Dynamic traffic scheduling and resource reservation algorithms for output-buffered switches
Scheduling algorithms implemented in Internet switches have been dominated by the best-effort and guaranteed service models. Each of these models encompasses the extreme ends of the correlation spectrum between service guarantees and resource utilisation. Recent advancements in adaptive applicati...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2002
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/8688/1/FSKTM_2002_6%20IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Scheduling algorithms implemented in Internet switches have been
dominated by the best-effort and guaranteed service models. Each of these models
encompasses the extreme ends of the correlation spectrum between service
guarantees and resource utilisation. Recent advancements in adaptive applications
have motivated active research in predictive service models and dynamic resource
reservation algorithms. The OCcuPancy_Adjusting (OCP_A) is a scheduling
algorithm focused on the design of the above-mentioned research areas. Previously,
this algorithm has been analysed for a unified resource reservation and scheduling
algorithm while implementing a tail discarding strategy. However, the differentiated
services provided by the OCP _A algorithm can be further enhanced. In this
dissertation, four new algorithms are proposed. Three are extensions of the OCP _A.
The fourth algorithm is an enhanced version of the Virtual Clock (VC) algorithm,
denoted as ACcelErated (ACE) scheduler. The first algorithm is a priority
scheduling algorithm (i.e. known as the M-Tier algorithm) incorporated with a multitier
dynamic resource reservation algorithm. Periodical resource reallocations are
implemented. Thus. enabling each tier's resource utilisation to converge to its desired Quality of Service (QoS) operating point. In addition. the algorithm
integrates a cross-sharing concept of unused resources between the various
hierarchical levels to exemplify the respective QoS sensitivity. In the second
algorithm. a control parameter is integrated into the M-Tier algorithm to ensure
reduction of delay segregation effects towards packet loss sensitive traffic. The third
algorithm, introduces a delay approximation algorithm to justify packet admission.
The fourth algorithm enhances the VC scheduling algorithm. This is performed via
the incorporation of dynamic features in the computation of the VC scheduling tag.
Subsequently, the delay bound limitation of the parameter is eliminated. |
|---|