that is heavily reliant on system functions that perform; bandwidth management, weight adjustment, link adaptation, queue management, traffic classification and traffic monitoring. Several novel attributes are introduced with the TPP-WFQ algorithm, these include: (i) A novel lag/lead calculation model that makes use of knowledge of the traffic profile to accurately measure the amount of service lost and gained for different sessions when an error occurs. (ii) A novel compensation approach that adopts five sources ofcompensation. This allows for a fast recovery time which is the fundamental attribute that has motivated extensive research in the field of wireless scheduling. (iii) A classbased weight adjustment function that manages the weights for all sessions in the system when partial errors occur on channels due to link adaptation. (iv) A priority-based compensation approach that ensures that compensatory resources are prioritized according to the class oftraffic being compensated. This ensures that delay-sensitive sessions recover more quickly than delay-tolerant ones. Detailed simulations incorporating multiple worst-case scenarios illustrate that these areas of novelty enable TPP-WFQ to perform significantly better than selected high ranking schedulers presented in the literature at providing strong QoS to sessions in an error prone environment. In particular, we demonstrate that TPP-WFQ allows delay-sensitive real-time sessions to becom~ more resilient to poor channel conditions.
Titel: |
Packet scheduling for wireless QOS
|
---|---|
Autor/in / Beteiligte Person: | Walsh, S. M. |
Link: | |
Veröffentlichung: | Queen's University Belfast, 2008 |
Medientyp: | Hochschulschrift |
Schlagwort: |
|
Sonstiges: |
|